Telemetry from this host is read in Grafana Cloud -- section 7 already verifies
by querying Cloud, and nothing uses the host's own Tempo, Prometheus, Grafana or
Loki. The base compose file defines all of them, so following section 5 as
written starts five services nobody looks at. Found running for seven days on
the box, started by an earlier session rather than by a deploy.
Records that the collector alone is what this host needs, while keeping the
existing warning that both compose files must be used whatever is brought up --
the base alone exports nothing to Cloud and logs no error. Notes the local
validation harness as the one consumer that does need the full stack.
CI's clang-tidy job failed on this branch. clang-tidy itself succeeded; the
step that failed was the gate that fails the job when findings exist, and the
uploaded diff named exactly two missing includes.
Event.h calls notify(std::uint64_t) but never included <cstdint>, relying on
it arriving transitively. This is pre-existing on the branch rather than new,
surfaced now because clang-tidy only inspects changed files and this one has
not been touched since.
InboundTransactions.cpp uses uint256 throughout and had no direct include for
it either. The round-request work added further direct uses, which is what
brought the file into clang-tidy's changed-file set.
Both fixes are the ones clang-tidy generated itself, applied with the repo's
angle-bracket include style; the include-style and clang-format hooks accept
the placement unchanged.
Note on why this was not caught before pushing: TIDY=1 clang-tidy was run
locally and passed, but against the compile database under
~/sourceCode/.clangdcache/<worktree>/, which was stale -- generated before
these edits. A stale database makes a local clang-tidy pass meaningless, as
project-map.md warns. CI holds the only current one.
Commit c5829df68f added a per-round event to the txset.acquire span but
updated only three of the surfaces that describe that span. Fills the rest.
The operator glossary gains the round-naming behaviour, and loses a claim it
should not have made: it described a tx-set fetch as being on the critical
path of a round. It is not. acquireTxSet returns immediately and the round
continues without the set; a set that arrives after the round has accepted is
discarded. The cost is a round resolving its disputes on less information,
not a round held up. Two later sentences restated the same claim and are
reworded to match.
The Tempo config records that the new event attributes are queryable through
event:name / event.<attr> on the pinned 2.9.4, but that a dedicated parquet
column accepts only span or resource scope, so those filters scan rather than
index and no event-scoped column should be added.
The collector config adds them to its deliberately-not-promoted list, with
the reason that matters: they are per-round and so unbounded, but more
fundamentally spanmetrics derives metrics from spans and not from span
events, so an event attribute cannot be a metric dimension at all. The
Grafana Cloud collector only cross-references that note, so it is unchanged.
No dashboard change: this adds no metric, so there is no panel to add and no
new filter dimension.
Verification: both YAML files parse; pre-commit passes on the changed files.
A tx-set fetch carried no key tying it to the consensus round that needed
the set, so attributing a stalled fetch to a round meant guessing from
timestamps. One fetch is wanted by many rounds -- it is keyed by set hash,
survives the round sweep, and the round never blocks on it -- so a single
parent, link or attribute cannot describe the relationship.
Instead the fetch span records one timestamped event per requesting round,
carrying the round's parent-ledger hash and the ledger it is building. Both
attribute keys already existed in the shared telemetry namespace with
exactly this meaning, and the existing addEvent API is used as-is, so no
new telemetry surface is added and the whole feature compiles out with
telemetry disabled.
The event fires once per round rather than once per peer proposal, keyed on
the round's parent-ledger hash: that distinguishes rounds started on
different forks at the same height, which a ledger-height compare cannot.
A mid-round wrong-ledger recovery re-enters consensus without re-caching the
round identity, so a fetch begun after that switch is attributed to the
pre-switch round; the limitation is documented where the values are cached.
Also fixes the fetch span's end time, which depended on when the C++ object
was destroyed. Three of the four exits that stop pursuing a fetch -- the set
arriving from elsewhere, the round sweep, and shutdown -- ended the span
only via the destructor, so the recorded duration included however long any
reference happened to be held. Each exit now ends the span itself, plus
cancel() and container teardown, and the destructor asserts the span is
already closed rather than closing it: a fallback that can never legitimately
fire should fail loudly instead of hiding a missed exit. An abandoned fetch
also no longer asks peers for a set nobody wants, which previously led to
charging those peers for answering our own request.
Verification: pre-commit and TIDY=1 clang-tidy pass; levelization is
unchanged. NOT compiled -- the branch is blocked by a gcc-15 internal
compiler error in the unrelated xrpl.libxrpl.rdb unity translation unit.
Runtime behaviour is unasserted: xrpl_tests links only xrpl.libxrpl, so
TransactionAcquire is unreachable from GTest; the added tests cover the new
span-name and attribute constants only.
The workload harness gates regressions on histogram_quantile over
job_queued_us / job_running_us, so re-cutting the microsecond ladder changes
what those queries return and the stored baselines no longer describe the
same measurement.
baseline-timings.json's job.acceptLedger.queued.p95 was 96.79us, which is
0.95 / 0.9926 x 100 -- the old 100us bucket edge scaled by the quantile, with
99.3% of samples beneath it. It was never a latency. Keeping it would make the
gate LESS sensitive rather than more: a genuine regression from a real 40us to
90us would still sit under 96.79us + 50% and pass.
Removes the four job.* entries and records why, including their values. The
comparer reports a metric absent from the baseline as "new metric (not in
baseline)" and skips it, so the span baselines stay live and gating continues
for everything unaffected. is_placeholder() still returns False, so this does
not disable the gate wholesale. Recapture the job.* numbers on a node running
the re-cut ladder.
Also corrects _bucket_note in regression-thresholds.json. It described the
spanmetrics ladder as 15 edges starting at 1ms; the collector config has 20,
including five sub-millisecond edges. The note's own reasoning was void too --
it justified the 10ms absolute span bound as "~2 low-end bucket widths", but
the low-end bucket width is 0.01ms, not 5ms. The bound is kept and justified
on the band where span quantiles actually sit, rather than on a derivation
from a ladder that no longer exists.
The microsecond ladder's first edge was 100us, which sat ABOVE the mass of
every instrument using it. Measured on devnet: 99.3% of job_queued_us
samples, 92.5% of job_running_us and 90.4% of getobject_lookup_us fell in
that first bucket. histogram_quantile then interpolated inside bucket 0 and
returned `quantile / fraction_in_bucket_0 x first_edge` -- p75/p95/p99 of
job_queued_us read 75.52/95.66/99.69us against a prediction of
75.53/95.67/99.70. Three-decimal agreement: those panels were reporting
arithmetic on the bucket edge, not latency.
The fix was already half-written. kSubMillisecondBoundaries had been parked
in MetricsRegistry.cpp as [[maybe_unused]] with a comment noting exactly this
problem for nodestore reads. Its edges are now folded into kMicrosecondBuckets
rather than deleted, so the parked intent is carried forward: 1..1000us
resolution where the mass is, upper edges unchanged so multi-second stalls
stay measurable.
Also moves the GetObject count and charge ladders into HistogramBuckets.h, so
all five ladders have one owner and one set of invariant tests (29 now).
Adds check_bucket_parity.py, wired into the existing OTel naming workflow.
The C++ millisecond ladder and the collector's spanmetrics ladder are
specified to agree over their shared range; they were identical when shipped,
then the collector side alone was extended and nothing noticed for eleven
phases. The check asserts containment rather than equality, because jobs
outlive spans -- jobq_updatepaths averages ~60s, which no span approaches, so
demanding equality would force a ceiling that censors it. Verified it rejects
a missing collector edge, a bogus in-range edge, and a return to the 5s
ceiling.
ledger-data-sync's "Job Queue Wait p95 By Type" moves off the beast
jobq_*_q_milliseconds pair onto job_queued_us filtered by job_type. Those
beast metrics are ms-quantised at the source (Event rounds up to a whole
millisecond), so 94-100% of their samples sat in the first bucket and no
ladder change could fix them. Note the label values are camelCase
(job_type="ledgerData"), not the lowercase metric-name fragments.
Both histogram-fed alert thresholds re-validated and left unchanged, with the
measured basis recorded so neither gets tuned against the old artefact: only
0.0022% of job_queued_us samples exceed the 1s threshold, and every edge
bracketing the 1000ms ios_latency threshold survived the ladder change.
Docs: the rpc_size "known issue -- tracked separately" notes in the runbook
and 09-data-collection-reference are now resolved notes, the stale 10-edge
span_duration bucket list is corrected to the collector's real 20, and the
runbook gains a "Reading A Histogram Percentile" section covering both
saturation traps and the expected discontinuity after a ladder change.
This is the change that actually lifts the 5 s ceiling. Until now the
millisecond ladder and the Unit type existed but nothing consumed them.
Telemetry.cpp registered ONE histogram view: instrument name pattern "*",
unit exactly "ms", boundaries {1, 5, ..., 1000, 5000}. Verified against the
installed SDK, "*" matches every name and "ms" matches exactly, so that view
governed every beast::insight Event -- all 54 of them, whatever they measure.
Measured on devnet: 24.9% of rpc_size samples and 100% of jobq_updatepaths
samples fell above 5000. A quantile landing in the `+Inf` bucket reads back
as the second-highest edge, so those p95s reported a flat 5000 rather than a
measurement, and the 1 s to 5 s span was a single four-second-wide bucket
that any quantile inside it had to interpolate across.
Replaces it with one view per unit, keyed on the unit an instrument declares:
- `ms` gets kMillisecondBuckets: every representable edge of the collector's
spanmetrics ladder, plus 60 s and 120 s. The extensions are deliberate --
jobq_updatepaths was measured averaging 59,956 ms, which no span
approaches, so parity alone would still censor it.
- `By` gets kByteBuckets, placed from the measured response distribution
(mean 2131 B, half under 1 kB, tail mean bounded at 7538 B).
OTelEventImpl now derives its declared unit AND its description from unit()
instead of hardcoding "Duration in ms"/"ms", so rpc_size exports as
rpc_size_bytes on the byte ladder. rpc-pathfinding's "RPC Response Size"
panel follows the rename; its unit was already decbytes and is now truthful.
Also corrects Phase7_taskList.md, which still specified the 5000 ladder as
"matching SpanMetrics". That was true when written and became false when the
collector ladder was extended on its own -- implementing the plan as written
reproduced the bug, so the spec is where the defect had come to live. The
edges now have exactly one owner and the plan points at it.
beast::insight::Event documents itself as carrying "a millisecond time, or
other integral value", but both backends assumed the first case: the OTel
bridge declared every instrument with unit `ms` and StatsD tagged every
sample `|ms`. One Event does not measure time -- ServerHandler's "size"
records the serialized RPC response length -- so it exported as
rpc_size_milliseconds and inherited the millisecond bucket ladder. A quarter
of its samples landed above that ladder's top edge, and since Prometheus
returns the second-highest edge for a quantile in the `+Inf` bucket, its p95
panel showed a flat 5.00 kB rather than a measurement.
Adds beast::insight::Unit (Millis, Bytes) plus otelUnitCode(), carried on
EventImpl and selectable at makeEvent(). Naming the unit at creation is what
lets a backend pick the export unit and, through it, the bucket ladder.
- Collector gains a virtual makeEvent(name, Unit) whose default delegates to
the millisecond overload, so a collector that cannot act on a unit keeps
working unchanged. NullCollector and the Groups wrapper override it.
- The Groups override matters most: call sites reach a collector through a
Group, so forwarding only the prefixed name would silently drop the unit.
A test covers that hop specifically.
- Event gains notify(std::uint64_t) for non-duration samples, replacing
ServerHandler's `Event::value_type{response.size()}` -- wrapping a byte
count in a std::chrono::milliseconds compiles but reads as a duration to
everything downstream.
- EventImpl::value_type stays std::chrono::milliseconds. Widening it would
change the wire value of every existing StatsD timer, and metrics needing
finer resolution use the OTel-native microsecond instruments.
The StatsD collector deliberately keeps emitting `|ms`: that path is retired
here (its UDP port is commented out of the compose file and the integration
test fails if anything listens on 8125), so changing its wire format would
alter a legacy contract with no consumer and no way to verify it.
The exported name does not change yet -- OTelEventImpl still hardcodes its
unit. That follows with the unit-keyed histogram views.
The bucket edges for the OTel histograms lived as file-local `namespace {}`
constants, unreachable from any test, and they drifted from the collector's
spanmetrics ladder they were specified to match. The millisecond ladder
stayed capped at 5 s after the collector side was extended to 30 s, so any
quantile above 5 s read back as a flat 5000 -- Prometheus returns the
second-highest edge for a quantile in the `+Inf` bucket, which looks like a
measurement rather than an error.
Adds include/xrpl/telemetry/HistogramBuckets.h as the single owner of the
ladders, with a constexpr validator plus static_asserts so a descending or
duplicated edge cannot compile, and gtest coverage that pins the floor and
ceiling against the measured distributions:
- kMillisecondBuckets carries every representable collector edge and extends
to 120 s, because the updatepaths job type averages ~60 s and a 30 s
ceiling would censor it exactly as 5 s does today. Sub-millisecond
collector edges are omitted: beast::insight::Event rounds durations up to
whole milliseconds, so they would collect nothing.
- kByteBuckets is new, for Events whose samples are sizes rather than
durations. Edges follow the measured RPC response distribution (mean
2131 B, half under 1 kB, tail mean bounded at 7538 B) rather than a guess,
so the resolution sits between 512 B and 64 kB.
No behaviour change yet -- nothing consumes the header until the views are
rewired.
Match the other dashboards: `rgb(15, 122, 102)` instead of the brighter
`rgb(25, 158, 112)`, still clear of the separation floors against the JMeter grey.
Use `rgb(15, 122, 102)` instead of `rgb(25, 158, 112)`: the brighter step drew
too much attention for a background band.
This is the darkest teal that still separates from the JMeter grey by a readable
margin -- normal-vision dE 15.6 against a floor of 15, CVD dE 12.3 against a
target of 8, and at least 3:1 on the dark surface. Dimmer steps fail: rgb(25,
100, 90) lands at dE 9.5, and a grey-derived rgb(25, 70, 70) at dE 5.8, which is
indistinguishable from the JMeter grey even with full colour vision.
Both checks selected on {job="xrpld"}. Loki's OTLP ingestion promotes
service.name to the label `service_name` and keeps a `job` attribute as
structured metadata, which a stream selector cannot match, so the selector
returned zero streams whatever had been ingested. The collector config and
TESTING.md already say to select on `service_name`.
Invert the cross-reference. Picking an arbitrary trace from Tempo and
expecting it in Loki fails even when correlation works, because a log line
carries a trace_id only when emitted inside a sampled span and most spans
log nothing at `warning` level. Start from a logged trace_id instead and
resolve it in Tempo, which is the invariant worth asserting, and try every
id found so one unexported trace does not fail the check.
Bound the log queries in time. Nothing here set start/end, so every query
relied on Loki's one-hour default and returned nothing when re-run later to
investigate a result.
Match the other dashboards: restore `rgb(70, 70, 70)` on `Perf Runs (JMeter)`
so regions that rendered before keep their colour. `Perf Runs (Locust)` stays
aqua.
The driver split changed the existing perf-run regions from grey to violet,
which was not asked for. Restore `rgb(70, 70, 70)` on `Perf Runs (JMeter)` so
every region that rendered before keeps its colour; `Perf Runs (Locust)` stays
aqua, since it is new.
Grey separates from aqua well (dE 22.8 deutan, 25.9 tritan, 26.1 normal), but it
sits at 1.98:1 against the dark-theme surface, below the 3:1 floor, so its region
edges read faint there. Noted in the runbook.
Apply the same two-layer split as the other dashboards to the sync-health board,
which is introduced on this branch: `Perf Runs (JMeter)` and
`Perf Runs (Locust)`, each matching ["perf-iac", "<driver>"] with
matchAny:false, in place of the single generic `perf-iac` layer.
A single "Annotate perf-iac runs" layer matched only `perf-iac`, so a Locust
load window was indistinguishable from a JMeter one. perf-iac now tags every
region with its load driver, so each driver can have its own layer and colour.
- Replace that layer with `Perf Runs (JMeter)` and `Perf Runs (Locust)`, each
matching ["perf-iac", "<driver>"] with matchAny:false, on 12 dashboards.
- job-queue, ledger-data-sync and log-derived-insights had an empty annotations
list and drew no perf regions at all; they now carry the builtIn layer plus
both driver layers.
- Grafana tag matching is a superset AND with no negation, so a generic
`perf-iac` layer also matches every driver region. Keeping one alongside the
driver layers would draw each load window twice, so it is replaced, not kept.
- Document the layers in the telemetry runbook, including two rendering limits:
annotations draw only on timeseries, state-timeline and candlestick panels,
and the shaded fill is 10% opacity so the region edges carry the colour.
- Add `jmeter` to the cspell dictionary; the hook rejects the bare word.
clang-tidy runs misc-include-cleaner with WarningsAsErrors, so a symbol
reached only transitively fails CI. Add the direct includes for JLOG,
beast::Journal, StartUpType, TokenType, toBase58, std::exception and
std::size_t.
JLOG is defined in xrpl/basics/Log.h, and libxrpl.beast cannot include
xrpl.basics -- basics depends on beast, not the reverse. Use the journal
stream idiom the rest of the file already uses.
detachCallbacks() flips a flag that each observable callback checks on entry,
which leaves a callback already past that check running while the twelve
service stops below it tear down the state it reads. Stop the provider at the
same point instead: that joins the reader thread, so once it returns no
callback is running and none can start. Metrics recorded during the remaining
shutdown steps are no longer exported, which is the cost of the guarantee.
Build metricsRegistry_ in the member-init list rather than assigning it in
setup(). getMetricsRegistry() is read from the job queue and io threads, which
are running by then, so the later assignment was an unsynchronised write to the
handle those reads follow.
beast::insight instruments are created during ApplicationImp's member-init
list, and opentelemetry-cpp 1.28 never rebinds an already-vended Meter, so an
instrument created before the MeterProvider is published records nothing for
the rest of the process. Observable instruments carry the opposite constraint:
registering one arms the SDK reader thread, and its callbacks run hook handlers
that read services which do not exist that early.
Publish the provider in Telemetry's constructor, ahead of every producer, and
defer only the observables. Collector gains onCollectionReady() and
onCollectionStopping(); OTelCollector arms and disarms its gauges in response.
StatsDCollector starts its polling thread in its own constructor and had the
same hazard, so it uses the pair to gate that thread.
The metrics resource carries service.instance.id and is immutable once built,
so the node public key is resolved in Main.cpp, where a config error can still
be reported, and passed to makeApplication(). getNodeIdentity() remains
authoritative; both paths now share readNodeIdentity(), so telemetry cannot
report a key the node has abandoned.
An explicit ~ApplicationImp stops observing and stops telemetry, covering the
setup() failure paths that never reach run(). Telemetry::stop() is once-only
and no longer clears another instance's global pointer. The histogram view's
meter selector now matches the meter actually in use, so its bucket boundaries
apply for the first time.
makeTelemetrySetup() rejects a contradictory [telemetry] mutual-TLS
setup by throwing, but it is called from ApplicationImp's
member-initializer list. A try/catch in the constructor body cannot
reach a throw from there, and nothing further up the stack caught it
either, so a config mistake reached std::terminate: the default handler
printed a terminate dump and raised SIGABRT, leaving a core file
instead of a startup error.
Catch std::exception around makeApplication() in run(), report the
reason on stderr and return -1, so the failure is a clean non-zero exit
with a message an operator can act on. Only the construction is
wrapped. setup() starts subsystems whose shutdown order is delicate and
is left outside deliberately, because unwinding a half-started
Application would skip the normal stop sequence.
Gate both validation guards on enabled. A node with telemetry switched
off previously refused to start over certificate paths that nothing
would read.
Document both throws on makeTelemetrySetup(), state in
cfg/xrpld-example.cfg and the configuration reference that a partial
mutual-TLS setup is fatal and that the checks apply only when
enabled=1, and add a runbook troubleshooting entry keyed on the two
error messages.
Tests cover both guards with the message asserted so the two are told
apart, both enabled=0 paths, and the default plaintext configuration.
Companion to the same fix on phase-9. These two uses exist only on this
branch, so they survived the merge-forward: develop moved TempDir from
beast:: to xrpl:: and deleted xrpl/beast/utility/temp_dir.h.
Database.cpp already includes xrpl/basics/FileUtilities.h and already
spells TempDir unqualified at line 288, so only the qualifier was wrong.
develop moved TempDir from beast:: to xrpl::, deleting
include/xrpl/beast/utility/temp_dir.h in favour of
include/xrpl/basics/FileUtilities.h. The merge kept this branch's
references to the old API, so the tree no longer compiled: the missing
header is a fatal include, and because DatabaseConfig_test.cpp lands in
the xrpld unity blob it broke the xrpld target itself, not just the tests.
Swap the include and drop the stale beast:: qualifier on 17 uses. The
three src/tests/libxrpl/nodestore files already include FileUtilities.h
and already spell TempDir unqualified elsewhere, so only the qualifier
was wrong there. DatabaseConfig_test.cpp needed the include as well; it
sits in namespace xrpl::node_store, so unqualified TempDir resolves to
xrpl::TempDir through the enclosing namespace.
Two further uses exist only on the sync-diagnostics tip and are fixed
there rather than here.
Node identity reached the OTel resource only as service.instance.id, which is
config-overridable and carries a deployment-chosen label rather than the node's
own identity. Add xrpl.node.id, set unconditionally from the node public key
(base58, TokenType::NodePublic), so traces and metrics share a stable per-node
key independent of [telemetry] service_instance_id.
Set on the tracer resource via Telemetry::setNodeId(), called from
ApplicationImp::setup() once nodeIdentity_ is known, and on the MetricsRegistry
resource via an added start() parameter. The beast::insight meter provider is
built in TelemetryImpl's constructor, before the wallet DB exists, so its
resource cannot carry the value; that path is left for later and the attribute
is omitted rather than stamped blank.
Also drops the transform/spanidentity collector processor added in
4a361a496d: per-node identity belongs on the resource, not copied onto every
span.
Consensus spans share one deterministic, ledger-derived trace_id, so a
single trace holds spans from every node and the resource-level node id is
not a reliable per-span discriminator in stored traces.
Add transform/spanidentity to both collector configs, copying
service.instance.id onto every span as service_instance_id so TraceQL can
filter per node with the same value the $node dashboard variable already
uses on the metrics side. Wired into the traces pipeline locally and into
traces/store (after tail_sampling) on the Grafana Cloud variant.
The collector bind-mounts one log root, data/logs, and identifies each node by
the subdirectory inside it. The second instance wrote to data2/logs, which is not
mounted at all, so its logs have never reached the pipeline under any name --
while its metrics and traces flowed normally, making it look like a quiet node
rather than an uncollected one. Its own config comment already claimed the
directory was "the same dir the compose stack bind-mounts into the collector",
so the intent was the single root and only the path was wrong.
Both nodes now log under data/logs. The nodestore stays split across data/ and
data2/, which is what those directories are for; the logs gain nothing from the
split because both sit on the same disk, and lose collection entirely.
The installer's log-directory check was wrong in the same way: it verified a path
under the host's /var/log/xrpld, which nothing reads -- that path exists only
inside the container, where the root is mounted. It now checks the directory the
collector actually reads, and the docs no longer ask for host symlinks that have
no effect.
Also drop the trailing slash from the data ignore rules and add data2. A
trailing-slash pattern matches only real directories, so on a host that follows
the runbook and symlinks both to a fast disk, neither was ignored -- leaving them
permanently untracked in a tree the runbook says should read clean, and one
`git add` away from committing a nodestore.
A node's service_instance_id should say which box the data came from, so a
dashboard can tell two otherwise-identical instances apart. A machine name is
also exactly what this public repository should not carry, and the value cannot
come from the environment: it is read only from the [telemetry] section, and the
config parser has no include directive.
So the tracked configs keep a generic identity and name no host, and the
installer renders each into a .host.cfg beside it with the id substituted from
.env.devbox. The units run the rendered copies, which are gitignored. The
tracked configs are never edited on the host, so an update cannot conflict and a
rebuild loses nothing.
The identity is substituted in two places from one value -- the setting and the
log directory name -- because they have to agree: the collector derives identity
for the logs pipeline from the log path, so a mismatch costs that node's logs
their service_instance_id label while its metrics keep theirs, which reads as
"no logs" rather than as a misconfiguration. The installer counts the
occurrences it expects to replace and verifies the result, so a config reshuffle
fails loudly instead of yielding a copy that quietly kept the generic identity;
on the dashboards that would look like the node had disappeared. It also rejects
an id that is not a safe directory name, and rejects two nodes sharing one id.
Also record what the last deployment needed and the runbook did not say: the
session-bus variable a rootless container runtime needs before its user units
will install, the link from the unit's expected binary path to the preset's
build directory, and that a config merge which redeclares service.extensions
drops the cloud authenticator and stops the collector exporting anything at all.
These files also had not been through the formatting hooks, which want 4-space
indentation and reflowed tables; that is fixed here too.
The collector exits at startup whenever the Grafana Cloud stack is brought up:
cannot start pipelines: failed to start "otlphttp/grafanacloud" exporter:
failed to resolve authenticator "basicauth/grafanacloud": authenticator not found
The developer stack passes otel-collector-filestorage.yaml as a second --config to
persist filelog read offsets. The collector replaces lists on merge rather than
appending them, which that file notes, and it repeats the BASE config's
service.extensions -- [health_check, file_storage/filelog]. Layer the cloud
overlay on top and the base config is swapped for the cloud one, whose list also
carries basicauth/grafanacloud; the storage overlay still comes last, so the
authenticator is dropped and the exporter that needs it cannot start.
Nothing degrades: the collector refuses to start at all, so a host brought up this
way exports no telemetry anywhere, local or cloud.
Adds a cloud variant of the overlay naming all three extensions, mounted by
docker-compose.grafanacloud.yaml over the base one so the command line is
unchanged. The base variant stays as it is for stacks that do not use the cloud
exporter.
Following the previous version on a freshly provisioned host failed three times,
each in a way that pointed away from the actual cause. Replacing the build section
with what actually works, and recording why each wrong turn misleads.
The daemon is not built by default: the Conan recipe defaults its xrpld option to
off. Omit the option and both Conan and CMake report success, then the build fails
with "No rule to make target 'xrpld'" because the target was never created.
Configure through the Conan-generated preset rather than a hand-written toolchain
path. The toolchain does not sit where the single- versus multi-config layouts
suggest, and a wrong path fails with "Could not find toolchain file" followed by
"CMAKE_CXX_COMPILER not set", which reads as a broken compiler.
The build directory is .build/build/Release, not .build. Building the wrong one
reports "Generator: execution of make failed", which reads as a toolchain problem.
The CMake cache is sticky: re-running conan install with a changed option does not
update an already-cached CMake variable, so the target stays absent even though
the option was accepted. Added the check and the cache-clearing steps.
Standing up and updating one of these hosts involves a handful of steps that fail
quietly when missed, and the knowledge was living in people's heads and in an
untracked file on the host itself. That file died with the last rebuild.
Covers updating the checkout, re-applying the host overlays a checkout can
disturb, building, and starting the collector and both nodes, then verifying all
three signals actually arrive rather than trusting the configuration.
The failure modes it calls out are ones already hit here: bringing the collector
up without the cloud overlay, which exports nothing and logs no error; omitting
--force-recreate, so a config change appears applied but is not; a log directory
whose name does not match the instance id, which leaves logs unlabelled while
metrics stay labelled; and reading the container command line to check which
config is active, which cannot distinguish the two cases because the overlay
mounts over the same path.
It also records two signals that read backwards: a higher acquisition timeout
count can mean a node is retrying forward more cheaply rather than stalling, and
outbound byte counters are recorded at the decision to send rather than at the
wire, so under a flood they overstate egress by orders of magnitude.
Deliberately names no credential or environment files and carries no host
identity, so it is safe in a public tree. Which file holds what is documented
where it is needed, next to the installer that reads it.
Running two nodes on one host is how the storage backends get compared with
everything else equal, but only the first instance's config was ever committed.
The second node's config, both systemd units, and the collector's per-node log
identity lived on the host alone and were lost when it was rebuilt.
Adds xrpld-telemetry-mainnet2.cfg, derived from the first and differing only
where two nodes on one host must differ: ports, data and log paths,
service_instance_id, and [node_db] type. Ports continue the offset-by-ten scheme
already in use -- devnet on 5005, Mainnet on 5015, so this one on 5025 -- so all
three can bind together.
Adds the units as templates plus install-units.sh, which fills them from an
untracked .env.devbox. This repository is public and f64f4b35e7 already removed a
personal home directory from shipped config, so the run user and checkout path
must not come back into git. The installer refuses an env file that is not mode
600 and refuses to install a unit still holding an unsubstituted placeholder, so
a half-configured host fails loudly rather than producing a unit systemd never
starts.
Both units carry RequiresMountsFor for the data mount. That mount's fstab entry
normally uses 'nofail' so a missing disk does not block boot, which also means it
fails silently; without the guard a node starts anyway and writes its nodestore
to the root filesystem until it fills.
Re-adds the collector's per-node log identity: include_file_path plus operators
lifting the log directory basename into service.instance.id. Without it the logs
pipeline carries no service_instance_id label while the metrics pipeline does, so
the dashboards' $node filter matches nothing for logs and reads as "no logs"
rather than as a misconfiguration.
Renames the first instance's log directory from data/logs/mainnet to
data/logs/xrpld-mainnet, the one change to that file: the derivation above reads
the basename, so it has to equal the service_instance_id for a node's logs and
metrics to carry the same label.
No credentials or host-identifying values here. .env.devbox and .env.grafanacloud
are both covered by the .env.* ignore rule; only the examples are tracked.
The transitions panel used increase(...[$__rate_interval]). $__rate_interval is
defined as max($__interval + scrape, 4 * scrape), i.e. deliberately one scrape
longer than the step so rate() windows overlap and lose no counter increase.
That overlap is harmless for rate(), but this panel reads the value as a count
of discrete events, and the overlap counts each event in more than one bucket.
Measured against a log-derived ground truth of 106 syncing transitions on
devnet-otel-usw2-01 over 2026-08-11T11:05Z..2026-08-12T23:04Z, the old query
reported 111.3 at a 300s step and 133.7 at a 60s step -- the error grew to +26%
as you zoomed in, because the overlap is a larger fraction of a smaller step.
Switch to $__interval so the buckets tile exactly, and wrap in round() because
increase() extrapolates to the window edges and so reports fractional counts for
an integer counter. The same measurement now gives 106 at 300s, 105 at 60s and
107 at 900s. Every state and both nodes land within a few counts of truth at any
zoom, and the legend Total is now a meaningful figure.
Pin Min step to 1m: the real scrape interval is 60s while the datasource
declares 15s, so without a floor $__interval can fall below one sample.
Draw as bars with 0 decimals -- the value is a discrete count per bucket, and a
line implies interpolation between counts that does not exist.
Nine conflicts, resolved as follows.
src/xrpld/app/ledger/detail/InboundLedger.cpp -- kept this branch's version.
phase10 sets the span's outcome/timeouts/peer_count attributes inline at each
exit; this branch replaced that with the idempotent finalizeAcquireSpan(), called
on all four exits (init, done, give-up, destructor). Taking phase10's blocks
would have set the outcome twice against a helper documented as not overwriting
what the real exit recorded. phase10's comment explains why peer_count must not
be read in a destructor; the helper solves that structurally by taking
std::optional<std::size_t> and being passed std::nullopt from there.
src/xrpld/telemetry/MetricsRegistry.cpp -- kept metric::ledgerEconomy over
phase10's "ledger_economy" literal. This branch added the naming check that
requires constants for converted families, so the literal would regress it. Took
phase10's comment cleanup.
src/xrpld/telemetry/MetricsRegistry.h -- kept registerRotationStateGauge(), which
only exists here, and took phase10's removal of the stale task-number comment.
validate_telemetry.py -- combined both. phase10 replaced serial metric polling
with a concurrent fan-out on one shared deadline, because 58 metrics x 45 s of
additive timeout overran the CI budget; that is kept. Its target list filters on
SKIPPED_METRIC_GROUPS rather than the two literals it hardcoded, so the
sync_diagnostics group stays owned by assert_sync_diagnostics_metrics() instead
of being polled and reported twice. Both SYNC_DIAGNOSTICS_GROUP and
METRIC_POLL_CONCURRENCY are needed and both are kept.
check_otel_naming.py -- both sides extend the rule docstring. Took phase10's
fuller Rule E text (doc discovery, allow-dotted markers) and re-appended rules
I/J/K/L, which exist only here.
expected_metrics.json -- the two sides add disjoint sibling groups, so both are
kept: sync_diagnostics alongside node_health_gauges, overlay_reduce_relay,
overlay_overflow, validation_lifetime_counters and not_asserted. Both dashboard
uids are kept, giving 16 asserted uids against 16 dashboards on disk.
expected_spans.json -- kept this branch's span set, a superset that adds the
acquire phase spans, ledger.serve, txset.acquire and peer.dial, and expands
ledger.acquire's required attributes. Took phase10's description, which documents
what the totals mean, and its note on how the RPC wildcard span is created.
total_span_types and total_unique_attributes are recomputed for the union: 48 and
74, since each side's figure counted only its own spans.
Docs: took phase10's more accurate wording on what the dashboard check actually
covers, and corrected the dashboard count from 15 to 16 where the merge made it
stale.
Verified: no conflict markers remain, both JSON contracts parse, both Python
files compile, asserted dashboard uids match the dashboards on disk exactly, and
the OTel naming check reports all layers consistent.
The Operating Mode Transitions panel queried state_accounting_*_transitions
directly. Those are monotonic counters, so the panel drew a slowly rising line
and a few transitions per hour were invisible against a total in the hundreds.
It also fell off a cliff whenever xrpld restarted and the counters reset to 0,
which reads as missing data rather than a restart.
Wrap each target in increase(...[$__rate_interval]) so each point is the number
of transitions in that bucket and the series survives a counter reset. This is
what the sibling panels on the same row (Operating Mode (Time Share), State
Duration Rate) already do.
Verified against devnet-otel-usw2-01/02 over 2026-08-11T11:01Z..2026-08-12T16:23Z:
the fixed expression reports 107 and 123 syncing transitions, matching the
counter deltas, and stays continuous across the 12:07 restart where the raw
counter dropped 630 -> 1.
Brief mode flaps remain invisible on Operating Mode (State Timeline) because a
~2 s dwell cannot be captured by a 60 s scrape; this panel is the place to read
them.
This branch had already made the same corrections independently, and in
richer form, so the resolution keeps this branch's version nearly throughout:
- 09-data-collection-reference.md: this branch already documents the
state-accounting gauges as cumulative **microseconds** with an explanatory
callout, and already names `jobq_job_count` with its `jobq` group. Kept.
- telemetry-runbook.md: already carries `jobq_job_count` in both tables. Kept,
along with this branch's larger additions.
- OpenTelemetryPlan.md: kept this branch's rewritten section 9 blurb, which
describes the inventory without hardcoding counts and so cannot drift.
- consensus-health.json: kept this branch's rewrite. It deliberately removed
the four TraceQL close-time detail panels and renamed the agreement panel;
the incoming side would have resurrected them. Panel count unchanged at 26.
- integration-test.sh: this branch's unprefixed native metric names were kept,
but it still asserted `job_count`, so the `jobq_job_count` correction was
carried over. That check would otherwise always fail.
The troubleshooting step queried `job_count`, which returns no series. The
gauge is registered as `makeGauge("job_count")` but `Application.cpp` passes
`collectorManager_->group("jobq")`, so the exported name carries the `jobq`
segment. The two metric tables in this file were corrected when phase-6
merged forward; this example was missed because it sits outside the tables.
Conflict resolution kept this branch's evolution and re-applied phase-6's
fixes on top of it, rather than taking either side wholesale:
- consensus-health.json: kept the native `span_calls_total` metric name and
the `interval: 15s` and point styling from this branch; added phase-6's
`close_time_correct` PromQL filter and the NetClock axis labels. The
TraceQL boolean-regex filter stays removed and the `byRegexp` overrides
carry over. Panel count unchanged at 27.
- 09-data-collection-reference.md: kept this branch's headings, its more
detailed consensus attribute table (which already types
`consensus_round_id` as int64) and its section numbering, including the
deliberate removal of the SpanNames inventory. Carried over only the
correction that the state-accounting duration gauges are cumulative
microseconds, not seconds.
- telemetry-runbook.md: kept this branch's native metric names
(`span_calls_total`, `span_duration_milliseconds_bucket`); carried the
`rpc.request` -> `rpc.http_request` span-name fix and the `jobq_` segment
on the job-queue depth metric.
- integration-test.sh: kept this branch's `check_otel_metric` form and
carried the `jobq_job_count` correction.
The integration test asserted `rippled_job_count`, which never reports any
series, so that check always failed. `JobQueue` registers the gauge as
`makeGauge("job_count")`, but `Application.cpp` passes it
`collectorManager_->group("jobq")`, so the emitted StatsD name is
`jobq.job_count` and the exported Prometheus name is
`<prefix>_jobq_job_count`.
Corrected the same name in two runbook tables that also dropped the `jobq`
segment. `09-data-collection-reference.md` already had it right, which is why
the two documents disagreed.
Routed here rather than to the phase-10 PR where it was reported: the wrong
name is present in `integration-test.sh` on every branch from phase 6
onward, and this is the branch that introduces the file.
Left alone deliberately:
- `statsd-node-health.json` still queries the old name, but that dashboard is
deleted at phase 7 in favour of `node-health.json`
- `06-implementation-phases.md` names `job_count`, which is accurate as the
code-level makeGauge argument rather than the exported metric name
The file_storage extension was added to otel-collector-config.yaml, which
every stack mounts. That made the extension mandatory: the collector image
runs as 10001:10001 and ships no writable directory, so any stack without a
prepared volume would fail to start rather than merely lose offsets. The
workload-validation stack mounts this same config and has no such volume.
Offset persistence is only useful where logs outlive a restart. The workload
harness creates a fresh log directory per run, so it has nothing to resume
from. Move the extension, the receiver's storage reference and the extended
service.extensions list into otel-collector-filestorage.yaml, layered as a
second --config by the developer stack alone. The base config keeps
start_at: beginning, which is what actually fixes the reported defect, and
stays self-sufficient for every other stack.
Verified against the pinned collector image: the base config validates and
runs on its own with no volume mounted and still ingests a line written
before startup; base plus overlay validates, preserves the base receiver's
operators through the merge, and re-ingests that line zero times on a second
run against the same volume.
Conflict resolution, OpenTelemetryPlan/09-data-collection-reference.md
Known Issues table: kept phase-9's expanded 17-row table rather than the
incoming 4-row version, and re-applied the incoming correction to the one
row it changed. The other three incoming rows are already present in
phase-9's table, and phase-9's version is more current (jobq_job_count,
not job_count).
The rpc_requests row now says [insight] server=otel; server=statsd
reaches a collector with no statsd receiver on this branch. The section 8
Quick Reference and the fallback caveat auto-merged cleanly.
09-data-collection-reference.md contradicted itself. Section 2 presents
server=otel as the recommended transport with StatsD as a fallback, but
the Known Issues table and both Configuration Quick Reference examples
still prescribed server=statsd, which on this branch reaches a collector
with no statsd receiver and an unpublished 8125/udp.
Switch the Known Issues row and the Minimal and Production examples to
server=otel with the OTLP metrics endpoint. Keep the labelled fallback
block, and state what it actually requires: re-adding the statsd
receiver and republishing the port. Also record that StatsDCollector
applies prefix to metric names while OTelCollector does not, so the two
transports do not produce the same series.
The phase-6 copy is left alone; server=statsd is correct there.
check-otel-naming fails Rule D on this branch: ledger-data-sync.json
aggregates by xrpl_work_item, which nothing in-tree emits because
perf-iac's alloy pipeline stamps it, so it cannot be derived from
*SpanNames.h. The sibling perf-iac identities xrpl_branch and
xrpl_node_role are already allowlisted; this one was missed when the
by() clauses were reintroduced.
Add it to EXTERNAL_INFRA_LABELS alongside them. Verified by removing
the entry again, which restores the failure.
Conflict resolutions:
- docker/telemetry/xrpld-telemetry.cfg: relocation conflict. phase-9 had
already moved [insight] to the end of the file with server=otel, so the
incoming block was dropped rather than inserted. Keeping both would have
produced two [insight] sections, which merge last-wins into a single
effective section, silently reviving the bug this branch just fixed.
phase-9's per-branch service_instance_id=xrpld-devnet is preserved.
- OpenTelemetryPlan/06-implementation-phases.md: kept both corrections.
phase-9's "Tempo" is right (no Jaeger anywhere in the stack) and
phase-8's "active, sampled span" is right: Log.cpp:328 injects only
when spanCtx.IsValid() && spanCtx.IsSampled().
- OpenTelemetryPlan/09-data-collection-reference.md and
docs/telemetry-runbook.md: kept phase-9's structured-metadata LogQL.
The collector's filelog regex_parser already extracts partition,
severity, trace_id and span_id, so phase-8's inline regexp forms are
redundant, and a line filter matches the literal text in a message body.
The generated node config carried two [insight] blocks. Duplicate ini
sections do not replace one another: parseIniFile emplaces the section
name (a no-op when it already exists) and appends the lines to the same
vector, then Section::append writes each key with insert_or_assign. The
effective section was therefore server=statsd with the first block's
endpoint and service_instance_id surviving but unused.
CollectorManager selects StatsDCollector for that value, so the nodes
emitted beast::insight metrics over UDP to 8125, which has no receiver
in the collector pipeline and no published port. The script's own check
asserts that 8125 is not listening, and its insight metric assertions
fail on zero series.
Keep only the server=otel block so the config matches what the script
verifies.
This branch removes the collector's StatsD receiver and un-publishes
8125/udp, but xrpld-telemetry.cfg still selected server=statsd, so the
sample config sent beast::insight metrics over UDP to a port nothing
listens on. Phase7_taskList.md:132 lists this switch as required work.
Select server=otel and replace address= with the OTLP metrics endpoint.
Document that endpoint and prefix are informational only, since
OTelCollector records on the global MeterProvider that [telemetry]
configures and formatName() applies no prefix, and note that beast
instruments are not exported yet because the collector is constructed
before the MeterProvider is registered.
Six findings from the review of #6494 survived independent verification.
Each was checked against the branch tip, and where behaviour was in
question, against a live collector and Loki rather than from the
reviewer's claim or from documentation alone.
Plan-doc section numbering. 06-implementation-phases.md used "## 6.9"
twice: for the new Phase 8 section and for the pre-existing Risk
Assessment. Three references already pointed at 6.8.1 and none at 6.9,
and the later phases are numbered 6.8.2 through 6.8.4, so Phase 8
becomes 6.8.1 and the sequence is monotonic. Renumbering to 6.10, as
suggested on the PR, would have collided with Success Metrics.
filelog read position. The receiver relied on the upstream default
start_at=end, which skips everything a node wrote before the first poll
and reads nothing at all from a log that has stopped being written to.
Read from the beginning instead, paired with a file_storage extension so
a restart resumes at the last offset rather than re-ingesting the file.
The collector image runs as 10001:10001 and ships no writable directory,
and a fresh named volume is root-owned, so a one-shot init service
prepares the volume first. It reuses an image the stack already pulls,
adding no new dependency.
Loki log stream label. The job resource attribute did not become a Loki
index label, so the documented {job="xrpld"} queries matched nothing.
Verified against grafana/loki:3.4.2 with its default config: only
service_name and deployment_environment are indexed, and job arrives as
structured metadata, which a stream selector cannot match. Dropped the
attribute and moved the twelve queries this branch introduced to
{service_name="xrpld"}. Three further occurrences in
07-observability-backends.md originate on the phase-1a branch and are
left for a commit there.
Trace ids on unsampled spans. Logs::format emitted trace_id and span_id
whenever the span context was valid. A span dropped by the
ParentBasedSampler still carries its parent's ids, so log lines
advertised traces that were never exported and the log-to-trace link
resolved to nothing. Require the sampled flag as well, and correct the
task list and the documentation that promised the fields unconditionally.
The remaining two findings were refuted. The reported risk of signing
material reaching Loki does not hold: Logs::format already scrubs seven
sensitive fields, and there is a single write path to the log file, so
every JLOG site is covered. The suggestion to add internalLink to the
Loki derived field is not applicable, because that key is not part of
Grafana's schema.
All five TraceQL panels on this dashboard returned nothing, and did so
without any visible error: they filtered on
span.close_time_correct=~"$close_time_correct", but close_time_correct is
a boolean attribute (RCLConsensus.cpp:601 passes a raw bool), and Tempo
restricts the regex operator to string operands, so the spanset resolved
to false. With the variable defaulting to All the clause rendered as
=~".*", so the panels were empty out of the box and looked exactly like a
node with no consensus activity.
Note this is the opposite of PromQL, where an absent or empty label does
match ".*" — which is why the 17 Prometheus panels on this same board were
unaffected and the dashboard appeared healthy.
Dropped the clause from all six queries, matching phases 9 and 10 where it
is already gone. The $close_time_correct variable now filters the
Prometheus "Close Time Agreement" panel instead, which already grouped by
that label but never filtered on it, so the control stays useful rather
than becoming dead UI.
Two defects were masked behind the empty panels and are fixed too:
- "Close Time: Raw Proposals" and "Close Time: Effective / Quantized"
carried unit dateTimeFromNow over close_time_self/close_time, which are
NetClock seconds (Ripple epoch), while Grafana's dateTime formatters
expect a millisecond Unix epoch — every point would have rendered as
roughly 1970. They now plot as plain numbers with the axis labelled
"NetClock Seconds (Ripple Epoch)", and the descriptions give the
946684800 offset for converting to Unix time.
- "Close Time Vote Bins & Resolution" matched its unit and axis overrides
byName against "Vote Bins" and "Resolution", which are not field names;
TraceQL select() yields close_time_vote_bins and close_resolution_ms, so
neither override applied. Switched to byRegexp so the match holds
whichever scope prefix Grafana emits.
No panel was added or removed: the (type, title) multiset is unchanged at
22. resolution_direction keeps its regex filter, which is correct there —
it is set from a std::string whose values are exactly the variable's
increased/decreased/unchanged.
The integration test's span assertions never actually ran. check_span()
built a Tempo /api/search call with --data-urlencode but no -G, so curl
POSTed the params as a body; Tempo answers 200 and ignores the query, so
every span name looked present. Verified against a live Tempo 2.9.4: the
buggy form returns the store's total trace count for any name, including
"zzz.does.not.exist"; with -G a real name returns 1 and a bogus one 0.
Fixed alongside it: the RPC check asserted "rpc.request", which is never
emitted (ServerHandler.cpp builds "rpc.http_request"). These two had to
change together, since -G turns the bogus name from a silent pass into a
hard failure.
Also in the script: a consensus timeout logged two failures and counted
two, because a post-loop else re-reported what the timeout branch had
already reported; and three unguarded curl calls aborted the whole script
under set -euo pipefail, making the ACCOUNT_ZERO fallback dead code with
no cleanup. Guarded the curls and wired an EXIT trap to the existing
cleanup(). The trap deliberately fires only before the summary, so a
completed run still leaves the stack up as the header documents.
Docs corrections, all re-derived from code:
- span inventory heading 35 -> 38, attribute heading 83 -> 89 rows
(78 unique keys), and the section 6 header table now carries the
missing TxApplySpanNames.h row so its columns sum to the same figures
- two stale paths: ConsensusSpanNames.h is under include/xrpl/consensus/,
TxSpanNames.h under src/xrpld/telemetry/
- consensus_round_id is int64, not string (RCLConsensus.cpp sets
prevLgr.seq() + 1); the runbook's TraceQL examples now use a numeric
literal instead of an unparseable bare <round_id>
- state-accounting duration gauges are cumulative MICROSECONDS, not
seconds (NetworkOPs.cpp declares std::chrono::microseconds and
publishes dur.count() raw)
- sampling_ratio is not a config key; head sampling is fixed at 1.0 and
the shipped collector has no tail sampling, so the caveat was rewritten
- the plan blurb referenced Jaeger; this stack is Tempo
Both the CMake option and the Conan recipe default telemetry on, so the
build worked, but this workflow exists to exercise telemetry and should not
depend on a default it does not control. If that default ever flipped, the
binary would build cleanly and then every span and metric assertion would
fail for a reason no log names.
Stated explicitly, the failure mode inverts: CMakeLists.txt does
find_package(opentelemetry-cpp CONFIG REQUIRED) under this option, so a
dependency graph without it fails at configure time with a clear message.
The shared build-deps action is left alone deliberately — forcing the Conan
option there would change every other workflow that uses it.
The setting was flagged in review as unauthenticated admin access. It is
deliberate, and it matches the sibling stack in docker-compose.yml, which
carries the same two variables and the same published port with its intent
in comments. This copy had none, so the reasoning lived only in a review
thread and was rediscovered as a finding each time the file was looked at.
Viewer would break the harness rather than harden it: the validation suite
drives the Grafana API against this instance to confirm each dashboard
provisions and loads, and the dashboards and datasources come from the
read-only mounts on the same service.
Two defects reported against the harness, both confirmed.
The TPS field was computed with `bc` at scale=2, and bc omits the leading
zero: it prints ".25", not "0.25". A bare ".25" is not valid JSON, and this
was the normal case rather than an edge case — ledgers close every few
seconds, so ledger-advance over elapsed-seconds is well under 1 for any
realistic window. It survived earlier checks because those piped the file
through jq, which accepts the malformed form; Python's json rejects the whole
file. awk's %.2f always pads, so the field is now produced with awk. Audited
the other numeric fields at the same time: CPU average and memory peak
already used awk, and the p99, sample count and consensus mean are integers,
so TPS was the only one affected.
Separately, a failing attribute fetch was reported under the span's own check
name, which had already recorded the trace as found. That produced two
entries for one name, one passing and one failing, inflating the check total
and blaming the trace-existence check for a failure in a later network call.
The fetch now carries its own error handling and reports under
`span.attrs.<span>`, matching where its successful counterpart reports. It
moved into a helper rather than growing `validate_spans`, which was already
well over the line limit.
Three consecutive validation runs timed out at Step 3 with nodes stuck at
"unreachable", and the reason was not recoverable from the logs. The node
logs showed the failing nodes stopping at an identical point, immediately
after JobQueue initialisation and before the debug log is opened, with no
error text at all. The harness knew each node's pid and never used it, so a
crashed node was indistinguishable from a slow one.
The readiness loop now checks whether each node process is still alive and
fails as soon as one is not, instead of waiting out the remaining window and
burying the cause under two minutes of progress output. Liveness is not a
bare `kill -0`: an exited-but-unreaped child keeps its pid, so a zombie
answers `kill -0` and reads as alive for the whole window, which is exactly
how a crashed node came to look like a slow one.
On failure each stopped node reports its wait status and the tail of its
stdout. The status is the discriminator that was missing: 137 for a SIGKILL,
139 for a segfault, 134 for an abort, anything below 128 for a deliberate
exit. stdout is printed inline rather than left to the artifact upload,
because a node that dies before its debug log opens writes nothing else and
a cancelled run uploads nothing at all.
This is instrumentation, not a fix. The failure is not attributable to the
recent changes on this branch: the first red run touched only the two Python
files used at Steps 4 and 5, both of which run after this gate, and the same
harness passed 5/5 twice before that.
Run with no arguments the script iterated an empty list, found no
violations and printed "OK: 0 dashboard(s) passed" with exit 0 -- a clean
bill of health for reading no files, indistinguishable from a real pass.
A bare run now defaults to every dashboard beside the script, and a run
that still ends up with nothing to check exits 2 rather than reporting
success. Passing paths explicitly behaves as before.
Single-value panels carry the filtered network type in their title because
they have no legend to put it in; multi-series panels carry it in
legendFormat instead. This gauge was the only single-value panel across the
fifteen dashboards without it, so its reading was ambiguous once more than
one network type was in scope.
A validation run timed out at Step 3 with only 4 of 5 nodes proposing, and
the reason was unrecoverable afterwards. Two gaps caused that.
The node-log artifact collected `node*/debug.log` but not `node*/stdout.log`.
A node that dies before its log sink opens never writes a debug.log at all,
so stdout is the only place its reason survives — and that file is written by
the harness and read by nothing, so it went to the runner and was discarded.
The failing node's log was simply absent from the artifact.
The readiness loop also fetched each node's `server_state` and threw it away,
reporting only a count. "4/5 nodes proposing" says a node is missing but not
which one, so there is nothing to grep for even once the logs are kept. The
timeout now names each node that is not proposing along with the state it
last reported, distinguishing a node that answered with a non-proposing
state from one whose RPC port did not answer at all.
Neither change affects a healthy run: the accumulator resets each attempt and
stays empty while every node is proposing.
The alerting example env file, the contact-point provisioning header and one
runbook line still pointed at a gitignored helper script and at a rollout
phase number, neither of which ships. The contact-point header now states the
policy-tree warning inline rather than deferring to a file the reader cannot
open.
Four files conflicted.
xrpl.proto and TraceContextPropagator.h: phase-9's note is a semantic
superset of phase-8's and already carries no plan-folder pointer, so
phase-9's text is kept and nothing phase-8 said is lost.
node-health.json: the only change phase-8 makes to this file is renaming a
row that phase-9 had already deleted, so phase-9's file stands. Panel count
holds at 59 and both Validated Ledger Seq panels survive.
TESTING.md: resolved per hunk rather than by side. Phase-9's replacement of
the drifted span table and its new Grafana Cloud section are kept, and
phase-8's de-numbered "Test 3: Log-Trace Correlation" heading is taken --
keeping phase-9 wholesale would have reinstated the phase number that
phase-8 removed.
The runbook pointed at a planning document and a helper script that are
outside the shipped tree, and at a dashboard script that no longer exists,
so an operator following it hit three dead ends. The Cloud alert workflow
is now described by what it does to the tracked rules.yaml.
The integration-test sections, alert provisioning headers and a naming-check
test comment were indexed by rollout phase and task numbers defined only in
planning documents outside the shipped tree. Each now names what it covers.
These comments were indexed by task, use-case and limitation numbers that
are defined only in planning documents outside the shipped tree. Nothing
in the repository defined them, so the cross-references resolved nowhere.
Each comment now states what the code does.
Two unrelated edits had crept in. The added `.claude/` line duplicates
`/.claude`, which the same file has already carried since well before this
branch, so it never changed behaviour. Removing the root `.env.*` rule did
change behaviour: contributors lost repo-level cover for any env file
outside docker/telemetry, and the local ignore that masks this on one
machine is not part of the repo.
Restore the rule and drop the duplicate, which leaves the root gitignore
byte-identical to the parent branch.
The scoped rules added under docker/telemetry stay: they keep the three
tracked .env.*.example files visible and stop the generated Grafana Cloud
dashboards being committed, which is telemetry work this branch owns.
Two defects reported against the validation harness. Both premises were
correct, but neither suggested fix was, so the remedies differ.
Dashboard panel count: `len(dashboard["panels"])` treated Grafana row
objects as panels and skipped the panels nested inside collapsed rows, so
every dashboard was over-reported by between 1 and 10 (`log-derived-insights`
read 41 against a true 31). The check also passed unconditionally on HTTP
200, so a dashboard that renders nothing would still pass. `_leaf_panel_count`
now walks row children and the result gates the verdict. Gating on the old
top-level length, as suggested, would not have caught the case it was aimed
at: a dashboard made only of collapsed rows counts its rows and reports a
positive number while rendering nothing.
RPC latency percentiles: `LoadStats.record` appended a latency for every
outcome, including requests that never got a reply, where the value is a
time-to-failure rather than a round trip. A timeout contributed the full
receive timeout, and at the error rate a real run shows this reported p95 and
p99 of 10000 ms where the true figure was 5 ms. `record` now takes an
optional latency and the timeout path passes none. The suggestion to append
only on success was not adopted: a reply carrying `status: error` is a
completed, timely round trip whose latency is a genuine measurement, and
discarding it would throw away real data. `per_command` is now keyed off the
request counts rather than the latency map, so a command whose every request
timed out still appears in the report instead of vanishing from it, and each
entry carries a `latency_samples` count.
The runbook told operators to read a missing outcome as "never went to the
network", which stopped being true once the abort path started setting it,
and the glossary still described outcome as a two-value split.
Document all three values, where each is written, and why peer_count is
absent on the abort path. Several claims were wrong and are corrected:
- Give-up is reached at about 18s, not 21s. There is no setTimer call, so
the first timer runs immediately and the old derivation counted a wait
that does not happen.
- A live aborted rate does not imply stalled acquisitions. A clean
shutdown clears every in-flight acquire, the admin fetch_info clear does
the same, and a full job lane stops timeouts advancing so give-up cannot
fire. The runbook already said the last of these elsewhere.
- A missing outcome does not mean exactly one thing. tryDB can set failed_
and return before done() runs, exporting a span with no outcome at all.
- failed covers unusable ledger data as well as exhausting the retry
limit, so a failed span can carry timeouts=0.
- The aborted lower bound of one minute holds only on the sweep path.
- The sweep measures time since anything last asked for the ledger, not
since data last arrived.
Also fixes the mainnet verification command, which still used the old RPC
port, and drops an inaccurate claim from the config comment about which
ports the workload scripts use.
The mainnet and devnet telemetry configs are host processes sharing one
network namespace and working directory, and they agreed on every port and
both database paths. Running them together left the second unable to bind,
and running them in sequence pointed a mainnet node at a store holding the
other network's ledgers, silently and with no error.
Offset the mainnet ports by ten and name both database paths after the
network, matching what debug_logfile already did.
Also narrow the attack surface this config carried. [port_ws_public] has
no admin key, so every caller on it is a guest, and it was bound to all
interfaces on a node following Mainnet; it is now loopback, and nothing in
the repo connects to that port. [signing_support] is removed: it only
affects non-admin callers, the admin ports already grant signing to local
callers, and upstream deprecates the commands it exposes. Verified against
a running node: signing still succeeds on the admin port and is refused as
notSupported on the public one.
Every aggregation that filtered on xrpl_work_item left it out of its
sum by() grouping, so PromQL dropped the label and the xrpl_ident legend
these panels build from it rendered without the work item. 217 clauses
across the touched boards.
Perf-iac gives each work item its own set of nodes, so service_instance_id
already separates the runs and the visible effect is the legend rather
than merged series. Four clauses are the exception and aggregate across
nodes, where the grouping does real work: Convergence, Lag Behind Network
Tip and Build Version now measure spread, lag and version distribution
within one work item instead of smearing two independent clusters
together. Lag Behind Network Tip also needed its on() join key extended to
match, or two coexisting work items would fail the query outright.
The five heatmap inner by (le) clauses are deliberately untouched: an
extra grouping label would superimpose several distributions.
Also in this change:
- transaction-overview gains per-type-per-stage rate and failure-rate
panels, so all three requested dimensions exist rather than latency
alone, and its stage-rate panel picks up the filters and legend its
siblings already had.
- The acquire-duration panel splits by outcome. An aborted acquisition is
open until the sweep reaps it, so mixing it into one percentile
conflated fetch latency with time-to-abandon.
- Ledger Acquire Rate by Outcome gets the same Title Case value mapping as
its neighbour, and node-health gains the $outcome variable the two
panels now need.
LedgerReq Wait by Handler and NodeStore Read Latency (Bottleneck
Discriminator) filter on xrpl_work_item but left it out of their sum by()
grouping, so PromQL dropped the label and the xrpl_ident legend lost that
segment.
These two targets took their current form on this branch, so they are
fixed here. The other nine in this file predate it and were fixed on
pratik/otel-phase7-native-metrics, which merges forward.
The discriminator panel divides two aggregations; both sides carry the
label so their label sets stay equal for vector matching.
Nine targets across four panels filtered on xrpl_work_item while leaving
it out of their sum by() grouping. PromQL keeps only the labels listed in
by(), so the label was dropped from the result and the xrpl_ident legend
these panels build from it lost that segment.
Perf-iac runs stamp xrpl_work_item and give each work item its own set of
nodes, so service_instance_id already separates the runs; the visible
effect is the legend rather than merged series. Adding the label makes the
work item identifiable, which is the comparison these panels exist for.
NuDB Read Latency and NuDB Read Found Ratio each divide two aggregations.
Both sides get the label so their label sets stay equal and vector
matching still works.
Two comment-only notes.
MetricsRegistry.cpp: record why the app and overlay includes exist and
what they cost. They are what makes loops.txt carry
xrpld.app <-> xrpld.telemetry and xrpld.overlay <-> xrpld.telemetry, where
ordering.txt previously had telemetry strictly below both. The observable
gauges are pull-model, so their callbacks need the concrete types to
sample live state. The cycle is confined to this translation unit: no
telemetry header includes app or overlay, and all of src/xrpld builds
into one target, so there is no header or link cycle. Inverting it needs
a metrics-source interface below overlay, which is left as follow-up.
Also note loops.txt is generated and must never be hand-edited.
docker-compose.yml: state the Compose >= 2.24.0 floor. The grafana
service uses the long-form env_file mapping, which older Compose cannot
parse, and it fails for the whole file rather than that one service. The
long form is required because .env.alerting is gitignored and absent in a
fresh clone, and the short form treats a missing env file as an error.
The test carried both <xrpl/proto/xrpl.pb.h> and the bare <xrpl.pb.h>.
Both resolve to the same generated header, because the proto helper puts
the generated tree and its prefixed subdirectory on the target, so the
second include expanded to nothing behind the header guard.
The bare form arrived from merging two same-day clang-tidy commits that
added the include with different spellings. Keep the prefixed spelling,
which is what the telemetry headers and the upstream phase branches use.
An InboundLedger destroyed while !isDone() recorded recordAbort() in the
metrics but left ledger.acquire carrying only the attributes set at
construction, so an abandoned acquisition was indistinguishable from one
still in flight.
Set outcome=aborted plus timeouts on that path, and add the val::aborted
constant. peer_count is deliberately omitted: reading it goes through
Overlay, and a destructor must not depend on Overlay still existing.
The status stays Unset for an abort, because InboundLedgers::stop()
clears every in-flight acquisition, so a clean shutdown would otherwise
report errors. done() now sets Error when failed_, which is an
unambiguous failure of the operation. Success is left Unset rather than
Ok, per the OpenTelemetry guidance that instrumentation should not
assert Ok.
rpcStart, jobQueue, jobStart and jobFinish each acquired a lock without
braces, so it stayed held to the end of the function and covered the OTel
recording calls this branch added. counters_.jobsMutex and
counters_.methodsMutex are process-wide, so every worker thread starting
or finishing a job serialised on the SDK's work.
That work is not a bare atomic add: each record builds a map-backed
attribute set and takes a spin lock inside the SDK, whose backoff reaches
a millisecond-scale sleep under contention.
Brace the lock plus the state it guards, then record after it releases.
None of the metric calls read lock-protected state, so this is
semantics-preserving, and rpcEnd() in the same file already had this
shape.
The row was titled after the rollout phase its metrics came from, a number
defined only in a planning folder outside the shipped tree. Grafana shows
this title to operators, so it named something no reader could look up.
Two panel descriptions pointed at a rollout phase and task number defined
only in a planning folder outside the shipped tree. The note still names
the file and the change it is waiting on, which is the part a reader can
act on.
The catalog tagged every span with a rollout phase number defined only in
a planning folder outside the shipped tree, so the column meant nothing
to a reader of the repository. The span name and source file identify
each entry.
These comments and headings identified the feature by a rollout phase
number defined only in a planning folder outside the shipped tree, so the
label meant nothing to a reader of the repository.
The section labels carried rollout phase numbers that are defined only in
a planning folder outside the shipped tree, so they meant nothing to a
reader of the repository. The descriptive half of each label already
identifies the section.
The label carried a rollout phase number defined only in a planning
folder outside the shipped tree. The descriptive half already identifies
the section.
These comments pointed at a planning folder and at its rollout phase
numbering, neither of which is part of the shipped tree, so the
references would dangle for any reader of the repository. Each comment
now states the fact it was pointing at.
The trace_state comment pointed at a planning document that is not part
of the shipped tree, so the reference would dangle for any reader of the
repository. State the reserved-and-inert fact on its own.
The rename check rewrites a bare pre-rename binary name in any processed
doc, which turned the sampler's selector description into "against xrpld
or xrpld". Describe the fallback without spelling the legacy token.
Fixes the review findings on this PR that belong to files it owns, plus
several defects found while verifying those fixes. Findings in files owned
by upstream branches are routed there and left untouched here.
Correctness:
- tx_submitter: advance the account sequence only on results that actually
consume one (tes*, tec*, terQUEUED). tem*/tef*/tel* never reach the
ledger, so advancing left a permanent gap that every later submit from
that account inherited. Add a re-fetch hatch so a repeated non-consuming
failure cannot livelock on the same sequence, and gate the account check
on funded-ness rather than list length.
- validate_telemetry: filter spans by name before collecting attributes, so
a per-span attribute contract can no longer be satisfied by a sibling
span; require exact name equality for non-wildcard children and glob
matching for wildcards; bounds-check every returned series instead of
only the first.
- collect_system_metrics: select xrpld by argv[0] rather than a substring
match on the whole command line, which averaged in unrelated processes
and reported their RSS as xrpld's. Count genuine 0.0 CPU readings, use a
clamped nearest-rank p99 index, and record RPC latency only on success.
- benchmark: return each verdict through a named variable instead of a
command substitution, so the pass/fail counters survive and the exit gate
can fire. Scale before dividing in the percentage math, which truncated a
1.26% impact to 1.00% and cleared a 1% threshold.
- compare_to_baseline: fall back to the absolute bound when the baseline is
not positive, so a 0 -> 500 ms jump is no longer "within bounds".
- rpc_load_generator: bound each connection to one in-flight recv(), drain
in-flight requests before closing, use a nearest-rank percentile, and
report delivery shortfall so an under-delivered run cannot pass with a 0%
error rate.
Fail loudly instead of silently:
- run-full-validation: treat a consensus timeout and a missing validated
ledger as fatal infrastructure errors, and fold the orchestrator and
benchmark exit codes into the final status. A degraded cluster previously
ran a full validation pass and reported misleading downstream failures.
- collect_system_metrics: warn per empty measurement source, emit
metrics_complete, and exit non-zero instead of substituting zeros that
pass every threshold. Require GNU date with %N rather than falling back
to a per-sample python3 fork that costs more than the threshold it is
measured against.
- benchmark: distinguish "could not measure" from "exceeded thresholds",
install a cleanup trap so a failure cannot leak nodes and ports, and
report an unusable baseline as inconclusive.
- workload_orchestrator: bound subprocess communicate() and fail the exit
gate on per-phase errors.
Also pins the workload compose images to the versions the sibling stack
already uses, hash-pins the Python dependencies, restricts the validator
config template to loopback, corrects the dashboard and metric counts in
the reference docs, drops a span from the regression gate that cannot fire
under a WebSocket-only workload, and narrows the teardown pkill pattern so
it no longer matches processes that merely mention the work directory.
Verified with a full harness run against a local five-node cluster:
158 of 158 checks passed with no regressions detected.
The Telemetry Validation workflow failed with three "0 series" checks:
rpc_method_us, job_queued_us and job_running_us. All three are Histograms,
and the Prometheus exporter emits a histogram only as the
_bucket/_count/_sum triple -- the bare instrument name is never a series,
so validate_metrics() could never match it.
Evidence from the failing run (31804450127): its own metric-name dump
lists rpc_method_us_bucket/_count/_sum and no bare rpc_method_us, while
the sibling counters recorded in the same function bodies passed with 100
and 67 series. capture_timings.py, which queries job_queued_us_bucket and
job_running_us_bucket, returned real values for the acceptLedger job type
in that same run. Every one of the 10 histograms present exposes the full
triple, so all three suffixes are safe to assert.
Name them the way the exporter does, matching what the spanmetrics group
above already does for span_duration_milliseconds and what
regression-metrics.json and the job-queue dashboard already query. The
metrics stay in their asserted groups because they are genuinely
unconditional, so `not_asserted` would be wrong.
`.github/scripts/rename/docs.sh` is a migration script, not a check: the
`check-rename` workflow runs it and then fails if the tree changed. Its
rule `s@([^/+-])rippled@\1xrpld@g` rewrites `rippled` unless the preceding
character is `/`, `+` or `-`, so the backtick-quoted historical filenames
added by 3153f3ef56 (`` `rippled-*.json` ``, `` `rippled-alerts.yml` ``)
were rewritten and the check failed on both plan docs.
Refer to those historical names by their path-qualified form
(`dashboards/rippled-*`, `prometheus/rippled-alerts.yml`) instead. The
leading `/` exempts them from the rule, which is the form the rest of the
repo and the surrounding lines in these same files already use. No meaning
is lost -- the notes still record which commit renamed what.
tasks/fix-validation-checks.md was committed with the phase-10 harness and
shows up in the PR diff. tasks/ is a scratch area -- roughly forty analysis and
plan documents live there locally and none of the others are tracked, on this
branch or on develop. The file stays on disk; only the tracking is removed.
Every telemetry build instruction told the reader to run
`cmake --preset default`. No `CMakePresets.json` is tracked in the repo, and the
only preset that exists is `conan-release`, generated into the gitignored
`CMakeUserPresets.json` by `conan install`. So the documented first build step
fails before it starts, in the runbook's quick-start, its disable procedure, and
the telemetry stack's testing guide.
Replaced with the flow BUILD.md actually documents -- `conan install ..
--output-folder .` followed by an explicit `conan_toolchain.cmake` invocation --
with `-o telemetry=True` / `-Dtelemetry=ON` as the telemetry delta, and a note
that `--preset conan-release` works as a shorthand.
The harness manifests asserted things the code cannot produce and missed most
of what it does. Two assertions were failing every run, and the metric set
covered 16 of the ~41 emitted names.
expected_spans.json: rpc.process was required with rpc.ws_message as its
parent, but it is created only in ServerHandler::processRequest() on the HTTP
path, so a WebSocket-only workload never produces it -- it is now optional and
parented to rpc.http_request, and the rpc.process -> rpc.command.* edge is
skipped with the real reason instead of a coroutine-context-loss diagnosis that
was never the cause. Adds the missing rpc.ws_upgrade span, corrects four
parents (consensus.mode_change, pathfind.request, and update_positions/check,
which are children of consensus.establish rather than consensus.round), and
demotes conditionally-set attributes out of required_attributes so a healthy
run stops failing. Counts recomputed from the file: 41 span types, 62 unique
required attributes.
expected_metrics.json: 16 -> 52 asserted entries across the job-queue, RPC
method, reduce-relay, overflow and validation families, plus the fifteenth
dashboard uid. Metrics the harness workload cannot exercise -- erroring RPC,
ledger-mismatch, TxQ overflow, and the lazily-created getobject_* instruments
-- are listed in a not_asserted group the validator skips, rather than as
assertions that would fail on a healthy node.
The workflow's push trigger listed two globs matching nothing
(include/xrpl/basics/Telemetry*.h, src/xrpld/app/misc/Telemetry*), so no C++
telemetry change ever triggered validation. Replaced with the paths the code
actually lives in, including src/libxrpl/beast/insight/** for the insight
export path the harness depends on. The four inert workflow_dispatch inputs are
now labelled UNUSED rather than looking like working knobs.
Docs: the workload README described a StatsD dirty-flag mechanism under a
member name that does not exist, on a code path the harness never uses -- it
sets [insight] server=otel, so gauges export through an observable-gauge
callback every cycle. Adds the missing txq-burst phase, reconciles three
different dashboard counts, and drops "posts summary to PR", which the workflow
has no permission to do. The runbook's phase-10 section loses the last
sampling_ratio reference (not a config key), gains a Regression Gate and CI
subsection covering the gate that can fail CI, and its compose-logs command now
names the workload compose file. cmake --preset default is left for a separate
change: no CMakePresets.json is tracked, so it is wrong everywhere it appears.
Also drops the dead exporter=otlp_http key the harness wrote into every node
config, and stops capture_timings.py defaulting --profile to a profile that
does not exist.
Three doc conflicts, all inside Phase 10's own sections. Resolved by keeping
phase-9's corrections and re-applying phase-10's ownership of those sections on
top, rather than taking either side wholesale:
06 §6.8.3 — Status now says Phase 10 is implemented on THIS branch (phase-9's
copy said "not merged into this branch", true there, false here). The stale
"71 checks" enumeration and "Current Status" list are replaced by phase-9's
dynamic-count description plus a Known Gaps list that states the real reason
the rpc.process hierarchy is unassertable: rpc.process is HTTP-only, so a
WebSocket-only workload never produces it. Phase-9's CI Deliverable subsection
is carried through; exit criteria keep phase-10's tick state with phase-9's
corrected wording.
06 §6.8.3 Architecture — the automatic merge had resurrected phase-10's stale
2-node cluster prose and diagram over phase-9's 5-node correction. Reconciled
to phase-10's topology (native xrpld processes vs the containerised backend,
which is accurate) with phase-9's node count: 5 validators, and the collector
labelled OTLP + filelog rather than StatsD, which the config has never had.
Dropped the "all 26 metrics required" label in favour of the manifest.
06 §6.8.3 Key Implementation Details — two claims corrected against the code.
The StatsD m_dirty gauge fix describes a member that exists nowhere in the
repo, and the harness sets [insight] server=otel anyway, so gauges export
through an observable-gauge callback. The tx.receive attribute keys are bare
suppressed and tx_status, not dotted, and tx_status is set only on the
reject/known-bad/dropped paths, so it is absent on a successful receive.
09 §5c — kept phase-10's four-column table shape, with phase-9's corrected
counts: 40 of 41 emitted spans, 67 required attributes, 14 of 15 dashboards.
Phase10_taskList — dynamic inventory totals, the real RPC span trees, and exit
criteria ticked where the code on this branch closes them. Per-RPC timings are
recorded as not gated: regression-metrics.json defines only spans and job_queue.
Verified every hunk of the phase-10 diff falls inside a Phase-10-owned section
(06 §6.8.3, 09 §5c, Phase10_taskList, and the runbook's appended Phase 10
sections); no phase-9-owned text is modified from this branch.
The reference docs had drifted from the code in ways that break the reader
rather than merely misinform: PromQL examples that return no data, a rollback
flag that is a no-op, a sampling knob that does not exist, and two span parents
that moved. Code is treated as the truth throughout; where the code is the
defective side, the doc now records it as a known issue instead of describing
the bug as intent.
Renames the docs missed: histogram names gain the exporter's unit suffix
(ios_latency_milliseconds_bucket and four siblings), ledger_history_mismatch
gains _total, the StatsD-era quantile label gives way to le buckets,
rpc.request becomes rpc.http_request, traces_spanmetrics_calls_total becomes
span_calls_total, and the nine dotted xrpl.* span attributes are recorded as
renamed rather than left as live keys.
Re-parenting: consensus.update_positions and consensus.check are children of
consensus.establish, not of consensus.round.
Units and labels: state_accounting_*_duration is microseconds, not seconds;
cache_metrics label values are case-sensitive; object_count carries demangled
C++ type names. Nodestore read and write latency stays microseconds -- the
nanosecond accumulator change did not move the exported unit.
Adds what shipped but was undocumented: the ledger.acquire span, seven
consensus.round events, twelve span attributes, node_writes_duration_us, the
7-day validation-agreement window, the TxQ admission and reduce-relay metric
families, metrics_endpoint, and the phase-10 validation workflow.
Corrects claims that never held: 10% head sampling (it is fixed at 100%),
configurable redaction (it is unconditional), -DXRPL_ENABLE_TELEMETRY=OFF
(the flag is -Dtelemetry=OFF, default ON), FindOpenTelemetry.cmake and the
xrpl_telemetry target (neither exists), Promtail and a StatsD exporter in the
pipeline (neither exists), and Loki stream selection on job= (only
service_name is a stream label).
Phase 9 is marked complete, its provisioned alerting is attributed to the
branch that shipped it, and Phase 11 stays at zero except the one prerequisite
its code closes. Counts are reconciled repo-wide: 41 emitted span families,
15 dashboards on disk with 14 asserted, 13 alert rules in 5 groups.
Hardens the gate that let this drift through: Rule E of the naming check now
covers the reference docs, its allow-dotted marker is key-scoped and warns on
stale or empty use, a missing checked file is reported instead of silently
skipped, the test suite runs in CI, and doc paths trigger the check.
C++ and CMake changes are comment-only: three MetricsRegistry instrument names,
eight OTelCollector claims of a metric-name prefix that formatName never adds,
and the telemetry option's inverted default.
Add two Peer Quality panels reading peer_disconnect_total: Peer Disconnect
Rate, the per-second teardown rate per node, and Peer Disconnects By Reason
& Direction, the per-interval increase split by cause and by which side
opened the connection. Both sit in the existing Disconnects & Connection Mix
row beside Resource Disconnects, which counts only the resource-charge
subset and carries no reason label.
The Ledger Sync Health board already shows the same split as a window
total, so it says how much of each reason but not when. These give the
time-shaped view, letting a reason spike be lined up against a stall.
Add disconnect_reason and disconnect_direction template variables for the
two new label dimensions and wire both queries to them, so the panels
filter on every dimension their series carry.
Update the 09 reference panel column and the _a7_note panel list to name
the panels that now render this counter.
Two runbook readings were misleading in ways that hid a real devnet issue.
The Online-Delete Rotation Window row said a healthy node shows the flag
"briefly 1 once per delete interval". That is true but not sufficient:
rotation_state{metric="in_flight"} is set immediately before
freshenCaches() and cleared by RotationExposureGuard on scope exit, so it
brackets only the freshen/swap phase. Measured on devnet-otel-usw2-01/02
with online_delete=256 and ~47.4M state nodes, the flag averaged
0.159/0.135 over 9 h while the node was inside a rotation ~93% of wall
clock, because the dominant visitNodes copy phase (median 651 s of an
~785 s cycle) emits nothing. The row read healthy on a rotation-bound
node. Add a scope note with the log triplet that does measure occupancy.
The Consensus Round Duration panel gained a p99 series, so retitle both
references and say when to read it: at ~19 rounds/min a single stalled
round is one sample in several hundred, so p95 stays at the normal close
time and can dip. An 11.4 s round measured p99 13400 ms while p95 read
3400 ms against a 2900-3787 ms baseline.
Instrumenting the copy phase is tracked in RIPD-7144; the underlying
rotation defects in RIPD-7720.
The "Consensus Round Duration" panel plotted only p50 and p95. At the
devnet rate of ~19 rounds/min, a single stalled round is one sample in
several hundred, so p95 over a rate interval stays near the normal 2s
close time and the stall is invisible.
Measured on devnet-otel-usw2-02 over a 40 min window: 761 rounds, of
which exactly one fell in the 7500-10000ms bucket while the mean never
exceeded 2270ms. The histogram already records the outlier; only the
panel could not show it.
Add a p99 series alongside p50/p95 and retitle the panel accordingly.
No metric or collector change: the existing
consensus_round_duration_ms_bucket histogram already has 20 boundaries
up to 120000ms.
The Peer Disconnect Rate By Reason panel anchored its LogQL capture on
"\] ", which only matches a reason logged immediately after the [NNN]
peer-id prefix. PeerImp does not log that way: PeerImp::fail emits
"[NNN] <name> failed: <reason>" and the clean teardown emits
"close: Closed". Only ConnectAttempt::fail, which logs the bare reason,
ever matched. The panel's Timeout series was therefore connect-attempt
timeouts only, Ping Timeout was invisible, and PeerImp's own Closed was
uncounted.
Match all three prefixes and separate Ping Timeout from Connect Timeout.
Recorded as LogQL trap 11 in the runbook, alongside the other silent
failures this dashboard exposed.
Also document six overlay observability gaps found while auditing what
ping/pong and gossip traffic is actually tracked. All are pre-existing
and none is fixed here: the code fixes belong in develop-owned overlay
files (TrafficCount, OverlayImpl, PeerImp, PeerfinderManager), not on a
telemetry branch, and one of them needs a public signature change.
- 09 section 6: six known issues, each marked NOT IMPLEMENTED with
file:line evidence -- mtCLUSTER counted as unknown (overhead_cluster_*
always zero, 8 panels flatline), squelch_ignored byte counts always
zero, inbound/outbound byte-basis asymmetry plus a stale Total header
comment, ping/endpoints instrumentation absent, peer span coverage,
and PeerFinder exporting 2 of ~17 available readings.
- 02 section 2.3.2: add a Status column to the span catalog. Of 36
catalogued spans, 16 are live, 15 were never built, and 5 shipped
under different names (consensus.phase.establish became
consensus.establish, ledger.close became consensus.ledger_close,
rpc.request split into rpc.http_request/rpc.ws_message, txq.apply
became txq.apply_direct/txq.accept_tx). The catalog is a design
inventory; 09 section 1.1 remains authoritative for what emits.
- Phase9_taskList: tasks 9.14-9.17 tracking the deferred work, with
exit criteria checked only for what is actually done.
- Glossary: new Ping / pong keepalive term distinguishing ping timeout
from connect timeout. Correct the Cluster and Squelch entries, which
described behaviour the metrics cannot show.
The glossary header pointed at tasks/telemetry_terms.py as its
generator. That file is in no branch and nowhere on disk -- tasks/ is
gitignored one directory up -- so the header now states the file is
hand-maintained and gives the entry format.
Gates: check_otel_naming.py passes all 9 rules (Rule D over 555
dashboard queries, Rule E over the runbook); 19 doc anchors verified;
dashboard JSON valid with a one-line diff. No C++ changes.
Resolves the nodestore latency-accumulator conflict by keeping this
branch's API surface and applying the incoming nanosecond fix to it.
This branch had renamed storeDurationStats() to recordStoreDuration() and
given it noexcept, an explicit relaxed fetch_add, and call-frequency and
thread-safety notes. The incoming change fixed the accumulators to hold
nanoseconds so sub-microsecond backend calls stop rounding to zero. Both
are kept: the name, signature and documentation from here, the nanosecond
accumulation from there.
The incoming duplicate getFetchDurationUs()/getStoreDurationUs() pair is
dropped in favour of this branch's noexcept versions, which now read the
nanosecond members and convert on read. recordStoreDuration()'s `if (us >
0)` guard is gone: it discarded every sub-microsecond store, which is the
same rounding the incoming change removes one level up.
Also updates a MetricMacros comment that explained the zero write mean in
terms of that removed guard; the injected totals it asserts on are
unchanged.
The fetch and store duration counters converted each sample to
microseconds before adding it, so any backend call finishing in under a
microsecond contributed zero. A warm nudb read answers in a few hundred
nanoseconds, so on fast hardware every read floored and the totals stayed
at zero no matter how many reads happened -- the same loss of resolution
the microsecond report was introduced to avoid, one decade lower.
Both accumulators now hold nanoseconds, the clock's own resolution, and
convert once in getFetchDurationUs() and getStoreDurationUs(). The public
accessors, the node_reads_duration_us and node_writes_duration_us JSON
fields, and the metrics that read them all keep microseconds, so nothing
downstream changes unit. storeDurationStats() takes the raw duration
instead of a pre-converted integer so no caller can round early, and
updateFetchMetrics() scales its microsecond input to match.
FetchReport::elapsed stays microseconds: it carries one fetch, not a
total, and that is the unit it declares. The reported sum is therefore the
accumulated total minus a sub-microsecond remainder per fetch, so the two
tests that asserted exact equality between them now assert that bound.
Both had assertions that depended on how fast the host reads; the bound
holds on any hardware.
* upstream/release/3.3.x: (41 commits)
chore: Bump version to 3.3.0
chore: Bump version to 3.3.0-rc7
fix: Increase manifest protocol message size cap and fix manifests relay
fix: Cap untrusted manifests per message and drop oversized ones
chore: Bump version to 3.2.1
chore: Bump version to 3.2.1-rc1
fix: Cap untrusted manifests per message and drop oversized ones
fix: Reject oversized validator manifest before decoding
fix: Reduce untrusted manifest cache cap to 100
fix: Bound untrusted manifest cache
chore: Bump version to 3.3.0-rc6
feat: Package validator-keys inside rippled
chore: Bump version to 3.3.0-rc5
fix: Switch SponsorshipSet to use a delta for sfFeeAmount
fix: Re-revert "fix: Set request size limits and differential pricing for get-object-by-hash calls"
chore: Bump version to 3.3.0-rc4
fix: Revert "fix: Set request size limits and differential pricing for get-object-by-hash calls"
chore: Bump version to 3.3.0-rc3
fix: Reduce untrusted manifest cache cap to 100
fix: Revert "fix: Reject oversized SHAMap nodes in gotStaleData and fetch-pack path"
...
The debug-log notice on Log-Derived Insights was two grid rows taller than its
content needs, pushing the first data row further down than necessary. Rendered
the panel to confirm the markdown still fits with no clipping.
It is the first panel in the dashboard, so no other panel's stored y needed to
change - Grafana's vertical compaction closes the gap.
The dashboard's sections and panels were rearranged directly in Grafana. The
in-repo copy still carried the old arrangement, so the next replace-upload would
have undone that work.
Panel and row positions (y, and x for non-repeating panels), plus panel ids, now
come from the live copy. Ids matter because the uploader keys on them: matching
them makes an upload an in-place update rather than a reshuffle.
Repeating panels stay stored at w=12, x=0 rather than adopting the live w=24. For
a horizontal repeat Grafana ignores the stored width and lays each copy out at
24/min(N, maxPerRow), so the two are indistinguishable on screen - verified by
rendering a single-network view, where a w=12 repeat still fills the row. w=12 is
the honest description of one panel at half width.
Verified: panel and row order identical to live, rendered geometry identical at
two networks (what the live stack runs), no panel rendering with an empty column
band to its left at one through four networks, and every query, description and
panel option untouched.
Carries the layout fix to the dashboards this branch owns. A repeating panel
claims the whole row once it expands per network, so the non-repeating panel
paired beside it was pushed down while keeping its stored x=12 - rendering on
the right with an empty gap on its left.
Repeating panels now get a row to themselves, keeping w=12 so their copies still
tile two across. Single-value panels are grouped to the top of each row section
so the charts that follow pair with each other rather than being separated by an
interleaved repeat.
Verified against origin/phase9: no panel lost, every targets block unchanged,
ids 1..N, no overlaps, no rows with a left-hand gap.
A repeating panel expands into one copy per network at view time and, with
maxPerRow=2, claims the whole row. The non-repeating panel paired beside it was
pushed to the next line but kept its stored x=12, so it rendered on the right
against an empty gap.
Two changes to how the layout is planned:
- a repeating panel gets a row to itself. It keeps w=12, so its copies still
tile two across inside that row.
- single-value panels are grouped to the top of each row section, so the
charts that follow pair with each other instead of being split up by an
interleaved repeat. Without this the gaps just become wasted half-rows.
Verified per dashboard: no panel lost, every targets block byte-identical, ids
1..N, no overlaps, and no row left with a gap on its left.
Panels were appearing on the right of a row with nothing to their left. A
repeating panel expands into one copy per network at view time and, with
maxPerRow=2, claims the whole row; the non-repeating panel paired beside it was
pushed to the next line but kept its stored x=12, so it rendered on the right
against an empty gap. Thirteen rows here were affected - more than every other
dashboard combined, which is why it showed up on this one first.
Two changes:
- a repeating panel is given a row to itself when the layout is planned. It
keeps w=12, so its copies still tile two across inside that row.
- single-value panels are grouped to the top of each row section, so the
charts that follow can pair with each other instead of being split up by an
interleaved repeat. Without this the gaps become wasted half-rows.
Result: no mixed rows, no left-hand gaps, 13 paired rows and only 3 panels
sitting alone, from sections with an odd number of charts.
Verified: 54 panels before and after, no panel lost, every targets block and
description byte-identical, ids 1..N, no overlaps.
Brings this dashboard in line with the other fifteen, which were normalized
together:
- id written as 1..N so panelId deep links address a specific panel
rather than whichever one happens to sit in that position
- gridPos at most two panels across; every visualisation h=10. Three
panels sat at w=8, the only three-across row in the set.
- line charts lineWidth=1, fillOpacity=0, pointSize=5, gradientMode=none
- repeat xrpl_network_type (horizontal, maxPerRow=2) with a
[$xrpl_network_type] title suffix on the stat and bargauge
panels, where two networks overlaid in one panel reads as
noise. Charts keep their networks as separate series.
The nine existing row groups are kept as they are. The heatmap already carried
the full plugin option set, so it is untouched.
Raw-text edits, so untouched lines keep their formatting byte for byte.
Verified: 54 panels before and after, no panel lost, every targets block and
description byte-identical, ids 1..N, and no two panels overlapping.
The GetObject Request Size Distribution heatmap, added on this branch, was
missing calculate, color and cellGap. Grafana's heatmap plugin needs them, and
without them the whole dashboard opens with "An error occurred within the
plugin" instead of rendering.
Same values as the other four heatmaps fixed upstream, taken from the one that
renders correctly on Grafana Cloud. The panel keeps its own axis label and unit.
Every heatmap carried only `tooltip` and `yAxis`, missing `calculate`, `color`
and `cellGap`. Grafana's heatmap plugin treats those as required, and without
them the panel fails to initialise: the dashboard opens with "An error occurred
within the plugin" rather than a chart.
The option values are taken from the one heatmap in this stack that does render
on Grafana Cloud (ledger-sync-health): calculate=false since the queries already
return histogram buckets, the Turbo 64-step scheme, and cellGap=1. Each panel
keeps its own yAxis label, unit and tooltip settings.
This is a long-standing defect rather than fallout from the recent layout work -
the same options are absent in origin/phase9 and in the cloud copies that were
already live.
Carries the sizing fix through to the dashboards this branch owns, after the
live Cloud copies showed charts clipping their legends at h=8 and stats too
cramped at h=4.
Every visualisation is h=10; tables and logs h=12; the log-derived-insights
instruction banner h=12 for its prose. A single height per visualisation also
means any two panels pair cleanly side by side.
Repeat stays on single-value panels only. The five state-timeline and heatmap
panels that had picked it up in an earlier pass are cleared, since repeating a
timeline broke ledger-data-sync on Cloud.
Verified against origin/phase9: no panel lost, every targets block unchanged,
ids 1..N, no grid overlaps.
Two problems showed up once these dashboards were live on Grafana Cloud.
Panels were too short. Charts at h=8 clipped their legends mid-row, and stats
at h=4 were cramped. Every visualisation is now h=10, with tables and logs at
h=12; the log-derived-insights instruction banner keeps h=12 for its prose.
Uniform height also means any two panels can pair side by side.
Repeat on a state-timeline broke the dashboard outright: ledger-data-sync
failed to open on Cloud at v45 and had to be restored to v44. Repeat is now
limited to single-value panels (stat, gauge, bargauge, table). Charts show
their networks as separate series instead, which is what a chart is for.
Repeat was also sticky: normalization only ever added the keys, so a panel that
picked them up in an earlier pass kept them even after its type stopped being
eligible. The keys and the title suffix are now removed from ineligible panels,
which is what actually cleared the two timeline panels here.
Verified per dashboard: no panel lost, every targets block byte-identical, ids
1..N, no grid overlaps, and repeat present only on stat/gauge/bargauge.
Applies the guideline-8 ordering to the two dashboards introduced on this
branch, and clears the last two labelling gaps found while auditing the set:
- peer-quality, validator-health: stats and bar gauges now lead their row
section instead of trailing the charts, matching the other dashboards.
- node-health "Peer Disconnects (Resources)": had an empty axisLabel, the
only chart in the set without one. Now reads "Disconnects".
- rpc-performance row "Aggregate RPC (all commands)" -> "(All Commands)",
the one title that was not in Title Case.
Reordering stays inside each row section so no panel changes category, and the
panel objects are cut and re-spliced as raw text, so their contents are
byte-identical. Verified against origin/phase9: no panel lost, every targets
block unchanged, ids 1..N, and every row section now leads with its
single-value panels.
Guideline 8 asks for gauges and stats at the top. Seven dashboards had them
scattered below charts, so the reader met a wall of time series before the
at-a-glance numbers that give those series context.
Stats, gauges and bar gauges now come first within each row section. The move
is deliberately scoped to inside a section: shifting a panel across a row
boundary would change which category it belongs to. Panels keep their relative
order otherwise, so the reading sequence within each group is unchanged.
Whole panel objects are cut and re-spliced as raw text, so their contents stay
byte-identical and only gridPos and id are recomputed. Verified per dashboard:
panel count unchanged, no panel lost, every targets block byte-identical, ids
still 1..N, and no row section left with a stat below a chart.
Applies the same normalization the phase-7-owned dashboards received, to the
five introduced on this branch: fee-market, job-queue, log-derived-insights,
peer-quality and validator-health.
- id written as 1..N so panelId deep links address a specific panel
rather than whichever panel happens to sit in that position
- gridPos at most two panels across; charts h=8, stats/gauges h=4,
tables/logs h=12 full width. Panels pair only with an
equal-height neighbour, so no row keeps a ragged empty cell.
validator-health and peer-quality were stacked single-file at
full width, which is why they were the longest to scroll.
- line charts lineWidth=1, fillOpacity=0, pointSize=5, gradientMode=none
- repeat xrpl_network_type (horizontal, maxPerRow=2) with a
[$xrpl_network_type] title suffix on stat/gauge/bargauge/
table/state-timeline panels, where two networks overlaid in
one panel reads as noise. Line charts keep their networks as
separate series.
- decimals 0 on panels counting discrete things (peers, quorum,
connection mix) - a fractional peer count is meaningless.
- rows category rows added to fee-market, job-queue and peer-quality
Panels with a right-hand legend stay full width; a side legend needs the room.
Raw-text edits, so untouched lines keep their formatting and escaping byte for
byte. Verified per dashboard against origin/phase9: panel count unchanged, no
panel lost, every targets block byte-identical, descriptions unchanged, ids
exactly 1..N, and no two panels overlapping on the grid.
These dashboards were hand-authored over time and had drifted apart: panel
heights spanned ten different values, nine dashboards had no row grouping,
line-chart styling was inconsistent, and no panel carried an id, so Grafana
assigned them positionally at load and every panelId deep link was only as
stable as the panel order.
Per panel, in document order:
- id written as 1..N so panelId links address a specific panel
- gridPos quantized to at most two panels across: charts h=8,
stats/gauges h=4, tables/logs h=12 full width. Panels are
paired only with an equal-height neighbour, so no row is left
with a ragged half-empty cell.
- line charts lineWidth=1, fillOpacity=0, pointSize=5, gradientMode=none
- repeat xrpl_network_type (horizontal, maxPerRow=2) with a
[$xrpl_network_type] title suffix, on the panel types where
overlaying two networks in one panel reads as noise
(stat/gauge/bargauge/table/state-timeline). Line charts keep
their networks as separate series, which is the point of a
line chart.
- decimals 0 where the value counts discrete things (threads, peers,
queue depths); a fractional thread count is meaningless.
- rows category rows added where a dashboard had none
Panels whose legend sits on the right stay full width: a side legend needs the
horizontal room, and squeezing it to half width clips the series names.
Edits were made as raw-text replacements, not a json.dump round-trip, so
formatting and escaping of untouched lines are byte-identical. Verified per
dashboard: panel count unchanged, every targets block byte-identical, all
descriptions unchanged, ids exactly 1..N, and no two panels overlapping on the
grid. The repo dashboard lint and the OTel naming check both pass.
Four defects found by a query-correctness audit of all 16 dashboards, all
scoped to panels this branch owns.
job-queue "Current Job Latency (p99 Gauge)": the histogram by-clause dropped
service_instance_id, collapsing every node into one fleet-wide p99 and hiding
a slow node. Measured: gauge read 1325us while the worst node was 1868us. The
panel's displayName already referenced service_instance_id, so it was also
rendering an empty label. Now groups by service_instance_id and xrpl_work_item,
matching sibling panels 5, 6 and 7.
ledger-data-sync "NodeStore Read Latency (Bottleneck Discriminator)": replaced
clamp_min(<denominator>, 1) with (<denominator> > 0). clamp_min clamps the
value, not just the zero case, so any node reading below 1/s got a fabricated
denominator. Demonstrated with a zero-rate denominator: clamp_min invents
40.2/40.3/11.2/9.2 where the > 0 guard correctly returns no data. This panel is
the bottleneck discriminator, read during a stall, which is exactly when the
read rate collapses and the clamp is most wrong. Panels 21 and 23 carry the
same defect but originate on phase-7 and are fixed there.
node-health thresholds: percentunit fields are compared against the raw value,
so a step of 80 needed 8000% and could never fire. Rescaled panels 74, 81 and
85 to 0.8. Panel 81 is a found-ratio where high is healthy, so its bands were
also inverted. Note these three panels use palette-classic with thresholdsStyle
off, so the steps are currently dormant rather than visibly wrong.
node-health panels 81 and 85 descriptions: both described the multi-series
panels they were split from. Panel 81 carried a byte-identical copy of panel
80's text, promising three plotted rate lines where it draws a single bounded
ratio; panel 85's text described read-thread gauges absent from its expression.
Rewritten to match the actual queries.
The Transaction Overview panel "Queue Bypass Ratio (Direct Apply vs
Enqueue)" reported a confident 0.50 on every node while the true bypass
rate was zero. The two spans it divided are not disjoint alternatives:
txq.apply_direct is a child of txq.enqueue. TxQ.cpp creates the
apply_direct span as the first statement of tryDirectApply(), ahead of
the account, sequence and fee-level guards, and tryDirectApply() is
called from inside the live enqueue scope. The span therefore counts
attempts, so the denominator direct + enqueue counts each transaction
twice and pins the ratio to one half algebraically.
Measured on a four-node fleet: 6082443 direct against 6082877 enqueue
over the same population, panel output 0.5000170 on three nodes and
0.5000000 on the fourth. Grouping txq.enqueue by txq_status over seven
days returns only "rejected" -- no transaction has ever taken the
direct-apply path.
Remove the panel rather than repoint it. A correct expression using
txq_status as the disjoint discriminator would render permanently
empty on this fleet, which reads no better than a wrong number.
Widen the band partner "TxQ Enqueue Rate by Transaction Type" from 12
to 24 columns so the y=48 band still fills the grid. Every band in all
ten dashboards sums to 24 columns; leaving a half-width hole would be
the only exception. Panel order and every other panel's position,
width and height are unchanged.
The runbook already listed txq.apply_direct as available but not
paneled, so that row becomes accurate. Rows describing the span itself
are untouched -- the span and its metric are unchanged.
Panels 21 (NuDB Read Latency) and 23 (NuDB Read Found Ratio) on the
Ledger Data & Sync dashboard guarded their divisor with
clamp_min(<denominator>, 1). clamp_min raises the value, not just the
zero case, so any node reading fewer than 1 block per second was
divided by a fabricated 1 instead of by its real read count.
Replace the clamp with the filter (<denominator> > 0). A comparison
without the bool modifier drops the sample rather than rewriting it, so
these panels now show no data instead of a wrong number.
Measured over 7 days: five nodes fall below 1 read/s. On validator-0 the
clamp reported 2.726 us/read against a true 5.493, and on nonUNLmalloc-tc
it reported 0 us/read, which cannot occur. The error is largest exactly
when panel 21 is used as the bottleneck discriminator during a stall,
because that is when the read rate collapses toward zero.
Matches the existing idiom on the same nodestore_state metric family in
the NodeStore Write vs Read Latency panel.
Grafana 13 enables the renderAuthJWT feature toggle by default, so the image
renderer silently drops render requests carrying the default token and alert
notifications arrive with no panel image. The runbook's alerting section
covered delivery but not this failure mode.
Adds how the token is wired (one variable feeding both services, so they cannot
drift), how to override it, and the two-command check for when images stop
appearing.
The renderer was the only service in the stack on a floating :latest tag, so
the stack was not reproducible: a renderer release could change under a
checkout that had not been touched. Pins it to v5.12.0, matching how every
other service here is pinned.
Also wires the shared render token. Grafana 13 enables the renderAuthJWT
feature toggle by default, and the renderer rejects requests unless both sides
present the same non-default token, so pinning alone would leave rendering
broken once the Grafana bump merges forward. Both values come from the same
GF_RENDERING_RENDERER_TOKEN variable and cannot drift apart; the local default
keeps `docker compose up` working with no setup.
Verified `docker compose config` resolves and both AUTH_TOKEN and
GF_RENDERING_RENDERER_TOKEN expand to the same value.
Bumps the Loki pin this branch introduced, and updates the two docs that
name the version so they do not drift from the compose file:
- OpenTelemetryPlan/09-data-collection-reference.md (log backend section)
- OpenTelemetryPlan/Phase8_taskList.md (compose snippet)
The service runs the config bundled in the image rather than one from this
repo; `-verify-config` reports "config is valid" under 3.7.6, and native OTLP
ingestion at /otlp is unchanged across 3.4 -> 3.7.
Moves off the v2.53 LTS line, which this branch pinned, onto the current
v3 release, and corrects the accompanying comment that named the old line.
Verified against the new image rather than assumed, since this crosses a
major version:
- prometheus.yml passes `promtool check config` under v3.13.2
- all 391 unique dashboard PromQL queries parse under the v3 parser
- all 13 alert-rule expressions parse under the v3 parser
The config uses none of the surfaces v3 changed: no remote_write, no
holt_winters, no offset modifiers.
Bumps the three images this branch pins:
otel-collector-contrib 0.121.0 -> 0.158.0
tempo 2.7.2 -> 2.9.4
grafana 11.5.2 -> 13.1.2
Verified locally against the new images: the collector config passes
`validate` under 0.158.0 and tempo.yaml passes `-config.verify=true` under
2.9.4, both unchanged. The collector's log path uses the generic otlphttp
exporter (otlphttp/loki), not the dedicated loki exporter removed upstream in
v0.147.0, so the pipeline is unaffected by that removal.
Grafana crosses two majors. Operator-visible consequences are handled on the
branches that own the affected files: Grafana 13 enables the renderAuthJWT
feature toggle by default, so the image renderer now requires a matching
renderer_token on both the server and the renderer container.
The Mainnet telemetry config carried ssl_verify=0, inherited when the file
was copied from the Devnet config. Mainnet fetches its validator lists from
public HTTPS publishers (vl.ripple.com, unl.xrplf.org) that present valid CA
certificates, so there is no self-signed certificate to work around.
With ssl_verify=0 both halves of TLS peer authentication are skipped on those
fetches: the certificate chain check and the RFC 6125 hostname match. The list
blob's own signature is still verified against the pinned publisher keys, so
this restores defense in depth rather than fixing a validator-injection path.
Restores the secure default and documents why it stays that way.
develop renamed CamelCase namespaces to snake_case with no compatibility
aliases, so lines this branch added still referenced the old names. The
merge was textually clean because develop never touched those lines, which
left the breakage invisible to git and visible only at compile time.
PeerFinder:: -> peer_finder:: Overlay.h, OverlayImpl.h, MetricMacros.cpp
Tuning:: -> tuning:: PeerImp.cpp (the other call sites in this
file already used the lowercase form)
Two further merge artifacts:
json_value.h develop added `using value_type = Value const` to
ValueConstIterator, and this branch had added
`using value_type = Value` for the same reason, so the merge
kept both and they redefined the alias with different types.
Keep develop's const-qualified type -- correct for an
iterator whose reference is `Value const&` -- and keep only
this branch's iterator_category addition.
InboundLedger.h <string> is no longer used directly by the header, but
five translation units that include it use std::string
without including <string> themselves. Mark it
`IWYU pragma: keep` rather than remove it, matching
suite_list.h.
develop renamed xrpl::Resource to xrpl::resource. MetricsRegistry.cpp
declared `namespace resource = opentelemetry::sdk::resource` at file
scope, but both use sites are inside namespace xrpl::telemetry, where
inner-scope lookup finds the enclosing xrpl::resource first and never
reaches the alias. That namespace has no ResourceAttributes or Resource,
so the build failed on all four platforms:
error: no type named 'ResourceAttributes' in namespace 'xrpl::resource'
error: no member named 'Create' in namespace 'xrpl::resource';
did you mean 'creat'?
Rename the alias to otel_resource. Moving it inside xrpl::telemetry would
also compile, but only by shadowing xrpl::resource -- a reader at the use
site still could not tell which namespace `resource::` meant, and a later
using-directive would reintroduce the ambiguity. A distinct name removes
the collision by construction and matches the metric_sdk / otlp_http
aliases already in this file.
Rule D validated every dashboard label against L1 (*SpanNames.h) and L6
(MetricsRegistry) labels. LogQL labels have a third provenance neither
layer can resolve: they are minted by the collector's regex_parser named
captures (partition, severity) or by an in-query `| regexp` stage
(action, pk, state, mode, phase, jobname, ip, pubkey). Checking them
against L1/L6 reported ten violations for labels correct by
construction.
Make the rule datasource-aware instead of allowlisting a filename. The
dashboard JSON is parsed so each query can be attributed to its
datasource, and queries on a log datasource are skipped. The exemption is
per query, not per file, so a dashboard mixing Prometheus and Loki panels
still has its Prometheus panels validated.
Parsing the JSON also fixed a blind spot: label filters are stored with
backslash-escaped quotes (`label=~\"$v\"`), which the previous raw-text
regex could never match, so only the `sum by (...)` form was ever
checked. With the strings unescaped, 555 queries are now validated where
far fewer were before. That surfaced three legitimate label sources the
rule did not model, each fixed at its source rather than allowlisted:
- deployment_environment / xrpl_network_type: resource attributes the
collector promotes onto metric datapoints. Derived from the config's
resource_metrics_key_attributes, so a new key is picked up
automatically, in both dotted and underscore forms.
- resource.service.instance.id: strips to a dotted service-identity
key, which builtins only held in underscore form.
- name: the TraceQL span-name intrinsic, alongside duration and kind.
A file that does not parse falls back to the raw-text scan, which checks
every query rather than skipping it; JSON validity is already enforced by
the prettier pre-commit hook.
Adds 10 tests: the exemption, per-query scoping in a mixed dashboard,
target-inherits-panel datasource, no sideways inheritance leak, nested
row panels, TraceQL intrinsics, the malformed-JSON fallback, and the
collector-promotion helper.
09-data-collection-reference.md: keep phase-9's structure, which relocated
the §2a call-site-metrics content into §5b (3860c93db2). The incoming §2a
block is superseded; its snake_case namespace fix is already applied to
phase-9's copy of the same text.
The stat block was relaid two-per-row in the previous commit, but the reflow
shifted subsequent panels by a delta computed against the block's old height.
The block shrank, so the delta left a 12-grid-line dead band between the stat
row and "Node Operating State Transitions".
Recompute every panel's y by walking the list in order and pairing half-width
panels, rather than patching offsets incrementally. The layout is now derived
from panel order and heights, so this class of drift cannot recur.
Verified: 192 grid lines, fully covered, no empty band, no overlaps, and no grid
line occupied by more than two panels.
* release/3.2.x:
chore: Bump version to 3.2.1
chore: Bump version to 3.2.1-rc1
fix: Cap untrusted manifests per message and drop oversized ones
fix: Reject oversized validator manifest before decoding
fix: Reduce untrusted manifest cache cap to 100
fix: Bound untrusted manifest cache
Fixes found while validating the Log-Derived Insights dashboard against live
Loki and Tempo. Every change was verified by executing the query, not by
inspection.
loki.yaml — the ConsensusLedgerHash derived field never matched anything:
- The TraceQL query named `.xrpl.consensus.ledger_id`. The attribute is
`consensus_ledger_id` (ConsensusSpanNames.h); live Tempo lists no dotted or
xrpl-prefixed variant among its span attributes.
- The bare `.` intrinsic scope cannot match a span attribute regardless of
name; TraceQL requires `span.`.
Corrected to `{span.consensus_ledger_id="..."}`, which returns the expected
consensus.round trace for a real ledger hash.
log-derived-insights.json:
- Stat units `suffix: <noun>` rendered raw integers, so large values printed
in full. `si:<noun>` is unsafe because Grafana parses the leading letter as
an SI prefix (`si:fee` rendered "29.0 Mee"). Switched to `short`, the
convention already used by 42 panels; the noun is in each panel title.
- Stat panels showed a single fleet-wide tile labelled "Value #A". Grafana's
Loki backend does not name binary-operation frames from legendFormat, so the
ratio panel could not resolve per-node labels. Replaced it with a plain sum
of duplicate fetches; the ratio remains available as a timeseries. Added an
explicit legendFormat to all eight stat panels.
- The ledger-fetch regex required `good:N dupe:N` together, but
SHAMapAddNode::get() omits any zero counter. That dropped 32% of lines --
disproportionately the efficient fetches -- biasing the ratio upward. Each
field now matches independently.
- state-timeline used `spanNulls: true`, which spans gaps indefinitely and
would render an outage as continuous. Set to the 30-minute threshold.
- Stat panels relaid two-per-row and given a tooltip, per dashboard guidelines.
telemetry-runbook.md:
- LogQL examples used line filters where structured metadata exists.
`|= "ERR"` also matches the literal in a message body: measured 423 ERR
lines plus 4 DBG lines per 6h. Replaced with severity, partition, and
trace_id field filters.
- Documented the new dashboard, its debug-log requirement, and the LogQL traps
these fixes exposed.
- Corrected the dashboard count, which was already stale at eleven against
fifteen on disk, and named the six with no reference section.
The heading still read "Future:" while the section body documents the
validation suite this branch ships and gives commands to run it. Section 6.8.3
in 06-implementation-phases.md already drops its "Future Enhancement" marker
here; this makes 5c consistent with it.
The Phase 7 section existed twice in this file; the copies disagreed on this
one label. Consolidating on the single copy kept the older wording, but Jaeger
was removed from the project earlier in this chain, so Tempo is the only trace
backend the diagram should name.
These changes were developed on the phase-10 branch but belong to content this
branch and its upstreams introduced. Carrying them on phase-10 made its PR diff
report churn in files phase-10 does not own, and left each PR claiming a scope
that did not match its contents.
Moved here from phase-10 (identical content, no functional change):
- Dashboards: all 14 existing boards plus the new log-derived-insights board.
- Docs: telemetry-runbook.md (minus the workload/benchmark sections, which
describe phase-10 tooling) and the new telemetry-glossary.md.
- Grafana Cloud + Alloy export path: collector config, compose override, the
two .env examples and alloy/config.alloy.
- Local stack: otel-collector-config.yaml gains sub-millisecond and
second-scale spanmetrics buckets, pins unit=ms, and promotes
close_time_correct; integration-test.sh and TESTING.md follow.
- Node configs: exported_instance -> service_instance_id in comments; the
mainnet sample now logs at warning to bound log volume.
- Metrics code: Telemetry.cpp builds the metrics pipeline in the constructor
via initMetrics() so the global MeterProvider is published before any
subsystem creates a beast::insight instrument, and the histogram view keeps
each instrument's own name instead of collapsing them under one series.
MetricsRegistry gains a last_close_time gauge and skips negative job-queue
durations. OTelCollector drops an unused accessor.
- Naming CI: xrpl_work_item joins EXTERNAL_INFRA_LABELS and Rule E accepts the
dotted perf-iac resource-attribute form. This must travel with the
dashboards and runbook that reference those labels, or the rules fail.
- Doxygen input glob no longer recurses dot-directories.
Sections describing phase-10 tooling stay on phase-10 and keep their
"Future Enhancement" / "Planned, not yet implemented" markers here; phase-10
removes those markers when it lands the tooling.
Update the catalogue table and prose for the two thresholds corrected by the
7-day backtest, and document why ValidatedLedgerStale must exclude the
1209600s sentinel that getValidatedLedgerAge() returns when no validated
ledger exists — that clause looks redundant and would otherwise be removed
as a simplification.
Also record the tuning lesson: justify thresholds from a 7-day sample, not
24 hours.
ValidatedLedgerStale fired on every node, healthy included.
LedgerMaster::getValidatedLedgerAge() returns weeks{2} (1209600s) as a
SENTINEL when no validated ledger exists, not as a measurement. The rule
read that as "14 days stale". Measured over six days it produced sustained
firing on all nine nodes. Excluding the exact sentinel value drops that to
zero while still tracking real staleness.
ManifestFloodInbound at 50 kB/s was routine paging: ~41 sustained 5-minute
samples across six healthy nodes in six days. Healthy p99 is 1.0-1.8 kB/s
and real storms peak at 2.7 MB/s, so 512 kB/s sits ~280x above normal and
~5x below the peaks, cutting sustained samples to 2.
Both thresholds were previously justified from a 24-hour window, which was
too short to expose either problem.
Correct the dashboard count (11 -> 15), note which boards lack a reference
section, and update the log-derived-insights panel/row counts. Replace line
matches with structured-metadata filters in the LogQL examples and document
the PeerFinder debug requirement plus which [MIXED] panels work at the
default log level.
Signed-off-by: Pratik Mankawde <3397372+pratikmankawde@users.noreply.github.com>
Bound the number of manifests carried in a single TMManifests message
(kMaxManifestsPerMessage). Trusted manifests are always included and
processed; untrusted gossip is capped per message on both send and
receive, and the sender is charged only when untrusted entries are
actually skipped. Oversized TMManifests messages are dropped without
penalty at the protocol layer so an unpatched peer is not disconnected.
Complements the cache bound from #276/#323.
Forty-eight panels spread over seven dashboards used the standard
description template but stopped after Function, leaving out Keywords,
Computation boundary and References. ledger-data-sync was the worst at
33 of 35 panels. They read as half-written next to their complete
neighbours, and the Keywords section is what defines the domain terms a
reader needs.
Fill in the three sections for each, in the order the already-complete
panels use. Keyword text is chosen from the panel's own metric family
rather than written per panel: job_* panels get the job-type and
deferred-job definitions, getobject_* and nudb_* get GetObject and
NodeStore, peer_finder_* and the traffic counters get peer and overlay,
and span-derived panels state that the value comes from the SpanMetrics
connector rather than a native metric.
Every panel now carries all ten sections. Panel count, layout, units and
queries are untouched.
Thirty-nine panels on this dashboard carry the full ten-section
description; twelve did not, so they read as half-written next to their
neighbours.
Eleven of them used the template but stopped after Function, missing
Keywords, Computation boundary and References. Fill those in, placed in
the same order the complete panels use. Content comes from each panel's
own metric: the job-queue panels get the job-type and deferred-job
definitions, I/O Latency and Database Sizes get the scheduler and
NodeStore ones, and each states whether its value is a native
beast::insight metric or computed through the OpenTelemetry SDK.
Server State had no template at all, only a one-line summary. Rewrite it
in full, documenting the 0-4 operating-mode encoding and its colour
mapping, and noting that it is an instantaneous state rather than a rate.
Panel count, layout, units and queries are untouched.
Three dashboards had layout faults that Grafana renders as visible
artifacts.
On Node Health, two row headers carried the same title as the panel
directly beneath them -- "Cache Hit Rates" and "Object Instance Counts"
each appeared twice, once as a row bar and once as a panel. Rename the
rows to "Caches" and "Object Instances"; the panels keep their names.
Worse on the same dashboard, seven bands summed to 48 columns on a
24-column grid, because the Extended Metrics block reused the y-space
already occupied by NodeStore I/O. Grafana cannot lay that out
deterministically, which is what produced the artifacts.
RPC Performance and Transaction Overview each had blank bands -- one
row and eight rows respectively -- left behind by earlier panel moves.
Repack all three grids so every band fits 24 columns, each band starts
where the previous one ends, and rows stay full-width single-height
anchors. Panel order, panel count and every panel's own width and
height are unchanged; only position moves.
Brings the telemetry startup-ordering fix forward. Two conflicts, both in
the metrics registry.
initSyncInstruments(): this branch had already replaced the registry-owned
state_changes_total with a call-site macro carrying {from,to} labels, and
the arriving branch moves the jq_trans_overflow_total observable out of
this function into registerJqTransOverflowCounter() so it is armed with
the other pull-model instruments. Kept both: the explanatory comment for
state_changes_total stays, the inline overflow block goes. The arriving
stateChangesCounter_ creation is dropped rather than merged -- this branch
removed that member, so keeping the line would not compile.
Test file: kept this branch's fuller header documentation, which records
what the disabled build can and cannot assert for the sync-diagnostics
gauges, and folded in the arriving branch's one new fact, that the
lifecycle is now two-phase. Include lists unioned.
Resolves the telemetry-startup conflict between the two branches. Both
sides move the telemetry start earlier in setup(); they disagree only on
how far the pipeline had been split at that point.
phase-1b (arriving) moved nodeIdentity_, setServiceInstanceId() and the
telemetry start up to just after the wallet DB is proven usable. phase-9
had split the metrics pipeline in two and left its copy of that block at
the old, later position.
Kept both intentions: the block stays at phase-1b's early position, and
metricsRegistry_ construction moves up with it so it precedes
startTelemetry() -- the metrics half is guarded on the registry existing,
so leaving the construction behind would have started tracing while
silently skipping metrics. phase-9's later copy is dropped as the stale
duplicate. The two-phase split is preserved: startTelemetryGauges() still
runs after overlay_ is constructed, because the observable callbacks read
it and getOverlay() asserts.
Net effect is that the metrics provider now starts earlier than on either
branch, and still before beginConsensus() emits the first spans and the
only operating-mode transition.
Sixteen bargauges on ledger-sync-health are titled "... (Count ...)" and their
descriptions promise a count over the selected time range, but the queries
used last_over_time(counter[$__range]). For a monotonically increasing counter
that returns the newest cumulative sample -- the lifetime total -- so every
panel read the same number no matter how the time picker was set. Verified
live: dns_resolve_total over [15m] and [24h] both returned 12/5.
Switches the 17 affected targets to increase(), which is the range-scoped
delta the titles and descriptions already describe.
Panels: DNS Resolve Outcomes, Outbound Dial Outcomes, Handshake Negotiation
Failures, UNL Fetch Outcomes, Server Stall Episodes, Acquire Stalls, Acquire
Source, Inbound Peer Accept Outcomes, Peer Disconnects, Ledger/Object Serve
Refusals, Byzantine Ledger Jumps, Replay Fallback, Replay Outcomes, Pre-Accept
Quorum Shortfalls, Sweep Heap-Trim Faults, Rotation Node Re-Stores.
Two live-verified dashboard defects found while auditing the devnet nodes.
Sub-millisecond spanmetrics buckets. The ladder started at 1ms, but most
xrpld spans are far below that: tx.preflight means ~0.012ms and tx.preclaim
~0.15ms. On the live nodes the first bucket held 13,761,009 of 13,761,419
tx.preflight samples (99.997%), so histogram_quantile pinned every p95/p99
at a constant 0.95ms and ~10 latency panels reported a boundary artifact
rather than a measurement. Adds 0.01/0.05/0.1/0.25/0.5ms ahead of 1ms in all
three bucket lists, which must stay identical or the connector misbuckets.
consensus_mode filter on spans that never carry the label. Only
consensus.round and consensus.ledger_close are stamped with consensus_mode
(verified against live series), so panels 1, 3, 12 and 14 -- which query
consensus.validation.send, consensus.proposal.send and
consensus.accept.apply -- blanked as soon as an operator picked any value
from the Consensus Mode dropdown. Drops the selector from those four; ids 2
and 11 keep it because their spans do carry it.
Note: existing histogram series retain the old le boundaries, so
span-latency panels show a discontinuity at collector restart.
base_fee_xrp was observed from LoadFeeTrack::getLocalFee(), which is the
local load-scaled fee escalation, not the ledger's base fee. The panel
built on it therefore tracked this node's load state rather than the
network's cost of a reference transaction, and read as a flat line
whenever the node was unloaded.
Read it from the validated ledger's fee settings instead, alongside the
reserve values already taken from there. The observation now only
reports when a validated ledger is available, which is correct: before
that there is no network fee to report.
MetricsRegistry::start() fused three steps with different prerequisites:
the exporter/provider and the synchronous instruments need only config
strings, while the observable gauges read live Application services. The
whole pipeline therefore waited on the latest prerequisite and ran near
the end of setup() -- after beginConsensus() had already recorded the
process's only operating-mode transition. state_changes_total was
emitted into a pipeline that did not exist yet, so the series never
appeared at all.
Split the two halves. start() keeps the provider and the synchronous
instruments and now runs as soon as the registry is constructed; the new
startAsyncGauges() registers the observable callbacks and runs once
overlay_ exists, still before the first consensus round. Application
gains a matching startTelemetryGauges() so each phase has its own call
site and its own precondition.
Move the jq_trans_overflow_total observable out of initSyncInstruments()
into the gauge phase. Its callback reads getOverlay(), which asserts
overlay_ is non-null, so creating it in the early phase armed the reader
thread against a half-built application -- an assert is not caught by
the callback's catch-all. The instrument is an observable counter rather
than a gauge, which is how it was mistaken for a push-only instrument.
Both start log lines are kept, one per phase, because that timeline is
what made the original ordering bug diagnosable.
Comments and preconditions are corrected to state the rule rather than
the current arrangement: start() may only create instruments whose
values are pushed, and any observable whose callback reads a service
belongs in the gauge phase. The gauge precondition now lists the
services the callbacks actually read.
telemetry_->start() ran at the end of ApplicationImp::start(), after
overlay_->start(). Spans are emitted well before that, during setup():
beginConsensus() runs the first consensus round there. SpanGuard drops a
span whenever the global Telemetry instance is not yet live, so that
round's spans were never recorded.
Move the start into setup(), behind a new startTelemetry() seam, right
after the node identity is known. getNodeIdentity() needs only the
cmdline, the config, or the wallet DB, and initRelationalDatabase()
has already created the latter -- the adjacent peerReservations_ load
proves it is usable -- so the identity block moves up with it.
The new position is bounded on both sides: after initRelationalDatabase()
because the identity needs the wallet DB and a DB failure aborts setup(),
and before beginConsensus() because that emits the first spans.
Grafana renders `unit: "ops"` as the literal string "ops/s" and `cps`
as "counts/s", so seven rate panels on this dashboard read as
"operations per second" regardless of what they counted -- nodes added
to a SHAMap, dial attempts, ledger-data requests served.
Replace them with custom-suffix units naming the quantity, following
the `suffix:/hr` and `si:drops` precedent already in this repo. Each
noun was checked against the emitting code rather than the panel title:
sync_addnode_total counts SHAMap nodes, peer.dial spans are connect
attempts, ledger.serve spans are peer requests.
Also spell out "Ops / Sec" as "Operations / Sec" on the NodeStore panel,
since abbreviating the noun defeats the point of naming it.
Queries are unchanged; the values were already correct.
The two key-job latency panels on node-health declared milliseconds
while querying `job_running_us` / `job_queued_us`, which record
microseconds. Every reading was a thousand times too large: the p95 for
acceptLedger, 241495us, rendered as "241 sec" instead of 241ms.
job-queue.json already read the same metrics as microseconds, so the
two dashboards disagreed by 1000x on identical data.
Also replace the generic `ops`, `cps` and `cpm` units, which Grafana
renders as the literal "ops/s", "counts/s" and "counts/min", with
custom-suffix units naming what each panel counts -- jobs, ledgers,
validations, lookups. The per-minute panels already scale correctly in
their queries; only the noun was missing.
Normalise the micro sign to U+00B5 throughout. Four axis labels used
the visually identical Greek mu, U+03BC, while every unit field used
the micro sign.
Ledger Close Rate plots ledgers closed beside ledger fetches, which one
unit cannot describe; give the fetches series its own unit on a
right-hand axis. Drop two field overrides on NodeStore Read Latency
that restated the panel unit as a custom suffix -- a suffix is appended
verbatim, so it would have suppressed magnitude scaling and left large
values unreadable, the same defect fixed above.
Queries are unchanged apart from the transaction path piechart, which
moves from rate() to increase(): a per-slice "per second" reading is
not a share of a total.
Alongside, widen the Complete Ledger Ranges table to full width, hoist
the stat panels above the fold, and bring the touched panels up to the
tooltip and null-spanning guidelines.
The four job-latency panels on node-health declared milliseconds while
querying `job_running_us` / `job_queued_us`, which record microseconds
(MetricsRegistry records the raw value, and the instrument description
says microseconds). Every reading was therefore a thousand times too
large: the p95 for acceptLedger, 241495us, rendered as "241 sec"
instead of 241ms. job-queue.json already read these same metrics as
microseconds, so the two dashboards disagreed by 1000x on identical
data. Switch node-health to microseconds to match.
Also replace the generic `ops` and `cps` units, which Grafana renders
as the literal "ops/s" and "counts/s", with custom-suffix units naming
what each panel counts -- messages, fetches, calls, mismatches.
Two panels plot more than one quantity on a single axis, which no
single unit can describe. Give each series its own unit through field
overrides: reads per second beside two queue depths on NuDB Read
Pressure, and ledgers beside fetches on Ledger Close Rate, the latter
on a right-hand axis.
State Duration Rate plots a seconds-per-second time share, which can
exceed 1.0 and so is not a percentage; label it as the ratio it is.
The normalised share already exists as its own panel.
Queries are unchanged; the values were already correct.
Alongside, bring the touched panels up to the dashboard guidelines and
hoist the stat panels above the fold.
Grafana renders `unit: "ops"` as the literal string "ops/s", so every
rate panel read as "operations per second" regardless of what it
actually counted. `Ledger Build Rate` showed "0.258 ops/s" where the
value is one ledger every 3.9s -- the number was right, the unit was
meaningless.
Replace the generic units with Grafana custom-suffix units naming the
quantity, following the existing `suffix:/hr` and `si:drops` precedent
in this repo. Nine of these are `stat` panels with no axis, so the unit
string was the only text a reader ever saw.
Also switch the two trusted/untrusted piecharts and the transaction
path piechart from rate() to increase(): a per-slice "per second"
reading is not a share of a total, counts in the window are.
Queries are unchanged apart from those three; the values were already
correct.
Alongside, bring the touched panels up to the dashboard guidelines:
tooltip mode/sort/max-height, 30-minute null spanning, and axis labels
in title case. Hoist the stat panels above the fold on
ledger-operations and rpc-performance.
Panels that a later branch in this chain removes are deliberately left
alone -- fixing them would only add merge conflicts.
Telemetry must read state, never change it. Two defects here did change it,
plus three smaller correctness and privacy fixes.
doPathFind and doRipplePathFind read source_account / destination_account off
context.params to hash them into span attributes. context.params is non-const,
so those reads selected json::Value's non-const operator[], which inserts a
null for a missing key. The same object is later validated by
PathRequest::parseJson, whose first checks are isMember(source_account) and
isMember(destination_account) — so a request that omitted either field looked
present and the client received Malformed instead of Missing. Reads now go
through std::as_const, whose overload returns kNull without inserting.
PathRequest::doUpdate emitted pathfind_dest_currency as
to_string(saDstAmount_.asset()). For a non-XRP asset that renders as
"<issuer>/<currency>" with the issuer as a plaintext Base58 address, so a
plain account address reached the span pipeline even though every other
account here is hashed first. The issuer is now redacted and the currency
kept; an MPT asset renders as its issuance ID and carries no address.
PathRequestManager::updateAll created pathfind.update_all with an unscoped
SpanGuard. An unscoped guard takes the ambient span as its own parent but does
not itself become the ambient parent, so the pathfind.compute spans that
doUpdate creates never nested under it, contradicting the documented hierarchy.
It is now a scoped guard, held in std::optional because ScopedSpanGuard is
deliberately non-movable and so cannot be produced by a ternary. The skip when
there are no active subscriptions is preserved. updateAll is dispatched via
addJob and doUpdate runs synchronously, so the guard is constructed and
destroyed under the same context store, as ScopedSpanGuard requires.
The WebSocket entry point emitted the client-supplied command string directly.
That value becomes a Prometheus label, so arbitrary request input could drive
unbounded label cardinality. It is now resolved against the handler registry,
collapsing anything unrecognized to "unknown", matching what the HTTP path
already does.
Also: the pathfind.discover comment claimed future child spans could be
parented off it, which its unscoped guard cannot do — corrected to say what
would be required instead. Config-reference and task-list docs named the
parser setupTelemetry(); the API is makeTelemetrySetup().
Conflict in processSession(): this branch added the request-payload-size
attribute on the same lines where the incoming change reverted
processRequest() from bool back to void. Kept both — the payload-size
attribute stays, and the call no longer captures a return value. The
trailing status block takes the incoming side, which leaves the
rpc.http_request span's status unset.
Six related defects in the RPC/gRPC span surface, all cases where a failure
was recorded as success or an attribute was missing on an error path.
GRPCServer: the non-exception branch set the span Ok unconditionally, then
sent a possibly-failed grpc::Status. The handler can return a non-OK status
without throwing, so every failed call traced as successful. Status now
follows result.second, with the error message as the span description.
ServerHandler: eight per-item error branches appended an error reply without
recording that the request failed. Batch responses and ripplerpc < 3.0 always
carry HTTP 200, so those failures were invisible and an entirely failed batch
ended its span as successful. Added an appendItemError() helper next to the
existing httpReplyError() lambda and routed all eight sites through it, so the
flag cannot be forgotten at a new call site.
ServerHandler: the early-return validation paths set the span error but not the
rpc_status attribute. Added it to httpReplyError() so every such path gets it.
RPCHandler: the fillHandler error path set only command and rpc_status, while
callMethod sets command, version and rpc_role. Error spans were therefore not
filterable by API version or role. The error path now mirrors that set.
RPCHandler: resolveCommandSpanName() checked only that command/method were
present, not that they agreed, while fillHandler rejects a mismatch as
rpcUNKNOWN_COMMAND. A request supplying both with different values was labelled
with one of the two names, misattributing the error to a command that never
dispatched. It now mirrors fillHandler's rule and collapses to "unknown".
ServerHandler: processRequest returned bool solely so the caller could set its
span status. Telemetry should read state, not shape the signature of the code it
observes, so the signature returns to void and rpc.process sets its own status
from spanHadError. The enclosing rpc.http_request span now leaves status unset:
the OTel spec has instrumentation leave status unset unless the operation itself
errored, and reserves Ok for an operator asserting verified success.
Conflict in src/libxrpl/telemetry/SpanGuard.cpp: both sides added a different
include in the same alphabetical slot — phase-4 added <initializer_list>, the
incoming std::format change added <format>. Both are required, so both are
kept in sorted order.
SpanGuard::span() and freshRoot() joined the prefix and suffix by hand with
reserve() + three append() calls. std::format expresses the same join in one
line; libstdc++ has shipped <format> since GCC 13 and the project floor is
GCC 15.2 (BUILD.md), so it is available on every supported compiler.
Both callers are noexcept and std::format can throw (std::bad_alloc, or
std::format_error on a malformed spec), so an escaping exception would
terminate the process. The call is wrapped in a joinSpanName() helper that
catches and returns std::nullopt; the caller then returns a null guard, which
is the same degrade-to-no-op path already taken when telemetry is disabled.
Telemetry must never bring the node down.
One helper rather than a try/catch at each site keeps both call sites a single
line and puts the exception-safety requirement in one documented place.
Addresses a review comment on PR #6437.
Conflict in OpenTelemetryPlan/03-implementation-strategy.md §3.9: both
branches independently fixed the same defect (stale hand-maintained line
counts) in different ways.
- 1b (0ff947454c) stripped the Lines Added / Lines Changed columns but kept
the §3.9.1 and §3.9.2 tables.
- 1a (3ad525a48a) removed both sections outright.
Resolved in favour of 1a's deletion. The tables were pre-implementation
estimates with no source of truth, so trimming the columns only defers the
next drift; the file/component lists duplicated §3.1's directory tree, which
1b already keeps current (DiscardFlag.h, FilteringSpanProcessor). Keeping the
upstream resolution also stops the same conflict recurring on 1c..10.
1b-specific content verified intact after the merge: §3.1 retains the
DiscardFlag.h entry, the FilteringSpanProcessor and discard() annotations,
and the TracingInstrumentation removal. No references to the deleted
sections remain in any plan doc.
The §3.9.1 "Files Modified Summary" and §3.9.2 "Detailed File Impact"
tables carried hand-maintained per-component line counts that had drifted
from the plan: the Lines Added column summed to 1,565 while the Total row
claimed ~1,670. Files (34) and Lines Changed (120) reconciled, so only the
Added total was stale — residue from expanding Core Telemetry 5 -> 11 files.
Rather than patch one cell, remove both tables. They were pre-implementation
estimates with no source of truth, so any figure in them drifts again on the
next phase. §3.1's directory tree remains the canonical list of the 11 new
telemetry files, and §3.9.3-3.9.7 (risk, architectural impact, backward
compatibility, rollback) carry the assessment without inventing line counts.
No references to the removed sections exist elsewhere in the plan docs.
Resolves src/tests/libxrpl/CMakeLists.txt test_modules by keeping both
sides: develop's beast/nodestore/protocol additions (nodestore moved into
alphabetical position) and this branch's ledger module.
Someone filling in .env.grafanacloud-alloy has no reason to know that
deployment.environment and xrpl.network.type are not env vars: they are
literals inside the tier transform, because OTTL statements do not expand
sys.env(). Left alone they ship as prod/mainnet and every dashboard
environment and network filter reads the wrong tier. The example now says
so, and notes service_instance_id needs nothing here.
Alloy declared 24 spanmetrics dimensions where both collector configs
declare 28. The four missing ones -- timed_out, object_type,
validation_status and accept_gated -- are emitted by the code and queried
by ledger-sync-health panels "Ledger Serve Rate by Object Type" and
"Trusted Validation Accept Rate by Status", so on an Alloy deployment
those panels collapsed to one undifferentiated series.
deploy-run-otel-xrpld.md already requires the dimension sets to stay in
sync; the config had drifted from it. Alloy's other differences from the
collectors (no filelog, no tail_sampling) are deliberate -- it is
cloud-only by design.
Five review findings.
The validation_accept span started inside the "outcome is Current"
branch, so validation_status could only ever read "current" and the four
rejected values were unreachable. That defeated the attribute: a node
whose trusted validations are all rejected emitted no span at all, so it
looked the same as a quiet node. The span now starts before the outcome
is checked, still for trusted validations only, so a rejected one is
recorded with its real status.
The rest are consistency fixes: ValidatorSite passes the parse_error
constant instead of the literal, the 28 gauge callbacks that spelled the
`metric` label key as a literal now use label::metric like the other two,
a stray blank first line is gone from six files, and MetricsRegistry.cpp
loses a duplicate Doxygen block that documented addHistogramView but sat
above an unrelated constant.
Two gaps meant the last failure produced no evidence of its cause.
The node-log upload was gated on `if: failure()`, but the validation step
sets continue-on-error, so the job is not failing at that point and the
condition never fired. Every failed run silently skipped the one artifact
that records why a node did not reach consensus. It now keys on the
validation step's own outcome, and also collects the harness logs.
Transaction failures were logged at DEBUG, which CI does not enable, so a
run where all 3052 submissions failed on a refused connection reported
nothing about it. The first occurrence of each distinct failure kind is now
a warning and repeats stay at DEBUG, so one refused connection says so once
instead of 3052 times.
Three template variables (xrpl_work_item, xrpl_branch, xrpl_node_role)
select their datasource via ${DS_PROMETHEUS}, but the variable itself was
absent from templating.list, so those label_values() lookups resolved to
no datasource and the filters stayed empty.
Add the datasource variable as the first templating entry, matching the
other dashboards in this folder.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The dashboard's template variables and panel targets select their
datasource via ${DS_PROMETHEUS}, but the variable itself was absent from
templating.list. An unresolvable datasource variable leaves those queries
without a datasource, so the panels render empty.
Add the datasource variable as the first templating entry, matching the
other dashboards in this folder.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The lane-attribution change added acquire_ledger_deferrals and
acquire_ledger_timeouts to the gauge but left this test expecting seven
labels, so it failed on every platform of both PRs: nine emitted against
seven expected. The exporters and the counters went in as separate
commits, which is how the test escaped the update.
The two new labels are asserted with values that differ from the all-lane
totals they are a subset of, so a counter wired to the wrong lane, or one
that ignored the flag and counted every lane, lands on the totals instead
and fails.
The round built each thread's batch inside the thread, before arriving at
the latch, so a throw there left the remaining threads waiting on an
arrival that never came -- the test hung instead of failing. A spawn loop
that ended early did the same.
Batches are now built before any thread starts, so nothing between spawn
and arrival can throw, and a guard counts down the shortfall for threads
that were never spawned before joining the ones that were.
The depth accounting having moved to insert entry, depthSamples is now
the denominator of the mean depth, so it gets its own assertions: equal
to insertCount once every thread has been joined, and moving with the
duplicate-key round. The overlap assertions are unchanged.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
HashRouter.h declared `class Config;` and never referred to it: the only
other mentions of the name in the file are two prose comments. Nothing
that includes this header depends on it either -- of the eighteen files
that do, the only one that names the type needs the full definition and
includes it itself. The declaration is vestigial, left behind when
setupHashRouter() moved to setup_HashRouter.h, which carries its own.
Left alone it is now an error rather than dead code. This branch's
SlotCensus work made Overlay.h include PeerfinderManager.h, because
Overlay::slotCensus() returns a SlotCensus by value. ValidatorList.cpp
includes both Overlay.h and HashRouter.h, and is the one translation unit
that reaches them without also pulling in xrpld/core/Config.h -- so
PeerFinder::Config and this declaration meet there, and
bugprone-forward-declaration-namespace reports a declaration with no
definition beside a same-named class in another namespace.
Removing it rather than including the real header: xrpl::Config lives in
xrpld, so an include would point the xrpl.libxrpl.core layer at the
layer above it.
The counters that attribute deferrals and timeouts to ledger acquisition
compared a string literal in TimeoutCounter against a separate literal in
InboundLedger. Renaming the job would have silently sent both metrics to
zero: no build error, no failing test, and documentation that quietly
became wrong. Both now use one shared constant.
Three measurement fixes landed with no doc or dashboard change, leaving text
that is now false and one fix unusable from a dashboard.
Deferrals and timeouts are recorded in TimeoutCounter, a base shared by five
subclasses, so the all-lane pair could show the documented livelock
fingerprint while ledger acquisition was healthy. The runbook procedure and
the reference doc now name acquire_ledger_deferrals and
acquire_ledger_timeouts and say why the all-lane pair misleads; a new panel
plots the ledger-scoped pair as rates on one axis, since the divergence is
the signal. The existing panel is retitled All Lanes and points at it.
Writer mean depth is depthSum over depthSamples, not over insertCount, and
the measured 1.60 came from the biased estimator, so it and the 37% queueing
share derived from it are lower bounds rather than values. The reference
table now marks them as such, and the decision rule is shown to survive the
correction rather than depending on the exact figures.
Completions were never counted for acquisitions satisfied from the local
store, so the run that read zero across 510 seconds had in fact reached
full. Every place that treated a zero as a symptom now says it only means
something on a build that has the fix.
Also corrects the sync-diagnosis label-value count from 13 to 15 and a stale
source line range; the instrument count stays 35, because both new values
multiplex onto the existing nodestore_state gauge.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
getSlotCensus() had no test that called it. The existing telemetry tests
cover the gauge callback by re-implementing it against a mock meter, so
they never exercise the accessor itself and would not notice a field
added to SlotCensus without a matching census assignment.
Four tests drive a real Logic and assert every field exactly:
- a fresh Logic publishes its configured ceilings and nothing else, so
"no slots configured" is distinguishable from "configured but empty";
- inPeers with wantIncoming=false yields an inbound ceiling of 0, per the
rule in Counts::onConfig;
- one fixed and one ordinary peer walked through dial, connect, handshake
and close, asserting all seven intermediate states. This pins the
non-obvious part: Counts::adjust guards the outActive_/inActive_ branch
with !s.fixed(), so a fixed peer raises fixedActive and never
outActive. A census that conflated them would overstate slot pressure.
fixedConfigured stays at 1 after the disconnect, which is the
"configured fixed peer I cannot reach" reading an operator needs;
- bootcache and livecache depths come from separate sources, so filling
one leaves the other at 0.
Four findings from a review pass over the PR.
The "Spans & traces" row was empty. Moving the row header down to clear
the back-fill panels was only half the change -- the seven span-derived
panels stayed at their old y, one unit below the native panels, so every
pair overlapped and Grafana parented all fifteen to "Back-fill &
persistence". The panels now sit below the row header, which restores
the split the runbook already describes: eight native panels answer "how
much", seven span-derived ones answer "which". Both rows stay expanded,
so the docs no longer call them collapsed.
metric_constants() excises each namespaced block before the flat
prefix pass. The flat pass classifies by identifier prefix and is meant
for headers that name the role in the identifier because they have no
`namespace metric`/`label`/`lval`; it was running over the whole header,
so a `kLabel`-prefixed constant written inside `namespace metric` landed
in both buckets and an instrument name became a valid label key for Rule
D. Nothing in the tree does that today, which is why it went unnoticed,
and why the guard is a test rather than a fix for an observed failure.
The `site` label now keeps a non-default port and drops userinfo residue
from the host. Omitting the port unconditionally merged two local sites
that differ only by port; printing it unconditionally would have renamed
the existing `https://vl.ripple.com` series. Comparing against the
scheme default distinguishes a configured port from the one the Resource
constructor fills in. parseUrl's host group also permits '@', so a
malformed URI with two of them leaves part of the userinfo in `domain`.
Both counters are recorded in TimeoutCounter, a base shared by five
subclasses with different job limits, so they pooled every lane
together. The documented fingerprint for a stalled acquisition is
deferrals rising while timeouts stay flat -- but a saturated replay lane
reproduces that shape while ledger acquisition is healthy, and the two
counters being compared were not drawn from the same population.
Adds ledger-acquisition-scoped counts alongside the totals, discriminated
by the job name already held at both call sites, so no new state and no
signature change. The all-lane counters stay for callers that want them.
Their exporters went in with the preceding commit, which shared a file.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Depth was observed at insert entry but its sample was only counted at
insert exit, so an insert still in flight contributed nothing to the
mean while completed inserts -- disproportionately the fast, shallow
ones -- all did. The reported mean understated queueing exactly when
queueing was worst: with every writer inside its first insert the gauge
was omitted entirely, while writers-in-flight correctly showed them all.
Depth and its sample count are now both folded in at entry, so the mean
is taken over one population. Mean depth is the L in Little's Law, so a
biased L understated the derived queueing share of each insert.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two findings from the CI clang-tidy job, both mechanical:
- MetricsRegistry.h no longer includes <ranges>. The only ranges use in
the header is std::ranges::all_of, which is declared in <algorithm>
and already included, so the include was genuinely unused.
- The computeGetObjectByHashFee() expectation table now uses designated
initializers. FeeCase has five fields, two of which are plain ints in
adjacent positions, so naming them at each row also makes a
requested/found transposition impossible to write by accident.
No behaviour change.
The workflow pinned its container image and compiler in a second place:
a hardcoded gcc-13 tag alongside the one in the build matrix. That copy
went stale when the rest of CI moved on, so this job kept building on
gcc 13 and failed on code the main CI compiled cleanly.
The tag now comes from the strategy matrix, so there is one source of
truth. The nix image carries several toolchains, so the compiler is also
selected explicitly for Conan to detect the intended one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
recordCompletion() was reached only from done(), but init() can satisfy an
acquisition entirely from the local store and return without ever calling
done(). Those completions went uncounted, so acquire_completions read zero
for a whole class of outcome. On a live clean-database sync it read 0 for
510 seconds while the node reached full, which was misread as the node
completing nothing.
Counted in init() rather than by calling done(), because done() drives the
state machine: it stores the ledger, dispatches AcqDone, updates the
fetch-rate average, and sets signaled_, which would disarm a later genuine
done(). A counter must not change behaviour. A completionCounted_ latch,
separate from signaled_, makes the count idempotent across the two
independent exits so a completion is counted exactly once however it was
reached.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
BEAST_EXPECT(getFetchDurationUs() > 0) failed on the macOS arm64 runner at
DatabaseConfig_test.cpp:879. It is an assertion about the host, not the
code: Database::fetchNodeObject() truncates each fetch to whole
microseconds, and a read served from NuDB's in-memory buckets can
genuinely measure under one microsecond. On a host fast enough for every
read to truncate, the total legitimately stays at zero.
Both fetch-duration checks now compare the accumulator against the total
the scheduler's fetch reports carried, via a CountingScheduler. Each
fetch is measured once and that one value feeds both, so the equality
holds at any host speed while being strictly stronger than '> 0': a
dropped fetch, a double-count, or the write member being returned all
break it. The same reasoning is already recorded in the GTest at
src/tests/libxrpl/nodestore/Database.cpp.
The store-duration '> 0' checks are left alone: those writes reach disk
and were not what failed.
All three were rejected by clang-tidy as errors:
- readability-math-missing-parentheses on 2 * kNumStored + kNumMissing
- misc-const-correctness on the never-mutated AcquireStats 'quiet'
- bugprone-unchecked-optional-access on the read_threads_running
dereferences: branching on BEAST_EXPECT's return value hides the
has_value() check from the analyser, so the guard is now a plain if and
the macro asserts separately.
The FeeCase table used positional aggregate initialisation, which
modernize-use-designated-initializers rejects for a five-field aggregate;
the clang-tidy job failed once per row. Naming the fields also makes the
requested/found/derived/literal ordering readable at each row instead of
relying on the reader tracking four ints positionally.
The header was added alongside sanitiseHandler(), which calls
std::ranges::all_of -- but that algorithm lives in <algorithm>, which the
same commit also added and which is still needed. <ranges> itself is never
used, so misc-include-cleaner rejects it and the clang-tidy job fails on
MetricsRegistry.h:146.
Adversarial validation of the previous commit found one of its two code fixes
was diagnosed wrongly and the other incomplete. Both are corrected here, along
with the layers the first pass missed.
1. The new dial outcome was named for the wrong condition. It was added as
`duplicate` on the belief that PeerFinder had already granted a slot for the
address. It has not: `Logic::onConnected` contains exactly ONE false-returning
path and it is the self-connect check, which logs "Logic dropping as self
connect" (include/xrpl/peerfinder/detail/Logic.h). The duplicate check lives
in `newOutboundSlot`, evaluated before a ConnectAttempt exists, so a real
duplicate can never reach this branch.
That mattered beyond the name: the previous commit told operators the outcome
was benign churn to ignore, when it actually reports a local misconfiguration
-- this node has its own address in [ips_fixed] or behind its advertised
endpoint, and every dial to it is wasted. Renamed to `self_connection`,
reusing the slug `handshake_negotiation_fail_total` already publishes for the
same fault so it reads identically on both signals, and every description
corrected to say so. The fail() string now reads "Self connection" too.
The first pass also missed three enforcement and contract sites: the
ConnectAttempt.h Doxygen state machine (which still mapped the slot branch
onto tls_fail), the LedgerSpanNames unit test (which pinned exactly five
values over a std::array<..., 5> and so left the new member untested), and the
span-derived twin panel plus two reference docs that still published the old
five-value domain.
2. The credential-free site label was incomplete twice over.
- It appended the port, and `Resource::Resource` DEFAULTS that to 443/https
and 80/http when the config omits one. The label would have become
`https://vl.ripple.com:443/` where Grafana Cloud currently holds
`https://vl.ripple.com`, silently renaming the series for every deployment
already scraping this metric. Verified against live label values before and
after; the port is now omitted.
- parseUrl's path group is `(/.*)?`, greedy to end of string, so a query or
fragment lands inside `path`. A list URL authenticated by `?token=...` would
have leaked exactly as userinfo did. The path is now truncated at the first
'?' or '#'.
Also updated the MetricNames.h usage example, which still taught the raw-URI
pattern to the next author, and the 09-doc row that described the label as the
configured URI.
3. Rule J hardening from the same review: `classify_instrument_kind` returns an
`other` sentinel for a non-factory macro, and storing it in the kind set could
render a future conflict as "created as counter and other". The sentinel is
now skipped, keeping it doing what it already did -- matching no shape rule.
Added a second regression test whose input the pre-fix code reported as CLEAN
(gauge-then-histogram on a `_us` name), so the guard is proven by a 0-vs-1
difference and not only by a changed message. Both new tests were run against
a reconstructed last-wins implementation and both fail against it.
Documented the conflict class in the Rule J rows of the checker README and
CONTRIBUTING, which previously described only the suffix conventions.
Verified: naming checker exits 0 with Rule J passing all 40 real names; 140
checker tests pass; 15 dashboards validate; both workload JSON files parse;
clang-tidy over the full compile database reports no finding on any changed line
of ConnectAttempt.cpp or ValidatorSite.cpp; pre-commit passes.
Not verified: not compiled. The label change adds string truncation and the
outcome rename touches a constexpr used across three translation units, so CI's
build remains the first real check on both.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Four defects from the automated review on PR #7875, each verified against the
current tree before fixing (one further comment, the row-63 dashboard overlap,
was already fixed by an earlier commit and needed nothing).
1. Rule J could not detect an instrument-kind mismatch. instrument_kinds() wrote
`kinds[wire] = ...`, so a wire name created through two different factories
kept only the kind visited last and whichever emit site the file walk reached
last silently decided the verdict. It now collects a set per name and reports
the conflict itself -- one name exporting two instruments is the defect, and
no suffix can be correct for both. Added a regression test that builds a name
as both a counter and an observable gauge and asserts the message names both.
2. A duplicate connection was reported as `tls_fail`. The TLS handshake had in
fact succeeded; PeerFinder simply already held a slot for that address, which
is ordinary churn on a healthy node. Conflating the two made a rising
`tls_fail` unreadable -- it could mean unreachable peers or merely a busy
PeerFinder, and those need opposite responses. Added a distinct `duplicate`
outcome and carried the widened vocabulary through every place that
enumerates it: the panel description, both filter descriptions, the runbook
branch table, the runbook outcome list and the expected_spans note. The
`dial_outcome` template variable is a label_values() query, so it picks the
new value up on its own.
3. ConnectAttempt::onShutdown had no `operation_aborted` guard, unlike the five
other handlers in the same file. A clean teardown was therefore counted as
`upgrade_fail`, inflating that outcome on any node shutting down with dials in
flight.
4. ValidatorSite used the raw configured URI as a Prometheus label.
[validator_list_sites] accepts credentials in the URI and ParsedUrl keeps them
in username/password, so a configured `https://user:pass@host` would have
copied the secret into a metric label and on into the collector, Prometheus
and every dashboard. The label is now rebuilt from scheme, host, port and
path -- everything needed to tell one site apart, and nothing more.
Verified: naming checker exits 0 with Rule J still passing all 40 real
instrument names; its unit tests now number 139 and all pass; 15 dashboards
validate; both workload JSON files parse; clang-tidy over the full compile
database reports no finding on either changed .cpp; pre-commit passes.
Not verified: not compiled. Item 4 introduces string concatenation and item 2 a
new constexpr, so CI's build is the first real check on both.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Brings in the phase-10 revert of the nodestore read-latency histogram plus the
nudb_bytes -> stored_object_bytes rename.
Conflicts in MetricsRegistry.{h,cpp} resolved keeping both intents:
- MetricsRegistry.cpp: dropped everything that existed only to serve the
reverted nodestore_read_us histogram -- the addSubMillisecondHistogramView()
helper, its call site, the kSubMillisecondBoundaries array and the
NodeStoreMetricNames.h include. Kept every view this branch registers
(consensus round duration, sweep_malloc_trim_us, dns_resolve_latency_ms,
overlay_dial_latency_ms) and the shared addHistogramView() base helper.
Took the rename at the storage_detail observe() call site.
- MetricsRegistry.h: took phase-10's move of the four nodestore_state observe
helpers and their ObserveFn sink from private to public, while keeping this
branch's enriched Doxygen on observeNodeStoreTotals().
Also corrected the registered-view count in the 09 reference doc: neither side's
arithmetic survives the merge, since this branch adds four views phase-10 never
saw and the revert removes one. Ten views are registered now, not six or seven.
Drops nodestore_read_us and everything added to reach it. read_mean_us already
carries microsecond precision and separated the two sync failure modes cleanly
in live testing -- 8.8 us on a clean store against a 223 us cold-store peak --
so the distribution added no signal that changed a diagnosis.
The cost of getting it was disproportionate. NodeStoreScheduler had no path to
the metrics registry, so its production constructor grew a ServiceRegistry
parameter: a metric addition changing a production signature. That in turn
forced an edit to a pre-existing test, src/test/app/SHAMapStore_test.cpp, whose
only stake in this is that it constructs a scheduler. Worse, the scheduler is
built in Application's member initializer list, long before metricsRegistry_
exists, so the registry could not be captured once and had to be re-resolved on
every fetch -- a lookup on a path that runs millions of times per sync.
The constructor returns to taking JobQueue& alone and SHAMapStore_test.cpp
returns to the single-argument call, leaving that file differing from its
pre-change form only by the NodeStore:: to node_store:: rename it picked up from
develop.
FetchReport::elapsed stays microseconds and onFetch keeps its explicit
duration_cast to milliseconds for addLoadEvents, which takes milliseconds. That
widening was a separate fix and is what makes read latency measurable at all.
kSubMillisecondBoundaries loses its only consumer and regains [[maybe_unused]],
which is the state the commit that introduced it left it in; without the
attribute an unused constant is an error under wextra with werr.
Also removes the ledger-data-sync panel that charted the histogram and the
fetch_type and found template variables, which filtered on labels no metric
emits any more, plus the runbook and reference-doc sections and the two
instrument and view counts that named it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The Sync State panel only ever showed Connected and Full. Two causes:
1. state_tracking{metric="state_value"} is emitted with values 0-6, where
5 is FULL+validating and 6 is FULL+proposing, but the panel declared
max: 4 with value mappings for 0-4 only. Values above 4 were pinned to
the axis ceiling and rendered unmapped. Most nodes sit at 6, so the
majority of series were clipped.
2. The gauge samples the instantaneous mode on a 10s export tick, so
states shorter than one tick fall between samples. A real sync showed
SYNCING for a single scrape and skipped TRACKING entirely. This is the
sampling hazard already noted on StateAccounting in NetworkOPs.h.
Extend the panel domain to 0-6 with Validating and Proposing mappings and
matching threshold steps, and add a colour-coded state-timeline panel where
each band's width is the time spent in that state, so brief states appear
as thin slivers rather than disappearing. The timeline reads
server_info{metric="server_state"} (raw OperatingMode 0-4) rather than
state_value, which folds 5 and 6 onto FULL and would split one Full band
into three colours.
Panel layout below the insertion point shifts down by 6 rows.
Note this makes short states legible, not lossless: exporting
StateAccounting's per-state duration accumulators is the sampling-immune
fix and is left as follow-up.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The 22 metric label values on the nodestore_state gauge were asserted
nowhere, and the depthSum assertions could not tell the real
fetch_add(depth) from a fetch_add(1) that would silently zero every
derived queueing time.
Make the four observe* helpers and their ObserveFn sink public rather
than private, so the seam the header already claimed is actually
reachable. All four are static and read only their arguments, so this
exposes no object state; a friend declaration would have granted access
to every private member instead. The assertions live in the existing
Beast nodestore suite because src/test compiles into xrpld, which
contains MetricsRegistry.cpp, while xrpl_tests deliberately does not
when telemetry is enabled.
Each helper now has its exact emitted label set asserted, so a typo in
any literal fails instead of silently producing a disjoint series, and
each derived mean is asserted ABSENT on a fresh store -- a refactor to
value_or(0) would draw a believable flat zero on a latency axis and
otherwise pass everything.
Replace the concurrent write-stats test with one that forces genuine
overlap through a latch. NuDB holds one global mutex for the whole
insert and doInsert reads the depth before entering it, so a blocked
thread records a depth of at least 2; asserting depthSum strictly
exceeds insertCount therefore cannot be satisfied by a constant 1. The
old bounds admitted that bug at their floor.
Also: cover the std::nullopt branch on the two backends that exist in
every build, bound the store-duration accumulator by the wall clock,
drop four assertions that cannot fail, and correct two comments that
claimed coverage the tests do not have -- the duplicate-key test is not
the throwing path, because nudb reports key_exists without throwing,
and no test drives NodeStoreScheduler::onFetch.
The nudb_bytes label value on the storage_detail gauge named something the
code never measured. It observes Database::getStoreSize(), which returns the
storeSz_ accumulator: the cumulative payload bytes of objects this process
handed to the NodeStore. That is not a NuDB file size. It excludes NuDB's
keys, bucket padding and log, and it resets with the process while the files
on disk do not.
The name caused two concrete errors. It invited sizing the store on disk from
a number that cannot do it, and it invited a write-amplification ratio against
node_written_bytes -- which reads the same accessor at MetricsRegistry.cpp:836,
so that ratio is a constant 1.0 and measures nothing.
The nudb_ prefix was wrong too. storeSz_ is written only by
Database::storeStats(), called from DatabaseNodeImp, DatabaseRotatingImp and
Database itself. No backend code touches it, so the value reads the same on
RocksDB. That distinguishes it from the real nudb_* family
(nudb_writers_in_flight and friends), which come from getWriteStats() and are
absent entirely on a non-NuDB backend.
stored_object_bytes says what the value is and claims nothing about the
filesystem. Docs already described the value correctly; they keep that
explanation and now also record the old name, so a query pinned to it can be
traced. Neither Backend nor Database exposes an on-disk size accessor and none
was added -- no metric reports the store's on-disk size today.
Updates the node-health panel title, description and PromQL, and the four docs
that name the label value.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Eleven local phase-10 commits, mostly documentation and dashboard-description
work, plus one real include fix. No conflicts (git merge-tree reported 0).
Reviewed for interaction with this branch:
- b50aa17aee adds <xrpl/consensus/ConsensusSpanNames.h> to RCLConsensus.cpp and
PeerImp.cpp. This is the same misc-include-cleaner finding my local clang-tidy
run reported on those two files, so the merge resolves two of the 74
pre-existing findings rather than adding any.
- 8ade2b43d9 / 7da5ac5992 / d18f713f91 rename the node-health found-ratio and
stored-bytes panels and correct the nudb_bytes and NuDB found-ratio
descriptions, then repoint the runbook at the new names. These are the same
correction phase-10 made earlier to node_reads_hit -- 'found an object', not a
cache hit -- carried through to the panels that render it.
- ledger-sync-health.json is untouched by phase-10, so none of this branch's
panel work is affected.
Verified after merging: naming checker exits 0, its 138 unit tests pass, all 15
dashboards validate, levelization produces no diff, and this branch's own work is
intact (4 touch() call sites, the ledgers_behind guard, 54 panels / 18 bargauges).
The panel was renamed from NuDB Cache Hit Ratio to NuDB Read Found Ratio,
which left the runbook naming a title that no longer exists and carrying a
paragraph saying the rename had not happened yet. Both are corrected, and
the explanation now says why the ratio is a found rate: the counter
increments whenever a fetch returned an object, and a node with
online_delete has no object cache at all.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Brings forward the node_reads_hit and nudb_bytes label corrections from phases 7
and 9.
Conflicts resolved keeping both sides:
- 06-implementation-phases.md: kept phase-10's unprefixed `storage_detail` gauge
name with phase-7's corrected getStoreSize() description.
- node-health.json: kept phase-10's rewritten panel structure and re-applied the
panel renames (NodeStore Read Found Ratio, NuDB Stored Bytes), legends and axis
labels on top of it.
node_reads_hit is a found count, not a cache hit. fetchHitCount_ increments
whenever a fetch returned an object, whatever served it, and a node with
online_delete uses DatabaseRotatingImp which has no NodeObject cache at all. The
ratio therefore reads near 100% while every fetch goes to disk, which made the
cold-read failure mode look impossible on the board.
- Phase9_taskList: node_reads_hit is a found count, not cache-served reads.
- node-health: "NodeStore Cache Hit Ratio" -> "NodeStore Read Found Ratio",
legends "Cache Hit Ratio" -> "Found Ratio" and "Reads Hit (cache)" ->
"Reads Found"; corrected the ratio panel's axis label, which read
"Operations / Sec" on a percentunit panel.
- node-health: "NuDB Storage Size" -> "NuDB Stored Bytes". nudb_bytes observes
getStoreSize(), the accumulator node_written_bytes also reads, so it is
cumulative object-payload bytes and not a filesystem measurement.
- ledger-data-sync: point the read-latency discriminator at the renamed panel.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
nudb_bytes was documented as a NuDB file size, one place even claiming a
filesystem stat. It observes Database::getStoreSize(), which sums the object
payloads this process has written. It excludes NuDB's keys, bucket padding and
log, and resets with the process. node_written_bytes calls the same accessor, so
the two series are equal by construction and a write-amplification ratio built
from them is a constant 1.0. Neither Backend nor Database exposes a file-size
accessor, so nothing reports on-disk size today.
The Ledger Data & Sync panel plotting node_reads_hit / node_reads_total was
titled "NuDB Cache Hit Ratio" and described as reads served from cache.
fetchHitCount_ increments whenever a fetch returned an object, whatever served
it, so the ratio is a found rate. It reads near 100% while every fetch goes to
disk, which made the cold-read failure mode look impossible. Renamed to
"NuDB Read Found Ratio" and rewrote the guidance to pair it with read latency.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Picks up two local phase-10 commits that were not yet on the remote, the
substantive one being clang-tidy cleanup in the new nodestore metric tests.
That cleanup addresses the same two check classes my own local clang-tidy run
surfaced as pre-existing on the base branch:
- bugprone-unchecked-optional-access: ASSERT_TRUE(x.has_value()) does not teach
the analyser that the later x->/x.value() is safe, because ASSERT_TRUE expands
to a return rather than a control-flow edge it can follow. Replaced with an
explicit 'if (!x.has_value()) FAIL() << ...', which both satisfies the check
and gives a message naming what was expected.
- misc-include-cleaner: <cstddef>, <algorithm> and <optional> added where the
translation unit uses them directly.
Also AcquireStats const-correctness and a kPassThroughHandlers loop rewrite.
git merge-tree reported 0 conflicts. Both sides edit
src/tests/libxrpl/telemetry/MetricsRegistry.cpp (theirs +16/-20, ours +213/-12)
but in disjoint regions, so the union applied cleanly.
The tests added on this branch tripped six checks under
WarningsAsErrors. All of them are in test code introduced here.
bugprone-unchecked-optional-access: gtest's ASSERT_TRUE returns an
opaque AssertionResult, so the dataflow analysis cannot see that a
following deref is guarded. Replaced with an explicit
`if (!x.has_value()) FAIL()`, which the analysis does follow, or with
a direct optional comparison where no deref is needed. Both keep the
original assertion strength and add a reason string.
readability-use-anyofallof: the two consteval helpers now use
std::ranges::all_of. The static_assert still evaluates at compile
time, verified by inverting the predicate and watching it fail.
misc-const-correctness, misc-include-cleaner,
modernize-use-designated-initializers: const on a never-mutated
local, corrected include sets, and named fields on the Expected
aggregate so its two adjacent bools cannot be transposed silently.
No production code changes, and no NOLINT added.
Docs-only from phase-10: two commits refining the sync-diagnosis decision rule
and nodestore_state metric semantics.
Reviewed for conflicts with this branch's work and found none. Two points worth
noting because they touch documentation this branch also edits:
- node_reads_hit is re-described from "reads served from cache" to "fetches that
found an object (not a cache hit)". That is the accurate reading of the
counter and does not change any query here.
- The cold-read decision rule is retuned (insert mean threshold ~1.2, and the
read-mean/found-rate split for telling cold-but-held from real misses). It
does not contradict the retirement of nodestore_latency on this branch, which
removed a duplicate gauge rather than any nodestore_state series.
git merge-tree reported 0 conflicts before merging.
Two metric-level defects the sync analysis identified, fixed at the source rather
than worked around on the dashboard.
1. InboundLedgers::sweep() destroys any acquire idle for more than a minute, and
that destruction is what telemetry reports as outcome=abandoned. But
lastAction_ was only refreshed by the constructor, update() and done() --
never by the receive path. With JtLedgerData capped at 3 concurrent jobs and
33 acquires in flight, an acquire whose peers were answering normally could
wait past the cutoff for its turn to apply data and be deleted for looking
idle. Measured on a fresh mainnet sync: 490 abandoned acquires against ZERO
expired retry budgets, so every one was a sweep, not a give-up.
gotData() now calls touch(). The sweeper's idle test measures real inactivity
instead of queue wait.
lastAction_ had to become atomic to allow this. It was a plain
clock_type::time_point written by the acquiring thread and read by sweep() on
the timer thread; adding a third writer on peer threads would have been a data
race. It is now std::atomic<clock_type::duration::rep> with relaxed ordering on
both sides -- the sweeper compares against a 60-second threshold, so a value
one tick stale cannot change its decision.
2. getLedgersBehindNetwork() returned the entire ledger sequence space on a fresh
node. The existing floor only guarded being ahead of every peer; it did not
guard having validated nothing at all, so validated=0 against a live tip gave
105,892,534 -- an accurate subtraction of a meaningless quantity. It
auto-scaled every consumer's axis and would trip any threshold. Distance to
tip is undefined before the first validated ledger, so it now reports 0 until
there is one, and the sync-state signals carry the initial-acquire progress.
The clamp_max(1e6) added to the Ledgers Behind Network panel as a stopgap is
removed: the metric is correct now, and leaving the clamp would hide a real
large backlog.
Verified: clang-tidy over the full compile database reports no finding on any
changed line in the three files (the pre-existing misc-include-cleaner and
misc-const-correctness findings elsewhere in InboundLedger.cpp are untouched by
this change). pre-commit passes including clang-format and the Doxygen style
check; validate_dashboards passes.
Not verified: not compiled -- per instructions.md the build needs approval, so CI
is the first real compile of the atomic change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two conflicts, both additive-vs-additive; each resolution keeps both sides.
check_otel_naming.py -- phase-10 taught the L6 label extractor to match the
label MAP first and to resolve a key hoisted into a `k...Label` constant,
scanning headers as well as sources. Our side had added the two-regex
first/subsequent literal scan and the `metric_constants(root)[1]` union that
covers the `namespace label` header style.
Kept phase-10's mechanism whole: METRIC_LABEL_MAP + the `(?:^|\{)` key regex
already subsumes what METRIC_LABEL_NEXT did, since matching inside the map body
makes every pair after the first open with a single `{`. So METRIC_LABEL_NEXT is
dropped as genuinely redundant rather than kept as a duplicate scan, and the
reason it existed is folded into METRIC_LABEL's comment. Re-added our
`metric_constants(root)[1]` union on top: LABEL_CONST_DEF only matches
`k`-prefixed identifiers, so it cannot see MetricNames.h's `label::jobType`
style, and without that union Rule D would reject dashboards querying labels
Rule I forced into constants. The two derivations are complementary and both
are now documented as such.
MetricsRegistry.cpp -- both sides added a new sibling view-registration helper
next to addMicrosecondHistogramView, and both added a registration call in
initExporterAndProvider(). Kept all four helpers
(addHistogramView/Microsecond/RoundDuration/SubMillisecond) and every
registration: phase-10's addSubMillisecondHistogramView + kNodeStoreReadUs
alongside our addRoundDurationHistogramView, sweepMallocTrimUs and the two
millisecond dial/resolve ladders.
phase-10's nodestore_read_us histogram does not duplicate our work. The
nodestore_latency gauge that would have overlapped it was retired in c4e434d520
before this merge, and the surviving nodestore_state gauge is complementary
rather than duplicative: both read the same fetch measurement, but the gauge
publishes only a since-boot mean via scaledMean() and cannot yield a
percentile -- the consequence observeNodeStoreTotals' own docs state plainly --
while the histogram buckets each fetch and can. The histogram also splits by
fetch_type and found, which the gauge cannot. phase-10 registered its
explicit-bucket View, so it does not inherit the SDK default ladder.
Each file keeps its own existing naming style: phase-10's k-prefixed constants
are left as-is, ours stay namespaced.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The decision rule for "slow to reach full" keyed on an absolute read-time
threshold and a found-rate threshold that misclassified the very run they
were written to explain: a populated-store run reading 31.8 us at 88.3%
found fell through both cuts and came out as "disk-bound" rather than the
cold-read case it is.
Replace it with two ordered questions -- is the read cost several times a
warm read, and is the write path queueing -- and demote the found rate to a
splitter that only applies once reads are known to be expensive. A high
found rate on its own is the normal state of a populated store, so it can
never be a trigger. The rule now classifies all four reference datasets
correctly, and the runbook shows the rule applied to each so the "confirm
against the reference points" step agrees with the table.
Also in the runbook:
- name the source of the devnet incident figures at the point of use, and
point forward to the caveat from the same paragraph
- state the provenance of the measured columns, and split the incident
figures into their own table marked as not our measurement
- say plainly that the compounding-factor explanation is an unconfirmed
hypothesis
- correct the deferral gate: it fires at the acquisition's own job limit of
5, not at the ledgerData lane cap of 3
- note that no read-max gauge exists, so the tie-break uses max_over_time
of the mean or the read histogram's p99
- split a PromQL block that put two expressions on adjacent lines, which
parses as one invalid expression
In the data-collection reference:
- node_reads_hit counts fetches that found an object, not cache hits
- nudb_bytes is cumulative payload bytes from the same accessor as
node_written_bytes, not on-disk size, so their ratio is a constant 1.0
- write_load and nudb_writers_in_flight are the same atomic on NuDB;
document that and what write_load means on RocksDB
- correct the instrument count to 8 and the view count to 7 after
nodestore_read_us and its view were added
- scope the Phase 9 query examples to one node with the regex form
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Second pass on the 54-panel audit. The first commit handled the count panels and
the description drift; these are the query and threshold defects.
Panels 4 and 6 (DNS Resolve / Outbound Dial p95) read NaN for the whole run.
Two faults compounded: both hard-coded [5m] instead of $__rate_interval, so they
ignored the dashboard time range entirely, and both wrapped the histogram buckets
in rate() even though DNS resolution and outbound dialling only happen during
startup -- a windowed rate over a series that stopped moving is 0/0. Dropping
rate() and reading the cumulative buckets gives the real distribution: p95 900 ms
for DNS, 1750 ms for dialling. Panel 52 kept its rate() (consensus rounds are
ongoing) but its hard-coded [5m] became $__rate_interval.
Panel 15 plotted 105,892,534 "ledgers behind" during the flagship window. The
underlying cause is in NetworkOPs.cpp -- getLedgersBehindNetwork() subtracts the
validated sequence from a networkTarget of 0 before any peer has reported -- and
that still needs a code fix. Meanwhile one sentinel spike flattened the real
0-20 backlog for the rest of the window, so the query now clamps at 1e6, far
above any true backlog. Reads 4 where it read 105 million.
Panel 17's sum by (from, to) dropped node identity, so with All nodes selected
every node's transitions summed into one bar. It now carries service_instance_id,
xrpl_branch and xrpl_work_item like every other panel.
Panel 38 divided by clamp_min(op_rate, 1), which turns "no operations in this
interval" into "one operation", reporting the whole duration total as if a single
op had consumed it. Replaced with a `> 0` gate so an idle interval draws a gap
instead of a fabricated latency.
Panels 13 and 45 had inverted thresholds: green began at 1 second, so every
sub-second time-to-full and time-to-first-validated rendered red -- the healthy
case was the alarming colour. Now green by default, yellow past 10 minutes, red
past 30.
Panel 48's p95 had no outcome filter, mixing abandoned and timed-out spans (which
sit at the retry ceiling by construction) into what reads as completion latency.
Restricted to outcome="complete".
Verified against Grafana Cloud: 66 queries, 0 parse errors, 56 with data, 9
legitimately empty (fault-only counters plus the write-timing series that
pre-dates both recorded nodes). The single remaining NaN is on panel 47 and is an
artifact of my verification harness forcing a hard 5m window; the panel's own
$__rate_interval returns 492 ms, so it needs no change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A 54-panel audit across 11 dimensions, each finding adversarially re-verified
against live data, returned 38 confirmed defects. This fixes them. Most were
introduced by the recent rate-to-count conversion itself.
1. increase() was the wrong function for these counters. A counter that only
moves at startup is born at its final value inside the window and never
rises, so increase() reports 0. Measured: dns_resolve_total reads 4 but
round(increase(...[$__range])) returned 0 -- the panel lost the signal
entirely. increase() also drops whatever accrued before the window opened,
which under-reported the rest (overlay_connect_total 110 against a true 130,
unl_fetch_total 10 against 14, peer_disconnect_total 7 against 8).
All 18 count panels now use last_over_time(...[$__range]), which on a
fresh-node board is the cumulative total since the process started -- exactly
what "count" means here.
EXCEPT panel 49. span_calls_total comes from the collector's spanmetrics
connector, which is collector-side state and does NOT reset when xrpld
restarts, so last_over_time would report the collector's lifetime across every
run: it read 2624 header completions where run C actually had 297. That panel
keeps round(increase(...)) and now reports 297/278/212/7, matching the
analysis. The distinction is process-level counter vs collector-side counter,
and it decides which function is correct.
2. Descriptions still described rates after the conversion, over three passes of
wording (Reading it / Healthy range / Watch for blocks, "Rate of", "per
second", "/s", "a rising rate"). 11 panels corrected; the two surviving uses
of "rate" are legitimate (a cache hit-rate reference, and panel 49 explaining
why a count reads better than a rate).
3. Ten descriptions pointed at panel titles that no longer exist, because the
conversion renamed the panels they cross-referenced. Two others named panels
that never existed on this board at all: "Total Jobs Queued" (now Worker Pool
Capacity & Total Backlog, panel 27) and "Fetch-Pack Peer Starvation" (now
Peers Able to Serve Needed Sequence, panel 28).
4. Panels 18 and 23 applied $acquire_metric on top of a hard-coded metric
selector, so the two ANDed: any selection outside the panel's own values gave
an empty graph and All was the only usable state. The redundant template
selector is gone; the panel's metric pair is its identity.
5. Panel 23 drew two series with different ranges (received_data_depth 0-20,
in_flight 13-49) under one yellow threshold at 16, so in_flight was
permanently yellow. The threshold is now scoped to received_data_depth.
6. The "Spans & traces" row sat at y=248, the same y as panels 38/39, so Grafana
folded the Back-fill panels into the wrong row. Moved to y=296, below the last
back-fill panel. Rows are now strictly ascending with no collision.
Verified against Grafana Cloud Prometheus: all 66 panel queries parse, 0 errors,
57 returning data (up from 56 -- panel 3 was one of the ones increase() had
silenced). validate_dashboards and check_otel_naming both pass.
Not verified: no PNG renders this round; the local Grafana and Prometheus are
down, so every check ran against Cloud data via the datasource proxy.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The phase-10 merge relocated PeerFinder's Manager interface from
src/xrpld/peerfinder/ to include/xrpl/peerfinder/, and SlotCensus moved with it
so libxrpl could see it. Logic.h uses SlotCensus as the return type of
getSlotCensus() but was never given the include, so every build failed:
build/modules/xrpl.libxrpl.peerfinder/xrpl/peerfinder/detail/Logic.h:192:5:
error: unknown type name 'SlotCensus'
...:197:16: error: use of undeclared identifier 'SlotCensus'; did you mean
'getSlotCensus'?
That single missing declaration was the whole failure. The 34 further errors in
the log were cascade: PeerFinderTest could not compile, so every TEST() in
src/tests/libxrpl/peerfinder/PeerFinder.cpp failed to instantiate against
gtest-internal.h. All four platforms (ubuntu-clang, ubuntu-gcc, macos-arm64,
windows-amd64) and clang-tidy reported the same root cause.
No include cycle: PeerfinderManager.h does not include detail/Logic.h, directly
or transitively. Levelization is unchanged because both headers are already in
xrpl.libxrpl.peerfinder, so generate.py produces no diff.
This is the risk called out when the merge landed -- moving SlotCensus into the
public header was the one resolution a static check could not confirm, and only
a compile would prove it. CI is that compile, and it found this.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
- JobQueue_test: drop the unused <functional>, and make the two
read-only GaugeFixture instances const. The other two stay mutable
because they submit jobs through fixture.queue.
- DatabaseConfig_test: drop the unused SystemParameters.h and include
ByteUtilities.h for megabytes(), which was reached only transitively.
Regenerate ordering.txt for the resulting edges: test.unit_test arrives
with SuiteJournal.h, and xrpl.protocol leaves with SystemParameters.h.
Rule D rejected the `handler` and `result` dashboard filters even though
both labels are emitted, because the L6 extractor missed them twice over:
- It matched only a key written as an inline literal directly after `{{`.
A label map is a braced list of braced pairs, so in
`Add(1, {{"job_type", a}, {"handler", b}})` only the first key follows
`{{` -- every later one was dropped. A key hoisted into a constant
(`{{kHandlerLabel, v}}`) was invisible in any position.
- It walked *.cpp only, so a constant declared in a header, as
kLabelResult is, could never be found.
Match the label map first and scan its pairs, resolve `k...Label`
constants through their `constexpr char k...[] = "..."` definitions, and
read headers too. Matching the map rather than any `{"key",` in the file
keeps ordinary brace initializers, such as the `{"http", "https"}` scheme
array, out of the label set -- they are not labels and must not license a
dashboard filter. Test code is skipped for the same reason Rule F skips
it: fixtures pass arbitrary literal pairs.
The new PromQL used an exact match on service_instance_id. The $node
template variable is multi-value, so an exact match returns nothing as
soon as more than one node is selected. Every one of the 442 filters
across the dashboards uses the regex form; the runbook now agrees.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A stalled ledger sync has two distinct causes that look identical from the
job queue: in both, the ledgerData lane sits at its concurrency limit of 3
with work waiting. Lane occupancy therefore discriminates nothing, and the
dashboard had no panel that separated the two.
Adds a Sync Bottleneck Discrimination row to Ledger Data & Sync with seven
panels built on the nodestore_state write-path and acquire counters:
- NodeStore Read Latency, lifetime and windowed, on a log axis with
threshold lines at 10 and 100 us. This is the discriminator: fast reads
mean the cost is on the write side, slow reads with a high hit ratio
mean every access is paying disk latency.
- NuDB Writer Queue Depth, the confirming half. NuDB takes one global
mutex per insert, so depth above 1 is queueing at that mutex.
- NuDB Insert Time, mean against true max, which quantifies how much of
each insert is wait rather than work.
- Acquire Deferrals vs Timeouts on one panel, because the livelock
fingerprint is deferrals rising while timeouts stay flat, and neither
counter shows that alone.
- Acquisition Progress and Discarded Acquire Work, the outcome side: no
completions while the lane is busy means acquisitions resolve neither
way.
- NodeStore Read Latency Distribution split by fetch_type and found, from
the nodestore_read_us histogram, since a mean cannot tell a uniformly
warm store from a mostly-warm one with a cold tail.
Every expression carries the existing service_instance_id and tier filters.
The histogram introduces two new label dimensions, so fetch_type and found
template variables are added to match.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two different bottlenecks both present as the ledgerData job lane pinned
at its concurrency cap of 3, so lane occupancy diagnoses neither. One is
write-serialized (NuDB takes one global mutex per insert, so inserts
queue), the other is cold-read-bound on a populated store. Telling them
apart needs the storage-side signals, not the lane.
Adds to docs/telemetry-runbook.md a "Slow to reach full" procedure: a
Mermaid diagram of the two modes, a decision table keyed on whether
acquisitions are completing, the measured reference values from both
runs, and the deferral/timeout pair that fingerprints the disarmed
give-up path. States plainly that node_reads_hit is a found count rather
than a cache-hit rate, which is why a ~100% "hit rate" at 113 us per
read is the cold-read signature and not a contradiction.
Records honestly that the populated-store run was twice as fast despite
slower reads, so cold reads alone do not explain the long incident.
Adds reference rows for the 13 new nodestore_state label values and the
nodestore_read_us histogram to
OpenTelemetryPlan/09-data-collection-reference.md, in the authoritative
Phase 9 OTel SDK section alongside the existing NodeStore I/O table.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
kSubMillisecondBoundaries existed but nothing used it, so per-fetch read
latency never reached Grafana -- only the coarse read_mean_us gauge did,
which cannot separate "every read took 9us" from "most took 2 and a few
took 900".
Add a nodestore_read_us histogram, register its view against the sub-
millisecond ladder rather than kMicrosecondBoundaries (whose first edge
is 100us, above the entire range a warm read occupies), and record into
it from NodeStoreScheduler::onFetch using FetchReport::elapsed, which a
previous change widened to microseconds for exactly this purpose.
The name and its labels live in a new include/xrpl/telemetry header
because the view registration (xrpld.telemetry) and the record site
(xrpld.app) sit in different levelization modules; a copy-pasted literal
would let them drift and silently drop the bucket override. Same reason
and same placement as GetObjectMetricNames.h. No new levelization edge:
xrpld.app > xrpl.telemetry already exists.
NodeStoreScheduler had no registry access, so it now takes a
ServiceRegistry and resolves the registry per call. It is constructed in
Application's initializer list, long before metricsRegistry_ is assigned
in setup() and started in startTelemetry(), so capturing a pointer at
construction would capture nullptr forever; the metric macros null-check
the registry, the meter and the instrument, so early fetches are simply
not recorded.
Labels are fetch_type and found, both already carried on the report --
4 series, fixed at compile time. A slow async read delays prefetch while
a slow sync read blocks a caller, and a miss can cost a read of every
backend, so neither dimension can be collapsed.
Negative elapsed times are skipped: the SDK rejects them and logs a
warning on every call, which on a per-fetch path is a log flood. Zero is
still recorded, since a page-cache-served read genuinely rounds to it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
nodestore_latency published six values that nodestore_state already
publishes from the same Database accessors, so the two gauges were
duplicate readings of the same atomics:
write_count -> node_writes getStoreCount()
read_count -> node_reads_total getFetchTotalCount()
write_duration_us -> node_writes_duration_us getStoreDurationUs()
read_duration_us -> node_reads_duration_us getFetchDurationUs()
write_mean_us -> write_mean_us store duration / count
read_mean_us -> read_mean_us fetch duration / count
nodestore_state is kept because its means go through scaledMean(), which
saturates at INT64_MAX instead of wrapping and omits a mean when the
denominator is zero rather than reporting a misleading 0 us.
Removes registerNodeStoreLatencyGauge, its instrument member, the
metric::nodestoreLatency constant and the lval::nodestore_latency label
namespace. The gauge-over-histogram rationale and the "p99 is not
obtainable" consequence are folded into observeNodeStoreTotals' docs.
Retargets the gauge-contract test onto nodestore_state rather than
deleting it: the scaledMean arithmetic is covered by the static_asserts
in tests/libxrpl/telemetry/MetricsRegistry.cpp, but nothing else asserts
that these named series multiplex onto one instrument keyed by `metric`.
The test now calls the production scaledMean instead of a copy of the
division, and its sub-microsecond case asserts scaledMean's actual
behaviour (a genuine mean of 0 on a zero numerator with a non-zero
count), which differs from the retired gauge's extra numerator guard.
Rewrites both ledger-sync-health copies' panel 38/39 queries and drops
the obsolete claim that the write numerator was never written: all three
concrete store paths call recordStoreDuration, so write_mean_us is live
on an ordinary node. The same stale [import_db] caveat is removed from
the runbook, the 09 reference row and the workload validator's note.
beast::Journal has no default constructor, so holding one as a plain
member deleted the suite's own default constructor and the Beast
registration macro could not instantiate it. SuiteJournal takes the
suite, converts implicitly where a journal is expected, and routes log
output into the test report.
Also removes a leftover unused local in run(), which is a hard error
under the warnings-as-errors build.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
(cherry picked from commit e3c2f8279a)
The nodestore suites moved from Beast to GTest upstream, which deleted
src/test/nodestore/TestBase.h. DatabaseConfig_test stayed on Beast and
still derived its journal and batch helpers from that base, so once both
sides met in a merge it referenced three symbols that no longer existed.
It now carries its own copies, matching the current API: node object
types are NodeObjectType::Ledger rather than the old hotLEDGER spelling.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
(cherry picked from commit 05d01de81a)
The phase-10 merge brought a relocation: ConsensusSpanNames.h moved from
src/xrpld/consensus/ to include/xrpl/consensus/. Three of its four consumers were
updated during conflict resolution, but this test still included the old path,
which no longer exists.
Also corrects the file header, which explained that the test is guarded on
XRPL_ENABLE_TELEMETRY "because that is the configuration in which this test target
has src/ on its include path". That reason no longer holds: the header is lib-side
now, so a libxrpl test can include it directly without reaching into src/. The
guard remains because it is what builds the telemetry test target.
Verified statically: every in-tree xrpld/ and xrpl/ include across src/ and
include/ now resolves to a file that exists (only the three generated protobuf
headers are absent, as expected before a build). check_otel_naming.py exits 0 and
its 134 unit tests pass; the one remaining reference to the old path is a
synthetic fixture path inside those tests, built in a temp dir, so it is correct
as written.
Not verified: this was not compiled. The clangd diagnostics that surfaced the
problem come from a compile database generated ~19 hours before the merge, so it
still indexes the pre-relocation layout and reports errors for headers that are
present; those specific errors are an artifact, but a real build is still the only
proof the merge compiles.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Phase-10 brought in the upstream nodestore/peerfinder/consensus reorganisation
along with its own write-path telemetry, which collided with the sync-diagnostic
signals on this branch. Twelve files conflicted; every resolution keeps both
intents rather than picking a side.
The nodestore write timing was implemented twice, independently. Both sides
added getStoreDurationUs()/getFetchDurationUs() to Database and both timed the
backend call in each concrete store(). Keeping both would have added twice to
storeDurationUs_ per store while storeStats() still counted one, so the mean
write latency would have read double on every dashboard -- silently, since no
test on either side asserts an exact microsecond figure. Resolved to one
accumulator API: recordStoreDuration(), which takes a duration, clamps a
sub-microsecond sample to zero and uses a relaxed atomic add. Phase-10's
storeDurationStats() is gone and its two call sites now use the survivor, so
all three store paths -- both store() overrides and importInternal() -- add
exactly once.
SlotCensus and its pure virtual moved from src/xrpld/peerfinder/ to
include/xrpl/peerfinder/PeerfinderManager.h, following the Manager interface
upstream relocated. The xrpld header is now phase-10's makeConfig shim, and
Overlay.h, MetricMacros.cpp and the getSlotCensus() override chain point at the
new location. ConsensusSpanNames.h and peerfinder Slot.h/Config.h include paths
followed their headers into libxrpl the same way.
InboundLedger gained phase-10's AcquireStats counters next to this branch's
span activations in both the destructor abort path and done(); neither
displaces the other. nodestore_state keeps the constant-based name this branch
requires of it and phase-10's fuller description.
Upstream #7292 deleted src/test/nodestore/Database_test.cpp, which held this
branch's testDurationAccessors. Phase-10 restored the per-store half of that
coverage in DatabaseConfig_test, but nothing covered importInternal -- it writes
through storeBatch() and never through store(), so it is a third store path that
has to time itself. That half is ported to a GTest in
src/tests/libxrpl/nodestore/Database.cpp, keeping the exact zero-before and
accumulate-after assertions and the per-instance negative check.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two panels were left out of the previous sweep because a blanket rate-to-count
rewrite would have been wrong for them. Handled per target here.
Panel 65 (Sweep Heap-Trim) carries two cumulative counters, minor faults and
reclaimed KB. Both are count-shaped, so it becomes a bargauge over
round(increase(...[$__range])) like the rest, and the legends drop their
"/ Sec" suffix now that the values are totals rather than rates.
Panel 66 (Online-Delete Rotation) stays a timeseries. Its two targets are not
the same kind of thing: target A reads rotation_state{in_flight}, a 0/1 flag
whose whole value is seeing when it is high and for how long, and target B
rates rotation_state{copy_forward}, a cumulative write total. A count bargauge
would destroy the flag's time dimension. Instead the shared "cps" unit -- wrong
for a flag -- is replaced by per-target overrides: the flag pinned to a 0..1
left axis, the write rate on a right axis in cps. This matches how the metric
is documented to be read (MetricNames.h:648): copy_forward climbing while
in_flight is 1 is expected, climbing while it is 0 means the flag leaked.
Verified against Grafana Cloud Prometheus: 0 parse errors on both panels.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Seventeen panels on ledger-sync-health drew low-frequency discrete counters as
timeseries with an ops/s unit. The rate is arithmetically right but unreadable:
"0.1394 ops/s" is 278 abandoned tree phases over 33 minutes, and no reader can
recover the 278. rate() also extrapolates, so whole events rendered as
fractions -- peer disconnects showed 8.008 for 8 actual disconnects.
Measured every candidate against live Prometheus over a 33-minute mainnet sync
before converting; none exceeded 0.48 events/s, so all are count-shaped:
serve_refused_total 955 events 0.4785/s
ledger_quorum_shortfall_total 670 0.3354/s
sync_acquire_source_total 511 0.2559/s
overlay_connect_total 130 0.0652/s
sync_acquire_no_progress_total 109 0.0545/s
unl_fetch_total 12 0.0061/s
peer_disconnect_total 8 0.0040/s
Each becomes a bargauge over round(increase(...[$__range])) with unit short and
decimals 0, matching the existing Mode Transitions panel. Legends follow
instructions.md OTel rule 7 -- "MetricName [labels]" -- and keep node identity
via xrpl_ident, so a nine-node view no longer collapses to one bar. Panel 49
also gained node identity in its aggregation, which it was missing entirely.
Label values stay as emitted (upgrade_fail, not "Upgrade Fail"). They are wire
identifiers: rewriting them in the legend would hide what the metric reports and
break silently when a new value appears. The Title Case sits in the metric name.
The five zero-valued panels were checked rather than assumed dead --
ledger_replay_*, sweep_malloc_trim_* and rotation_copy_node_restore_total all
have real emit sites and are feature-gated off in this configuration.
Verified against Grafana Cloud Prometheus: 0 parse errors across all panel
queries, whole-number results throughout.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Three defects, all found by rendering the panels and running their queries
against a live mainnet node.
1. Ledger Acquire Phase Outcomes had an invalid PromQL escape.
The saved JSON held "ledger\\.acquire\\.(.*)", which decodes to
`ledger\.acquire\.(.*)`. In a PromQL double-quoted string `\.` is not a
legal escape, so Prometheus rejected the whole query:
parse error: unknown escape sequence U+002E '.'
The panel therefore rendered an error badge and "No data". A PromQL string
needs two characters, so the JSON must carry four backslashes.
2. The same panel never grouped by outcome, despite its title.
`sum by (span_name, timed_out, ...)` omitted `outcome`, so complete and
abandoned collapsed into one line. Measured at 16:45 UTC that hid a 29x
difference: astree complete=2085 against abandoned=71, all drawn as a
single indistinguishable series -- and every phase then showed the same
0.2596/s value, which is what made the panel look meaningless.
Now grouped by outcome, giving four real series (verified live):
header complete 0.1439/s, header abandoned 0.0772/s,
astree abandoned 0.1404/s, txtree abandoned 0.1404/s
The selector moves from timed_out (always "false" here, so it carried no
information and its filter var was redundant) to the declared
$span_outcome. Legend becomes "<phase> <outcome>"; axis label reads
"Phases / sec" to match the ops unit.
3. Seven stat panels and one heatmap dumped the raw label set as the legend.
With no fieldConfig.defaults.displayName but textMode "value_and_name",
Grafana has no name to show and falls back to printing every label:
{deployment_environment="local", exported_instance="xrpld-mainnet",
exported_job="xrpld", instance="otel-collector:8889", ...}
Rendered PNGs of panels 10 and 26 confirmed it. Fixed on ids 10, 12, 13,
14, 26, 36, 45 and 52 with the board convention already used by 24 sibling
stat panels: "${__field.labels.series} ${__field.labels.xrpl_ident}".
Verified afterwards by executing all 66 panel queries on this board against
live Prometheus: 0 parse errors, 56 returning data. The 10 empty ones are
counters a healthy node never increments plus panel 29, which is gated to stay
blank until a ledger is validated.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The Peer Ledger Supply Window panel drew supply_min_seq, supply_max_seq and
nothing else on one linear axis. Measured on a mainnet node, those sit around
105,890,000 and roughly 300,000 apart, so the 588-ledger tip movement that
shows whether sync is progressing was 0.0006% of the axis and read as a flat
line. unit "none" also printed the sequences unabbreviated and clipped the
legend.
The panel's own "Watch for" text asked the reader to compare supply_min_seq
against this node's validated sequence, but that line was not on the panel at
all, so the comparison meant switching dashboards.
Plot the two distances instead, which is what the panel was always asking
about:
History Headroom = validated_ledger_seq - supply_min_seq
Tip Gap = supply_max_seq - validated_ledger_seq
Zero is now the boundary in both directions: negative headroom is exactly the
"every peer pruned what I still need" case the description warns about, and it
becomes a zero crossing rather than a line-order comparison. Tip Gap gets the
right-hand axis because the two ranges differ by orders of magnitude
(measured: 299999..300001 against -1..1).
Both operands are gated `> 0`. Ungated, differencing the documented
"unknown window" sentinel of 0 yields the whole sequence space: measured
-105854935 for headroom and 105890295 for tip gap during the first ticks,
which destroys the axis for the rest of the window. Gated, the panel stays
blank until the node has a validated ledger and a peer has advertised a
range, which is the honest reading for that state.
Both queries verified against a live mainnet node through the full template
substitution: refId A = 300001 legend "History Headroom [xrpld-mainnet]",
refId B = -1 legend "Tip Gap [xrpld-mainnet]".
Runbook branch-C table, step 11 walkthrough and the 09 reference row follow
the rename and the new reading.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The panel built its legend with label_replace from the `from` label alone, so
every edge leaving a mode collapsed onto one series: connected->syncing and
connected->full both drew as "connected". The whole point of the from/to pair
is to tell a healthy climb from flapping, and that was exactly what the
legend hid.
label_replace cannot concatenate two labels. label_join can, which is the
pattern the consensus board already uses for its multi-label legends, so the
series now reads "from -> to".
Also regenerates the Grafana Cloud copy of the board.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
beast::Journal has no default constructor, so holding one as a plain
member deleted the suite's own default constructor and the Beast
registration macro could not instantiate it. SuiteJournal takes the
suite, converts implicitly where a journal is expected, and routes log
output into the test report.
Also removes a leftover unused local in run(), which is a hard error
under the warnings-as-errors build.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The nodestore suites moved from Beast to GTest upstream, which deleted
src/test/nodestore/TestBase.h. DatabaseConfig_test stayed on Beast and
still derived its journal and batch helpers from that base, so once both
sides met in a merge it referenced three symbols that no longer existed.
It now carries its own copies, matching the current API: node object
types are NodeObjectType::Ledger rather than the old hotLEDGER spelling.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
FetchReport::elapsed was milliseconds, so every nodestore read rounded to
zero: a warm store answers in single-digit microseconds and a cold one in
low hundreds, and both became 0 ms. That difference is the whole signal
separating a cold-read stall from a healthy node, and it was being
discarded at the type. Database::fetchNodeObject now measures once and
uses that one value for both the cumulative counter and the report, so
the two can never disagree. The job-queue call still takes milliseconds
and now casts explicitly.
BatchWriteReport::elapsed stays milliseconds and is documented as such:
a batch write covers many objects and reaches the disk, so it belongs in
that range.
Also adds a sub-millisecond histogram ladder, because the existing bucket
edges start at 100 microseconds and put the entire warm range in bucket
0. It is not wired to a view yet: no sub-millisecond instrument exists to
name, so the edges wait for the instrument that records read latency.
The new test captures what the nodestore reports and asserts the reported
total equals the internal microsecond accumulator exactly, plus that at
least one report is not a whole number of milliseconds -- which a
millisecond-typed field can never satisfy on any hardware.
Both files use telemetry::consensus::span::* but reached the declarations
only transitively, which misc-include-cleaner rejects. Add the direct
include now that the header lives under xrpl/consensus/.
The nodestore namespace became xrpl::node_store when develop was merged
in, but one mock override still named the old spelling, so it did not
match the ServiceRegistry signature it overrides. The sibling mock in
TestServiceRegistry.h was already correct.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Adds the derived read and write means, the NuDB writer depth and insert
timings, and the seven acquisition counters to the existing
nodestore_state gauge. Every value multiplexes onto that one instrument
through its `metric` label, so no new instrument is created.
Means are omitted rather than reported as zero when their denominator is
zero, so a dashboard shows a gap instead of a plausible wrong number. All
four go through one new scaledMean() helper so the guard cannot be
forgotten at a future call site; it also saturates instead of wrapping,
because a wrapped gauge reads as a healthy-looking dip. A zero total over
real samples still reports zero, since a store fast enough to truncate
every sample must not look dead.
The NuDB write-path block is skipped entirely when getWriteStats() is
nullopt, which is every backend but NuDB, so absent labels distinguish
"not measured" from "measured, and idle". Writer depth is scaled by 100
and named accordingly, because it sits just above 1.0 and an integral
gauge would truncate the whole signal away.
The gauge callback body is split into four static helpers to stay inside
the per-function line budget and to make each domain testable with a
recording sink.
Also corrects nudb_bytes, which called getStoreSize() exactly as
node_written_bytes does, so the obvious write-amplification ratio was a
constant 1.0 and the old "on-disk size" comment was wrong. No file-size
accessor exists on Backend or Database, so the value is unchanged and the
comment now states what it really is.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
clang-tidy had been skipped on this branch while CMake configure was
failing, so these findings in makeStageSpan surfaced only now:
- brace the three single-statement if bodies (readability-braces-around-statements)
- compare the pointer parameter explicitly against nullptr
(readability-implicit-bool-conversion)
- include xrpl/protocol/Protocol.h for LedgerIndex (misc-include-cleaner)
A saturated ledgerData lane makes TimeoutCounter re-arm its timer
without running the timer body, so timeouts_ never advances and the
six-timeout give-up can never fire. Acquisitions then neither finish
nor fail until the one-minute sweep destroys their partial maps, and
the work restarts. Every step of that chain was debug-log-only, so a
node at warning level could not be diagnosed after the fact.
The counters are separate on purpose: deferrals rising while timeouts
stay flat is the signature, and no single counter shows it.
Completions are recorded in done() rather than at the "Done: complete"
log line, because that line also fires for failures and misses the
checkLocal and receiveNode paths; done() is the one funnel every
outcome passes through and its signaled_ guard makes it idempotent.
AcquireStats is only forward-declared in ServiceRegistry so libxrpl
still includes nothing from xrpld. The src/ include path for the test
binary moves out of the telemetry guard, since a header-only type
under src/xrpld/ is testable in every build.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The consensus headers moved into the isolated xrpl/consensus module and
took a dependency on xrpl/telemetry for the tracing span constants, but
two things were left behind:
- Four includes still pointed at the old src/xrpld/consensus/ location,
which no longer exists, so the build failed with
"fatal error: 'xrpld/consensus/ConsensusParms.h' file not found".
- xrpl.libxrpl.consensus never linked xrpl.libxrpl.telemetry. add_module
isolates each module's headers, so xrpl/telemetry/SpanNames.h was not
on the include path even once the include was repointed.
Repoint the stale includes at xrpl/consensus/, and declare the telemetry
module before consensus so consensus can link it. Regenerate ordering.txt
for the resulting edge.
NuDB serializes every insert behind one global mutex held for the whole
call, so a caller cannot see how long it waited. Record instead the
writer depth joined at and the wall time spent; with mean depth L and
mean insert time W, Little's Law gives service time W/L and queueing
W - W/L. That distinguishes a serialized write path from a saturated
disk: measured on a dev box the device sat 89 percent idle while
throughput stayed flat at 42k inserts per second.
The accounting runs from a ScopeExit guard because the insert can
allocate and therefore throw; leaking the depth would strand the gauge
above zero for the life of the process.
getWriteLoad also stops returning a hardcoded zero. It now reports
writer depth, which is bounded by the writing-thread count and so stays
far below the kMaxWriteLoadAcquire cutoff that gates history
acquisition, where returning bytes or microseconds would have silently
suppressed it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The phase-2 merge re-added the pre-telemetry `add_module(xrpl tx)` block
without removing it, leaving the module declared twice. CMake's
add_library rejects a repeated target name, so configure failed before
any compilation:
add_library cannot create target "xrpl.libxrpl.tx" because another
target with the same name already exists.
Drop the stale pre-telemetry block and keep the one that follows
add_module(xrpl telemetry), which links both ledger and telemetry.
libxrpl/tx needs the telemetry link for the tx.transactor span, and
levelization already records `libxrpl.tx > xrpl.telemetry`.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The isValidJson2 call site in RPCCall.cpp already forces instantiation
of std::all_of over json::ValueConstIterator, so a regression that
removed the iterator traits would fail the real build. A dedicated
static_assert test is redundant; remove it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
json::ValueConstIterator and ValueIterator declared difference_type,
reference and pointer but not value_type or iterator_category. Under
C++23, std::iterator_traits then classifies them as output iterators,
so std::all_of over a Value's members (isValidJson2 in RPCCall.cpp)
fails to instantiate on GCC 13/14 with:
cannot convert 'output_iterator_tag' to 'std::input_iterator_tag'
GCC 15 masks this via LWG-3798/P2609, but the perf CI image ships
GCC 13, so the source needs the traits regardless. The iterators wrap
a std::map iterator (++/-- only), so the category is bidirectional.
Add value_type + iterator_category to both iterators, include <iterator>,
and add a regression test asserting the traits and that std::all_of /
std::count_if compile and run over Value members.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The gcc debug-coverage job rejected an unbraced `if` whose body is a GTest
assertion: EXPECT_EQ expands to an if/else, so the outer `if` leaves an else
that could bind either way, and -Werror=dangling-else refuses it. clang does
not warn, which is why only that one job failed.
Braced the span-names case that failed, then swept every test file this
branch touches for the same shape and braced the two others found, so the
next gcc run does not fail on the next one down the list.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The gcc debug-coverage job failed where clang, macos and windows passed:
gcc's -Werror=type-limits rejects comparing an unsigned expression against
zero with >=, because it is always true.
The assertion was vacuous anyway. Both operands are unsigned, so the check
proved nothing, while the comment beside it says the intent was a non-zero
microsecond figure. Now asserts strictly positive, which is what it meant.
The same job also logs a CMake LTO capability probe failing on a missing
compiler-ar and prints the code-generation guard's echo line; neither is
related to this branch.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
All eleven were reproduced locally against the same checks before fixing.
Six unused includes, left behind when the merge unioned two sets of includes
and later edits removed their only users: algorithm, functional, numeric and
thread in the job-queue test, and cstdint in the sync-state test and the load
manager. Each verified unused by grepping for every symbol the header
provides, so none is a still-needed include being dropped.
The telemetry registry header included ranges for a std::ranges::all_of call,
but that algorithm comes from algorithm, which the header already included.
Two consteval handler-name loops became std::ranges::all_of, which reads as
the predicate it is, and the two flagged fixtures are const.
Also picks up the levelization baseline the check asked for: the consensus
span-name test adds one edge from the libxrpl tests to xrpld.consensus, which
is the exact line CI's diff requested.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two suspects from the 3.3.0 slowdown investigation had no signal. Both were
already computing the numbers and throwing them away, so this exposes them
rather than adding measurement.
Per-sweep heap trim. The trim runs after every cache sweep, and its cost
scales with resident heap, so it is the leading explanation for a node with
a populated database syncing slower than a fresh one. The report already
carried duration, fault deltas and reclaimed pages, but the whole
measurement sat behind a debug-journal check, so an ordinary node measured
nothing, and the call site discarded the result. The measurement now always
runs and only the log line stays gated. Records trim duration, minor faults
and reclaimed kilobytes. Measured cost of the always-on path is about six
microseconds per sweep against a trim costing milliseconds, at a cadence of
ten to a hundred and twenty seconds.
Honest limit, stated in the runbook: the fault delta spans only the trim
call, so it shows the trim itself faulting but not the faults that follow as
caches refill. The duration is the signal to correlate against sweep-job
queueing.
Rotation writes. Rotation copies archive-served reads forward and re-stores
nodes missing from both backends, both of which compete with sync I/O and
only happen on a populated online_delete database. The copy-forward count
existed but was reset by the rotation's own log line, so a metric reading it
would drop to zero on every swap; a never-reset total sits beside it now.
The re-store count was not measured at all. Rotation duration is
deliberately not recorded: the health throttle sleeps at eight points inside
the sequence and dominates exactly when the node is unhealthy, so the number
would conflate work with waiting.
Nothing added for the other two suspects. Get-object serving is already
covered by the handler label, the lookup histogram and the deferred and
saturation gauges; peer churn by the disconnect-reason counter.
Also replaces nine per-file cspell ignores with one ignoreRegExpList entry
for the telemetry macro names, and picks up the levelization baseline for the
consensus span-name test.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
fetchDurationUs_ had no getter, so telemetry reached it by building a JSON
object, stringifying a uint64 and parsing it back with stoll on every
collect tick. storeDurationUs_ was declared and never written or read at
all, along with the jss::node_writes_duration_us key.
Both now have accessors, both production store paths time their backend
call, and the registry reads them without the round trip. get_counts also
reports the write duration, so the RPC and the metric agree. Mean read
latency is the signal that separates a cold store from a warm one: warm
reads are single-digit microseconds, cold ones low hundreds.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Both were uint32_t while every sibling counter was uint64_t. On a
multi-day node node_read_bytes read 1,481,244,491 for 1,894,924,394
reads, i.e. 0.8 bytes per read, which is impossible: the counter had
wrapped about 350 times. Read hit rate is used to tell a cold-read
stall from a write-lock ceiling, so a wrapping numerator makes that
diagnosis wrong rather than merely imprecise.
getFetchTotalCount() was already backed by a uint64_t member, so its
32-bit return type truncated a correct value at the accessor. All
three getters now return uint64_t.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Found by reviewing what each metric actually measures, with attention to the
derived and bucketed ones. All five could report healthy while the node was
not, or the reverse.
- The nodestore latency panel took rate() of a mean. The gauge already
divides duration by count in code, so rating it produced a figure with no
unit, and Prometheus discards a gauge's decreases, so a heavy back-fill
read as roughly zero microseconds per operation. The cumulative duration
totals are now exported alongside the means, and the panel divides the
rate of the total by the rate of the count, which is the latency over the
panel's own window rather than a since-boot average that flattens with
uptime.
- The DNS-resolve and outbound-dial histograms had no explicit buckets, so
they inherited a ladder that stops at ten seconds while the dial timer is
fifteen. Every timed-out dial fell in the overflow bucket and p95 read
exactly ten seconds however bad it got. Both now have a ladder reaching
thirty seconds with fifteen on its own boundary, so a timeout is
distinguishable from merely slow.
- The missing-node counts only cleared when a tree completed, so a
timed-out or failed acquire left its last count latched. Since the gauge
reports the maximum across everything still in the collection, and
eviction waits on a grace period plus the sweep interval, a finished node
reported as stuck for minutes. That inverts the one signal that separates
stuck from slow. Cleared unconditionally on the terminal path instead.
- A disabled quorum published a sentinel so large that, on a timeseries
axis shared with the trusted-key count, it flattened the key line to the
baseline and hid the outage it was meant to mark. The series is now
omitted and a quorum_disabled flag carries the state.
- Two panel descriptions claimed a one-second export cycle. The reader is
configured for ten.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The write-duration test encoded the behaviour from before the store path was
timed. It populated the source database through store(), then asserted that
database's write duration was still zero, which was only true while
importDatabase was the sole timed path. Both concrete store overrides now
time their backend write, so an ordinary store run accumulates a duration
and the assertion failed.
Turned that stale assertion into a positive one: an ordinary storeBatch must
produce a non-zero duration, which is the case a real node actually
exercises. The negative half of the test, proving the accumulation is
per-database rather than a shared global, now checks that the import leaves
the source's store COUNT unchanged, since the source's duration is
legitimately non-zero from its own writes.
80552 tests across the six affected suites pass.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
These are the first failures found by actually building and running the new
telemetry tests. All three were defects in the tests, not the metrics.
- state_changes_total expected six series from seven transitions, but two
edges were traversed twice, so there are five distinct from/to pairs. The
repeat is the point of the label: flapping raises the count on one edge
rather than minting a new series, so the assertion now says five and
explains why.
- shamap_cache_hit_rate compared for exact double equality against a value
that reaches the gauge through a float hit rate, so 0.9 arrives as
0.89999997. Now compares within a tolerance far tighter than any
threshold a dashboard reads, since asserting exact equality was only
asserting the float representation.
- nodestore_latency shared one provider across four scenarios. The reader
reports cumulative temporality, so a mean observed by an earlier scenario
was still present in the next collection, which defeated the two
assertions that a mean is absent when its denominator is zero. Each
scenario now collects from its own provider.
139 of 139 telemetry tests pass.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Phase-10 independently instrumented the peer object-fetch path while this
branch instrumented fresh-node sync, so the two overlapped in three places.
Resolved by keeping each side's stronger implementation rather than shipping
both.
Per-job-type waiting/running/deferred existed twice. Phase-10's version
survives: it publishes per-type gauges from JobQueue::collect(), which
snapshots under the queue lock and publishes after releasing it, a
deliberate lock-order fix against the collector's own lock. This branch's
jobq_backlog gauge and the JobQueue::getJobTypeCounts() accessor that fed it
are removed, along with their panels, assertions and reference rows.
jobq_saturation stays: it reports the whole worker pool, which phase-10 has
no equivalent for.
The histogram view helper also existed twice with identical bodies under two
names; one survives, and the microsecond ladder is now the named array
rather than boundaries repeated inline. The job_type label was declared
twice, once as a file-local constant invisible to the naming check; both it
and handler now come from the constants header.
Two things phase-10 adds are complementary, not duplicates, and are kept as
they are: the handler label, which separates the two request kinds that both
report as the same job type, and getobject_rejected_total, which counts
malformed requests where this branch's serve_refused_total counts requests
this node declined to serve.
Also fixes two naming-check failures that pre-date this merge on phase-10.
The check derived label keys only from namespaced constants, so it could not
see the per-subsystem headers' flat k-prefixed style and rejected dashboards
querying labels the code really emits. It now reads both styles, with the
enforcement rules unchanged.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Metric names and label keys were bare string literals, repeated across the
emit site, the gauge registration, the unit test, the workload manifest, the
dashboard queries and the reference table. A rename touched six places and a
typo in any one of them failed silently: a metric that never appears, or a
label that never joins.
The span side already had this right, with names and attribute keys declared
once in the *SpanNames.h headers and a CI rule rejecting literals at call
sites. That rule only ever covered spans, so the metric side had no
equivalent and no suffix convention was enforced by anything.
- Adds MetricNames.h declaring every instrument name, label key and bounded
label value this story emits, grouped by subsystem, following the existing
span-name header layout.
- Converts the call sites subsystem by subsystem. The emitted strings are
unchanged: 75 names before, the same 75 after, verified by extracting the
wire strings from both trees and diffing the sets.
- Extends the naming check with three rules: no literal instrument name or
label key at an emit site, the duration and counter suffix conventions,
and every name in the workload manifest resolving to a constant. The
first rule is ratcheted per metric family so the pre-existing families
warn rather than block, keeping the remaining work visible instead of
forcing one unreviewable change.
Constants are character arrays rather than the span headers' StaticStr,
because the metrics API takes a string view that will not construct from it.
Two things the conversion exposed: a serve-refusal reason that the original
inventory missed because it is passed through a ternary, and a label whose
constant made it invisible to the checker's literal scan, which would have
failed a dashboard rule.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The board and runbook had grown by append across eight work packages, so
they read in the order the work was done rather than the order a node
progresses. This is the coherence pass; it adds no new instrumentation.
- Dashboard: 52 panels regrouped from two rows into nine that follow the
fresh-start sequence — bootstrap, peer supply, sync state, acquire and
SHAMap fetch, job queue, quorum and publish, terminal blockers, then
back-fill and spans collapsed since they answer conditional questions.
Layout only: no title, query or description changed.
- Runbook: the flat step list becomes a decision tree branching on the
observed symptom, with the amendment-block check first because it is
terminal. Each branch names the panels, what healthy and unhealthy look
like, and what to conclude. The existing steps are kept as the detail
bodies.
- Reference table: every signal name re-checked against the code and every
named panel against the board; four stale panel references fixed.
- Validation: every signal is now either asserted or covered by a note
explaining why a five-node local cluster cannot produce it.
Also fixes the write-latency signal, which was inert on a real node: the
store duration was only recorded on the database-import path, while the two
production store implementations did not time themselves, so an ordinary
node reported a write count with no latency. Both now time the backend
write, which is the disk work this signal exists to expose. Without it the
"existing database syncs slower than a fresh one" diagnosis had no primary
signal.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Completes the join left open when the acquire and consensus work landed in
parallel. The acquire span now derives its trace id from the ledger hash it
already records, so a fetch shares one trace with that ledger's validation,
acceptance and store, and its three phase children inherit the same id.
Reading one trace now answers the whole question for a slow ledger: whether
the data was slow to arrive, slow to be accepted, or slow to persist.
The acquire span stays an optional member of the join group, for the reason
its own entry already gives: a healthy cluster agreeing from genesis rarely
back-fills, so the span need not appear on every run.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A slow fresh-sync ledger produced spans scattered across threads with no
way to relate them. They now share a trace id derived from the ledger's own
hash, the one value every participating site already holds, so nothing new
is plumbed across threads. This is the pattern the transaction pipeline
already uses for its tx id.
Joined: ledger.validate, ledger.store, and a new
consensus.validation.accept recorded when a trusted validation arrives. In
Tempo, searching one ledger hash returns them together, so an operator can
tell whether the ledger was slow to arrive, slow to be accepted, or slow to
be stored. They are siblings rather than a chain because the accept gate is
entered from three different threads, so no fixed parent order exists.
consensus.validation.accept also records why an arriving validation did or
did not advance the gate, which makes "validations arrive but are all
rejected" visible for the first time.
consensus_round_duration_ms turns the existing round-time span attribute
into a histogram, so a fleet trend needs a metric query rather than raw
trace inspection. An explicit bucket view is required, not optional: the
SDK default tops out at ten seconds while consensus abandons a round at two
minutes, so slow rounds would all fall in one bucket and every quantile
would read exactly ten seconds. Cost is one record per round.
Record layer: the histogram is native and needs no collector change. The
two new bounded attributes are added as span-metric dimensions to both
collector configs. The ledger hash stays out of them, since a per-ledger
dimension mints a series per ledger; it is indexed in Tempo as the join key.
The ledger.acquire span is not joined yet, because that file was being
changed concurrently. It is registered as an optional member of the join
group so nothing fails, and switching it is a one-line follow-up.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Four blind spots in the sync exchange, each now a span:
- txset.acquire: transaction-set acquisition had no span at all, though it
is the sibling of ledger.acquire and runs every consensus round. A round
that falls behind because its tx set never arrived was indistinguishable
from one that deliberated slowly.
- ledger.acquire.{header,astree,txtree}: the acquire span was flat, so the
account-state tree, which dominates a fresh sync, could not be separated
from the transaction tree. These are children, closed before the parent.
- peer.dial: the outbound dial already had outcome counters; the span adds
the per-attempt timeline, so a slow stage is visible rather than only its
terminal reason.
- ledger.serve: serving a peer's ledger request was uninstrumented, so this
node's contribution to someone else's sync was invisible.
Every span finalizes exactly once. Outcomes come from shared compile-time
rules rather than a literal per branch, so no exit can mislabel itself and
an exit added later cannot omit one. Destructor paths are noexcept.
One rule needed care: the timeout path also sets the failure flag, because
that is how the timeout loop stops, so precedence puts timeout ahead of
failure or a timed-out acquire would read as a data fault.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The quorum and publish gauges were emitted but never surfaced: no panel, no
harness assertion, no reference entry. Completes those layers.
- Four panels: trusted validations against the quorum target on one axis so
a tally climbing toward quorum is visually distinct from one flat below
it; publish lag; pre-accept shortfall rate; and time to first validated
ledger.
- Both signals are asserted by the workload validator. The shortfall
counter does fire on a healthy cluster, because this node validates and
then immediately re-enters the accept gate before its peers' validations
arrive, so the first evaluation of every round tallies short. The panel
and note say so, and give the fault signature instead: the shortfall rate
outpacing the ledger-close rate while the tally stays flat and nothing
ever reaches first-validated.
- The quorum target is deliberately drawn as its own line rather than as a
headroom stat, so the disabled-quorum sentinel reads as an unreachable
target instead of an unreadable negative number.
Also removes three reference rows that were appended twice when two agents
each documented the same back-fill signals.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Registers the new gauges, renders them, asserts them and documents them, so
each signal reaches an operator rather than stopping at the emit site:
- MetricsRegistry: gauge registration for ledger_quorum_publish,
nodestore_latency, peer_ledger_supply, peerfinder_slot_census and
amendment_block, each guarded by the detached-callbacks check and
tolerant of services that are not ready yet.
- Ledger Sync Health dashboard: panels for the new signals, filtered by
the node template variable like every other board.
- Workload validation: the new series are asserted, so a signal that
regresses to absent fails CI. Signals the local cluster structurally
cannot produce, such as a replay fallback or an amendment block, are
noted rather than asserted, which would fail red on a healthy run.
- Reference, runbook and glossary entries, including the diagnosis order
for a node that has peers and validators but never validates.
- Regenerated levelization baseline: three new one-way edges from the
telemetry and test modules, no new cycles.
Also drops an unused cstddef include from the macro tests, which the
include checker rejects.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The span only recorded an outcome on the normal completion path. An
acquire that stalled and was later swept ended with no outcome at all, and
its duration stretched to the sweep interval rather than the real fetch
time. So the one case these signals exist to catch, a fetch that never
finishes, was the one case that could not be traced, and aggregate outcome
and timeout rates read low exactly when nodes are stuck.
- Adds an abandoned outcome value for the swept-while-fetching case.
- Routes every exit through one idempotent finalizer, so a span is
finalized exactly once whether it completes, fails, short-circuits on
local data, or is destroyed mid-fetch. The destructor path cannot throw.
- Adds the ledger hash to the span and backfills the sequence once known,
since by-hash acquires start without one and could not otherwise be tied
to a specific ledger.
- Record layer: outcome stays a span-metrics dimension in both collector
configs, which drift apart if only one is edited. The ledger hash is
indexed in Tempo for trace search instead, because a per-ledger value as
a metric dimension would mint a new series every ledger.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Quorum and publish (A5):
- ledger_quorum_publish gauge: the trusted-validation tally against the
quorum the candidate ledger must reach, plus the gap between the
validated and published sequences. The published sequence was never
exported, so a publish pipeline falling behind healthy validation was
invisible.
- ledger_quorum_shortfall_total: counts the pre-accept early return where
a node has peers and validators yet still declines to declare a ledger
validated. That path was log-only, and it is the difference between
accumulating toward quorum and never reaching it.
Back-fill and persistence (A6):
- nodestore_latency: read and write service time. storeDurationUs_ was
declared but never written and had no accessor, so there was no write
latency signal at all. This is the direct fingerprint of a node with an
existing database syncing slower than a fresh one, where the node cache
is cold and every tree step reaches disk. Distinct from the existing
NuDB read panels, which show volume and hit ratio rather than service
time.
- ledger_replay_fallback_total and ledger_replay_outcome_total: the replay
path silently falls back to plain acquisition on timeout or failure, so
a defeated optimisation looked like ordinary slow back-fill.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Whether the network can even serve this node, and whether the node is
about to be shut out of validation, were both invisible:
- peer_ledger_supply: how many connected peers advertise a range covering
the sequence being fetched. Peers each track a range from status
changes, but nothing aggregated them, so "nobody has what I need" looked
identical to "peers are slow".
- peerfinder_slot_census: outbound active against capacity, connection
attempts, inbound, fixed configured against active, and the bootcache
and livecache sizes. All were computed already; only two were exported,
read at unrelated instants, so they could not be compared.
- peer_disconnect_total{reason,direction} and peer_accept_total{outcome}:
every disconnect previously collapsed into one number, so our own
backpressure could not be told from topology or network faults. Reasons
are a fixed set of literals recorded on the peer and emitted once at
close, never data supplied by the remote end.
- serve_refused_total{request,reason}: the other half of the sync
exchange, when this node declines to serve a peer.
- amendment_block: whether an unsupported amendment is expected and how
long until it activates. Amendment-blocked is terminal for validation,
so the countdown is the only leading indicator. The amendment id is
deliberately not a label, since the network can vote an id this build
has never heard of; it is already logged.
- ledger_jump_total: repeated last-closed-ledger switches, which mean the
node is thrashing between chains.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Sync-critical job types run at very low concurrency limits (ledgerRequest
and ledgerData allow 3 each), so a node can stall simply because those
jobs are held back behind other work. Nothing exposed that until now:
the existing job metrics are rates and quantiles of jobs that already
moved, or a single queue-wide depth.
- jobq_backlog{metric,job_type}: instantaneous waiting, running and
deferred counts per job type. Deferred is the starvation signal and had
no exposure anywhere; it is set when a type is at its concurrency limit.
- jobq_saturation{metric}: running tasks, worker-thread count and total
waiting, so a slowdown spanning several subsystems can be attributed to
worker-pool exhaustion instead of being diagnosed once per victim.
Both read through two new const accessors on JobQueue that take the
existing mutex once and copy integers, so a single reading is internally
consistent and no per-job cost is added. The job_type label reuses the
same JobTypes name helper the existing job counters use, so the two label
sets join.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Post-deploy verification against Grafana Cloud found two problems on the
job-queue dashboard.
Every aggregating panel's displayName names four labels, but the queries
aggregate with `sum by (...)` lists that omit `xrpl_work_item`, so that
part of the legend rendered as an unresolved reference. Pre-existing, but
splitting the wait and execution panels by job type made the legend
load-bearing, so it is fixed here: `xrpl_work_item` is added to the ten
affected sum-by lists.
The two panels split by job type in the previous commit also kept their
seven-section descriptions. They now carry the same ten sections as the
rest, and their glossary links point at the real `job-queue-job-type`
anchor rather than a `job-queue` anchor that does not exist.
Verified no dashboard links a glossary anchor that is absent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The eight panels added for the GetObject and job-queue metrics did not
follow the conventions the rest of the dashboards use.
- Descriptions go from seven sections to the ten used by the other 159
panels, adding Keywords, Computation boundary and References. The
existing diagnostic guidance is kept; only the format changed.
- Six glossary entries added for the terms those References link to
(concurrency limit, deferred job, handler label, NodeStore lookup
hit/miss, resource charge), so no link is dead.
- displayName becomes `${series} ${xrpl_ident}`, the form 122 of 143
panels use. This needs the label_join wrapper that builds xrpl_ident,
which these queries lacked, so it is added to ten targets; without it
the legend would render an unresolved label.
- Legend blocks now match each dashboard's own convention rather than
being split three-with and four-without across the new panels.
Job Queue Wait Time and Job Execution Time dropped job_type from their
aggregation, so every queue collapsed into one line and the legend could
not say which queue was slow. Both now split by type, capped with topk
to stay readable, matching the per-type panel already on that dashboard.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Resolved five files. In each case both sides had content worth keeping,
so nothing was taken wholesale:
- MetricsRegistry.cpp: kept the incoming `handler` label on the job
instruments and re-applied this branch's `queuedDurUs >= 0` guard,
which the incoming side does not have.
- telemetry-runbook.md: took the incoming gauge table, which adds the
three per-job-type rows, and re-applied this branch's corrected
`jobq_job_count` name.
- 09-data-collection-reference.md: kept this branch's validation
inventory (newer counts, extra Config File column) and inserted the
incoming call-site and per-job-type gauge rows plus their explanation.
- node-health.json: merged structurally rather than by text. This
branch's panels are authoritative; only the two incoming job-queue
panels were appended, below the existing layout. The
`Validated Ledger Seq` panels added directly in Grafana are preserved.
- job-queue.json, ledger-data-sync.json: panel sets were identical, so
took the incoming side for its `$handler` variable, the handler filter
on existing queries, and the new panels. Verified no panel was lost.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Resolved OpenTelemetryPlan/09-data-collection-reference.md: both sides
document the same new metrics, reached by different routes. The incoming
sections came from the phase-6 edit merged forward; this branch already
documents the same surface in its own layout, with the job lifecycle
metrics and the GetObject request path under the MetricsRegistry
section rather than a separate 2a.
Kept this branch's sections to avoid duplicate entries for every metric.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Resolved OpenTelemetryPlan/09-data-collection-reference.md: this branch
replaced the StatsD-shaped Section 2 with OTel-native naming, so the
incoming StatsD tables and the StatsD per-job timer section do not apply
here. Kept this branch's names throughout.
The incoming per-job-type saturation gauge section is new content and is
retained, renumbered to 2.5 and rewritten for the OTel naming: bare
lowercase `jobq_<jobtype>_{waiting,running,deferred}` derived through
OTelCollector rather than the case-preserving `xrpld_`-prefixed StatsD
form.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Slowness on the peer object-fetch path could be observed but not
attributed. Job duration metrics carry only `job_type`, and both
`RcvGetLedger` and `RcvGetObjByHash` report as `ledgerRequest`, so a
queue-wait spike could not be traced to a handler. Nothing measured
NodeStore cost, request size, or the differential charge.
Latency now decomposes into three additive parts, each separately
measurable:
end-to-end = queue wait + NodeStore lookup + everything else
- `handler` label on job_queued_total/_started_total/_finished_total and
job_queued_us/job_running_us. The value is sanitised: a name passes
through only if non-empty and all ASCII letters, else "other". Two job
names embed a ledger sequence, so a raw label would mint one series
per ledger; the rule bounds the domain at 43 names plus "other".
- getobject_lookup_us, _request_objects, _lookups_total{result},
_rejected_total{reason} and _charge, recorded at their call sites.
All three histograms get explicit bucket views: the SDK default stops
at 10,000, which every one of them exceeds.
- Per-job-type waiting/running/deferred gauges for the 35 non-special
job types. `deferred` is the leading indicator, since addJob never
rejects -- it defers, so backpressure otherwise shows up only as
latency after the fact.
`JobQueue::collect()` snapshots the counters under the queue lock and
publishes gauges after releasing it. Writing them while holding the lock
would invert a lock order against the collector's own lock, which the
collector's flush thread already holds when it calls this hook.
Tests assert exact values, including that the charge is priced on the
requested count rather than the capped iteration count.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Add reference entries for the observability surface introduced on
phase-9: the `handler` label on the job instruments, the five
`getobject_*` request metrics, and the per-job-type queue saturation
gauges.
Names here follow this branch's StatsD pipeline, which preserves case
and carries the `xrpld_` prefix, so they differ from the lowercased
OTel-native names used from phase-7 onward. The sections state where
the implementing code lives, since it is introduced downstream.
Also correct pre-existing entries: `job_count` exports as
`jobq_job_count` via the collector group prefix, the non-special job
type count is 35 (not 36), and `JtLedgerData` has five producers.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signals that separate a sync that is merely slow from one that will never
finish:
- sync_acquire{missing_state_nodes_max, missing_tx_nodes_max, in_flight,
received_data_depth}: how many SHAMap nodes each in-flight acquire is
still waiting for. getMissingNodes already computed this and the callers
discarded it after a trace log. A count that stays flat means the
acquire is wedged; a shrinking count means it is progressing. Recorded
once per sweep, never inside the per-node walk, and reset when a tree
completes so a finished acquire does not read as stuck forever.
- shamap_cache_hit_rate{treenode}: hit rate of the in-memory tree-node
cache, which sits above the node store, so it is distinct from the
existing NuDB ratio. A cold cache on a fresh node sends every traversal
step to disk.
- sync_acquire_no_progress_total: timer ticks where an acquire made no
progress, previously only logged.
- sync_addnode_total{good,duplicate,invalid}: whether arriving nodes are
useful, duplicated or rejected, so wasted fetch work is visible.
- sync_acquire_source_total{local,network}: whether a ledger was served
from the local store or had to be fetched.
Adds getBad()/getDuplicate() to SHAMapAddNode and an acquireProgress()
accessor on InboundLedgers so the xrpld gauge can read these without
libxrpl depending on telemetry.
ledger_seq is deliberately not a metric label: it is unbounded. Per-ledger
identity stays on the ledger.acquire span; the metrics expose bounded
aggregates instead.
The full-below cache hit rate is not exported: KeyCache updates different
counters than getHitRate() reads, so it would always report zero. That
libxrpl bug is documented rather than papered over.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Five signals that explain why a node is not advancing toward full, none of
which were observable before:
- state_changes_total now carries {from,to} mode labels, emitted at
setMode using the existing strOperatingMode helper. A bare count could
not distinguish a healthy climb from a node flapping between tracking
and connected. Removes the now-unused incrementStateChanges wrapper.
- sync_state{initial_full_duration_us}: time to first reach full, which
StateAccounting already computed but exposed only in server_info.
- sync_state{network_ledger_gate}: whether the node is still refusing to
build ledgers because it has no network ledger.
- sync_state{server_stall_seconds} and server_stall_events_total: how
long the main thread has been unresponsive. LoadManager computed this
and only logged it, so a stall was invisible until the fatal threshold.
The episode rule is a pure function so it can be tested without adding
a test-only mutator to LoadManager.
- sync_state{ledgers_behind}: how far our validated sequence trails the
best sequence any peer advertises, read from already-cached peer ranges
so no extra network traffic is added.
Also fixes the naming checker: it derived only the first label of a
multi-label instrument, so a dashboard querying the second label was
wrongly rejected.
Note: the clang-tidy hook cannot run in this worktree (no build
directory); the remaining pre-commit hooks, the naming check, dashboard
schema and harness syntax all pass.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Three defects found reviewing the WP-A1 commit:
- Handshake.cpp moved a std::string into std::runtime_error, which has no
rvalue constructor. The move never happened and clang-tidy rejects it
under performance-move-const-arg, so CI would fail even though the
local hook only runs clang-tidy with TIDY=1. Takes the message by const
reference instead, and drops the <utility> include that existed only
for that move.
- ValidatorList disables quorum by returning SIZE_MAX. Casting that to
int64_t wrapped it to -1, so the headroom panel computed
0 - (-1) = +1 and coloured yellow on a node that can never validate:
the sign inverted in exactly the bootstrap failure these signals exist
to catch. Reports the disabled state as int64 max so headroom goes
strongly negative instead.
- The runbook claimed an expired list loads no keys. Expired counts as
accepted, so its keys are loaded and then dropped by the expiry sweep,
which calls for a different fix than replacing validators.txt. Pending
is likewise a future-dated refresh, not a rejection. Documents both,
plus how the quorum-disabled state now reads.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A freshly started node most often stalls before it ever peers or reaches
quorum, and that whole chain had no telemetry. Adds the six signals that
make it observable:
- dns_resolve_total / dns_resolve_latency_ms: configured-peer hostname
resolution, emitted from OverlayImpl so libxrpl stays independent.
- overlay_connect_total / overlay_dial_latency_ms: outbound dial outcome
by terminal reason, plus dial duration.
- handshake_negotiation_fail_total: protocol and network-id negotiation
rejections, labelled by reason, so a misconfigured network is no longer
indistinguishable from unreachable peers.
- unl_fetch_total and the unl_quorum gauge: validator-list fetch outcome
per site and trusted key count against the required quorum. Without
these a bad validators.txt leaves the node syncing forever with no
signal.
- clock_close_offset_seconds: network close-time offset, which server_info
hides below 60s but which stalls consensus participation.
Panels land in the Bootstrap row of the Ledger Sync Health dashboard, the
metrics are asserted by the workload validator, and both the reference and
the runbook flow describe them.
Levelization baseline regenerated: overlay now includes MetricMacros.h, so
the overlay/telemetry pair is reported one-way instead of bidirectional.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Match the panel height the operator settled on in Grafana (18 rows) so
all 20 ranked bars render without an inner scrollbar, and reflow the
Sync Diagnostics row and the panels below it accordingly.
Adds the anchors the sync-diagnostics signals attach to, with no signals
emitted yet:
- New "Ledger Sync Health" dashboard (uid ledger-sync-health) with the
standard template-variable block copied from an existing board, plus
empty "Bootstrap (Domain 0)" and "Sync pipeline" rows.
- Signal index section in the data-collection reference, an operator-flow
stub in the telemetry runbook, and a glossary anchor.
- A sync_diagnostics group in expected_metrics.json and a matching
assertion helper in validate_telemetry.py so CI fails when a signal
regresses to absent.
Also registers the new dashboard uid with the harness so the board is
covered by validation.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The bar labels on 'Overlay Traffic Heatmap (All Categories, Bytes In)'
were raw metric names carrying a redundant '_bytes_in' suffix, and the
half-width 8-row panel truncated both the category and the node
identity.
- Strip the '_bytes_in' suffix from the derived series label; the panel
title already states the metric is inbound bytes.
- Widen the panel to full width and grow it to 12 rows so 20 bars render
with their full category and node labels.
- Shift the Sync Diagnostics row and the panels below it down by 4 to
keep the layout gap-free.
- Note the label derivation in the panel description.
The panel plotted only Full and Tracking, so the states a node actually
passes through while catching up were invisible. On the dev box's cold
sync the node spent ~19 minutes in Connected and ~3 seconds in Syncing,
while Tracking totalled 2 microseconds -- the one non-Full line the
panel did draw was the least informative of the five.
Add Syncing, Connected and Disconnected series (state-ladder order,
matching the Operating Mode Transitions panel) and rename the panel to
"State Duration Rate (All States)". Because the node is always in
exactly one state, the five rates sum to ~1.0, so the panel now reads as
"which state is time going into right now", and a handover between two
lines marks a state change with its width showing the dwell time.
Description expanded to the full section set (keywords, computation
boundary, references) used by the other panels on this dashboard.
Verified live: rate(state_accounting_connected_duration)/1e6 reaches
1.0 then 0.387 across the catch-up window and syncing reaches 0.0103 --
both previously undrawable.
The dashboard showed 54 individual signals but no single answer to
"is this node healthy and doing its job?". A reader had to correlate
server state, ledger age and peer counts by eye.
Add a stat panel at the top that reduces the two conditions that define
a healthy observer node to one green/red verdict:
server_state == 4 (Full) AND validated_ledger_age < 30s
Both terms use `== bool` so each yields 0/1, joined with
`* on(service_instance_id)` so the verdict is per node. Value mappings
render 1 as "Healthy" (green) and 0 as "Not Healthy" (red) with
background colouring, so the state is readable at a glance.
Placed at y=0 per the dashboard convention that gauges/stats lead;
existing panels shift down by 4 rows with no other change.
Verified live against Grafana Cloud: returns 1 for aws-dev-xrpl-1
(Full, validated seq 105,824,596).
The four aggregated Sync Diagnostics panels grouped by xrpl_work_item,
which no layer in this repo emits (it is injected by the perf-iac
harness). That tripped Rule D of the telemetry naming check:
D ledger-data-sync.json xrpl_work_item
must exist in L1, a metric label, or be a builtin
Align with the convention used by every other aggregated panel in the
dashboard set: group by (service_instance_id, xrpl_branch,
xrpl_node_role) and leave the xrpl_ident label_join untouched. The
legend is unaffected -- label_join over a label dropped by the
aggregation contributes an empty segment, which the trailing
label_replace already strips.
Verified live: both the NuDB ratio and the job-queue p95 queries still
return per-node series rendering as "[aws-dev-xrpl-1]".
The 0-4 values now render as named states via value mappings, so the numeric
axis label is misleading. Rename to "Server State".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add value mappings (0 Disconnected, 1 Connected, 2 Syncing, 3 Tracking,
4 Full) and matching red->green thresholds to the Sync State panel, so the
line, tooltip, and legend render state names and colors instead of bare 0-4.
Keeps the minimal custom-key convention; only the Sync State panel changes.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rework the 10 Sync Diagnostics panels' legends to match the format used by the
existing panels: wrap each query in label_join + label_replace to build the
xrpl_ident label ([service_instance_id, xrpl_branch, xrpl_work_item] with empty
values stripped) and set displayName to "${series} ${xrpl_ident}". Verified
against a live node: renders as "Series [aws-dev-xrpl-1]" with no empty-label
gaps, identical mechanism to the original 15 panels. Only the new panels change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a 10-panel "Sync Diagnostics" row that shows ledger-sync slowdown and its
causes, laid out top-to-bottom as symptom -> latency -> root cause:
- Red flags: Sync State, Validated Ledger Age, Ledger Close Rate
- Latency: Job Queue Wait p95 by type, NuDB Read Latency, I/O Scheduler p95
- Root cause: NuDB Cache Hit Ratio, NuDB Read Pressure, Job Queue Depth,
Load Factor & Peers
All panels use native beast::insight metrics introduced on this branch
(nodestore_state, ledgermaster_*, jobq_*_q, ios_latency, load_factor_metrics,
peer_finder_*) plus the consensus.mode_change span metric. Queries filter by
the dashboard's $node and tier variables and were verified against a live
node. Nodestore ratio panels use sum-by(identity) matching to bridge the
metric= sub-label.
Format matches the dashboard convention: minimal custom keys, ${DS_PROMETHEUS}
datasource uid, and the standard What/How/Reading/Healthy/Watch/Source/Function
description structure. Pure-append; existing panels untouched.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Document the two new ledger-correlation attributes in the data-collection
reference: add them to the transaction and TxQ attribute tables (with per-span
coverage), update the attribute count, and correct the txq.enqueue parent note
(parents to tx.process on the submission path via explicit context; a root on
the open-ledger rebuild path, correlated via current_ledger_seq).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add current_ledger_seq / current_ledger_hash to the tx.process, tx.receive,
and txq.enqueue span-reference rows, correct the txq.enqueue parent note
(parents to tx.process on the submission path via explicit context; a root on
the open-ledger rebuild path), and add a "Correlating a transaction to the
ledger it was worked on" recipe joining the txID-keyed tx/txq spans to the
ledger trace via current_ledger_seq.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add shared current_ledger_seq / current_ledger_hash span attributes so a
transaction's work can be joined to the ledger trace that produced it, and
fix discrepancy D1 (txq.enqueue was a detached trace root).
- Define current_ledger_seq / current_ledger_hash once in SpanNames.h and
re-export via `using` from TxQ/TxApply/Tx span-name headers. These name the
ledger being worked on (open/tentative apply or in-flight consensus build),
distinct from ledger_seq (the built/validated ledger on ledger.build /
consensus.round). Named after the RPC field ledger_current_index.
- txq.enqueue: set current_ledger_seq/hash from the view, and parent the span
to the caller's tx.process span via an explicit captured SpanContext (new
trailing TxQ::apply param) instead of a detached root. The parent is
explicit, not ambient-inherited, and the ScopedSpanGuard scope is RAII-bound
to the synchronous apply, so it cannot leak onto a reused worker (D1 fix).
On the open-ledger rebuild path no tx.process context exists, so it stays a
root and the attribute provides the correlation.
- tx.preclaim / tx.transactor: set both attributes from their ledger view.
tx.preflight is stateless (no view) and is the documented exception.
- tx.process / tx.receive: set current_ledger_seq from the current open ledger
index at submit/receive time (no hash: not yet applied to a ledger).
- Contract test pins the two new attribute key strings.
Neither key is a spanmetrics dimension, so there is no metric-cardinality
impact. Dashboards/collector/docs land on the later phases per the chain split.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The colocated span-name headers (RpcSpanNames.h, ConsensusSpanNames.h,
PeerSpanNames.h, etc.) hold only compile-time constant strings and live
next to their subsystem rather than under a telemetry/ directory, so the
existing dir-prefix ignores miss them. Add a **/*SpanNames.h glob so they
do not count against patch coverage; like the rest of the telemetry
surface they are not exercised in the coverage build.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace the raw "trace_id="/" span_id=" prefix literals and the bare 32/16
hex-width magic numbers in the log trace-context injection with named
constexpr constants (kTraceIdPrefix, kSpanIdPrefix, kTraceIdHexLen,
kSpanIdHexLen), documenting that the widths are the W3C 16-byte trace_id /
8-byte span_id rendered as lowercase hex. No behavior change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The existing ignore list only covered src/xrpld/telemetry/, but the same
rationale — telemetry code is conditionally compiled behind
XRPL_ENABLE_TELEMETRY, which the coverage build does not enable — applies to
src/libxrpl/telemetry/ (SpanGuard.cpp, Telemetry.cpp, etc.) and the
include/xrpl/telemetry/ headers. Their lines are never executed in coverage
builds, so they were dragging patch coverage down on the telemetry PRs.
Extend the ignore list to match.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a "Protocol Span Flow" section to the telemetry operator runbook: 8
Mermaid diagrams that map every OTel span onto the real xrpld control flow
and XRPL protocol order (verified against code and docs/consensus.md), for
use as the canonical key when linking the span hierarchy.
- Master overview, client/peer ingress, shared apply pipeline, consensus
round, accept/build/finalize, and pathfinding/ledger-acquire side flows.
- Every node/branch is labelled with the span that represents that state or
transition (or explicit "(no span)"); drops/abandons are marked terminal.
- Shows real loops, retries, recovery, and drop branches: multi-round
consensus settling (avalanche threshold rounds + MovedOn/Expired retry with
wrong-ledger recovery), 3-pass tx apply retry, TxQ cross-ledger retry,
quorum-gated async validation with abandoned-ledger, ingress backpressure
drops, and cross-node context propagation.
- Adds a divergence table noting where OTel span parenting does not match the
real protocol flow (deterministic/hash trace-id roots, JtAccept/JtUpdatePf
job hand-offs, sequential peer->consensus receive stages).
Replace the stale planning-era diagram in OpenTelemetryPlan/08-appendix.md
(which named spans that were never built) with a pointer to the runbook.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Carries the review fixes forward to phase-10: no-alloc GetCurrent hot path,
activateIfLive helper deduplicating the four worker-body activation sites,
non-copyable CoroAwareContextStorage, and the clarified drop-counter doc.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Alloy spanmetrics connector (perf-iac nodes run Alloy instead of the
collector-contrib image) carried the same old [.. 1s, 5s] bucket list, whose
buckets are documented as "copied from" the collector config. Sync it to the
15-bucket layout (adds 2s, 3s, 4s, 10s, 30s) so Alloy-fed nodes get the same
accurate second-scale consensus and ledger.acquire percentiles as
collector-fed nodes. unit was already "ms". Kept strictly ascending.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
M1: ~PendingTraceId now increments a process-wide atomic counter when a
pinned deterministic trace_id is destroyed unconsumed, so the silent drop
stays observable in release builds where the existing XRPL_ASSERT is a
no-op. Exposed via unconsumedDeterministicIdDrops() for a future metric.
The deterministic trace_id bytes, the GenerateTraceId() consume logic, and
the unconditional reset() are all unchanged.
M2: tighten the Doxygen on SpanGuard::childSpan(std::string_view) to state
it parents to the current ambient context of this store (meaningful only
when a scope/ScopedSpanGuard/ScopedActivation is active) and to point
callers at childSpan(name, ctx) for explicit cross-store parenting.
Doc-only; no behavior change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Brings coroutine-aware context storage + tx/consensus worker-body activation.
Resolved: Telemetry.cpp keeps both meterProvider_ (phase-7) and contextStorage_
(coro-aware); doc-09 keeps phase-7 structure and applies the pathfind.request →
rpc.command.<name> correction to phase-7's own PathFind section.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Activate the accept span as ambient context at the top of doAccept using the
non-owning ScopedActivation, so doAccept's log lines (and any spans created in
the body) carry the accept span's trace_id. doAccept runs to completion on the
JtAccept worker with no coroutine yield, so the activation is thread-local and
safe; the acceptSpan still owns and ends the span. The accept.apply child span
keeps its explicit childSpan(acceptApply, acceptSpanContext_) parentage --
activation only adds log correlation and does not alter parent linkage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Brings coroutine-aware context storage, scoped rpc.command, coro-store-swap
tests, and the scoped pathfind.request forward.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add two GTests on the SpanGuardScopeTest fixture and install a
CoroAwareContextStorage so the tests exercise coro-aware ambient context:
- scopedGuard_survives_localvalue_store_swap: a ScopedSpanGuard's scope
is hidden when its LocalValue store is swapped out and visible again
when swapped back in, and pops cleanly under the owning store.
- activate_sets_ambient_without_owning: SpanGuard::activate() makes the
span ambient for the activation's lifetime without ending it; the
owning guard ends it exactly once.
Fix the cross-store death test to match the store-identity assertion
message ("constructing context store", not "constructing thread") after
the A3 refactor; the fixture's storage install is what lets the assert
fire at all.
Revert RipplePathFind's pathfind.request from freshRoot SpanGuard back to
ScopedSpanGuard: the coro-aware storage moves the scope with the
coroutine across yield, so it nests under rpc.command and stays
trace-correlated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
rpc.command.* becomes a scoped child so it nests under rpc.process, correlates
its log lines, and parents pathfind.request. No RPC::Context plumbing needed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Brings coroutine-aware OTel context storage forward: CoroAwareContextStorage,
its install in Telemetry, ScopedSpanGuard same-store assertion, and the
non-owning ScopedActivation helper.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add SpanGuard::activate() returning a ScopedActivation RAII helper that
activates an already-owned span (from a thread-free SpanGuard) as the
current context WITHOUT taking ownership. The activation pushes the span
onto the current LocalValue context store on construction and pops it on
destruction; it never ends the span (its owning SpanGuard does).
This lets a job-handoff span be made ambient for the duration of a
synchronous, non-yielding worker body so log lines there carry the
span's trace_id. Non-copyable and non-movable, mirroring ScopedSpanGuard.
owner is captured after the Scope push via declaration-order member
initialization (scope declared before owner), matching the A3
capture-after-materialization invariant. A #else no-op stub keeps the
API zero-overhead when telemetry is compiled out.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Coroutine-aware storage lets a scope resume on another worker within the same
coroutine store; store-identity is the correct pop-safety invariant. Same-store
equals same-thread for synchronous code, so no safety is lost.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On Windows the telemetry unit test src/tests/libxrpl/telemetry/MetricMacros.cpp
fails to compile with boost/asio/detail/socket_types.hpp C1189
"WinSock.h has already been included". Root cause traced with clang -E -H:
MetricMacros.h pulls MetricsRegistry.h -> OTel spin_lock_mutex.h, which
defines _WINSOCKAPI_ and includes <windows.h> (WinSock 1); it then pulls
ServiceRegistry.h -> <boost/asio.hpp>, whose socket_types.hpp requires
winsock2.h first. Production consumers (RCLConsensus.cpp, PerfLogImp.cpp)
happen to include Boost.Asio via other xrpld headers first, so only the test
TU -- which includes MetricMacros.h first -- hit it.
Pre-include boost/asio/detail/socket_types.hpp under _MSC_VER at the top of
the header so winsock2.h wins for every consumer. Mirrors the existing guard
in MetricsRegistry.cpp.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Install CoroAwareContextStorage as the process-wide OTel runtime context
storage in TelemetryImpl::start(), before SetTracerProvider and before any
span is created, so the ambient context follows JobQueue coroutines across
yield/resume. Hold it in a member for the process lifetime; not reset in
stop() because resetting while spans may exist is SDK undefined behaviour.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
misc-include-cleaner does not expand macros, so it flags MetricsRegistry.h,
ServiceRegistry.h, <functional> and <mutex> as unused even though every
XRPL_METRIC_* macro body uses them ((app).getMetricsRegistry(),
std::once_flag/std::call_once, std::function). Add IWYU pragma: keep, the
same suppression already used across xrpld/ for macro/transitive includes.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Backs the OTel active-context stack with xrpl::LocalValue so the ambient
context follows a JobQueue::Coro across yield/resume. Not yet installed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Consensus.h includes ConsensusSpanNames.h, which already includes
SpanNames.h (it is built on StaticStr/join()). Consensus.h uses no base
SpanNames.h symbols directly, so misc-include-cleaner flags the direct
include as unused. Remove it; SpanNames.h stays reachable transitively.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Rename file-local thread_local globals to the .clang-tidy convention
(GlobalVariablePrefix "g" + CamelCase), keeping the Tls marker:
tlsPendingTraceId -> gTlsPendingTraceId, tlsPendingConsumed -> gTlsPendingConsumed.
- Add direct includes for opentelemetry trace_id.h / span_id.h (header uses
TraceId/SpanId in signatures) and sdk/trace/id_generator.h (.cpp references
IdGenerator directly) to satisfy misc-include-cleaner.
Both verified clean with clang-tidy against a telemetry-enabled compile DB.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The check-rename CI gate requires all prose/comment references to use the
new naming. Fix the four remaining occurrences ("Ripple epoch",
"rippled's doAccept") introduced by this branch's dashboard/glossary
commits. Comment- and doc-only; no behavior change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The OTel naming check (Rule F) scans @code doc-comment examples and fails
on string-literal span names; Rule H warns on undefined SpanNames
constants. Replace the literal "subtask" and the undefined
rpc_span::op::dispatch with the defined rpc_span::op::process constant so
the examples model correct API usage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Brings in the telemetry/metrics init-order fix (start before [rpc_startup]).
Conflict in Application.cpp: phase10 had reordered start() to run
telemetry+metrics first; phase9 relocated that block into a new
startTelemetry() called from setup() before the [rpc_startup] loop.
Resolved by taking the fix — the block lives only in startTelemetry()
(called once from setup(), after overlay_ is built), removed from start().
telemetry_->start() and metricsRegistry_->start() each appear exactly once.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Metric-emitting code that first runs during ApplicationImp::setup() (the
[rpc_startup] loop) executed while the OTel meter was still empty, because
telemetry_ and metricsRegistry_ were started later in start(). A call-site
metric macro caches its instrument on first use via std::call_once, so that
first pre-meter use latched a null instrument for the process lifetime and the
metric silently never recorded (observed: rpc_in_flight_requests, emitted from
PerfLogImp::rpcStart, never appeared).
Extract telemetry_->start() and the metricsRegistry_->start() block into a new
private ApplicationImp::startTelemetry() and call it from setup() just before
the [rpc_startup] loop, so the meter is live before any metric-emitting code
runs. The call sits after overlay_ (and the other subsystems the observable-
gauge callbacks read) is constructed, since starting the metrics reader thread
earlier would let its callbacks call getOverlay() before overlay_ exists.
perfLog_->start() stays in start() — the macro call sites do not depend on it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a 'Computation boundary' section to every keyworded panel stating (a)
whether the result is per-node or aggregated across nodes, derived from
the query's grouping, and (b) where the value is computed along the
pipeline: xrpld code (MetricsRegistry OTel SDK, or beast::insight native),
the collector's SpanMetrics connector, or a Tempo TraceQL query over raw
spans. Classification is derived per panel from its actual queries and the
MetricsRegistry source, not guessed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
P95 of second-scale spans was a meaningless interpolation. The spanmetrics
histogram topped out at [.. 1s, 5s], so consensus.round (~3.9s) and
consensus.establish (~1.9s) all fell into one 1s-5s bucket and
histogram_quantile interpolated linearly across that 4s-wide gap — the
"Build vs Close" / "Ledger Close Duration" panels' P95 read ~4800ms purely
as an artifact (verified: sum/count avg = 3824ms). ledger.acquire was worse:
~17% of samples exceeded the 5s ceiling, so its p95/p99 were unmeasurable.
Add 2s, 3s, 4s (resolve the 1-5s pile-up) and 10s, 30s (give the
ledger.acquire catch-up tail a measurable home). All ten existing boundaries
are preserved and the list stays strictly ascending (the connector
binary-searches buckets and silently misbuckets otherwise). Pin unit=ms so a
future collector default-unit flip can't rename the metric to _seconds.
Buckets chosen from the live mainnet distribution, not guessed. Native
beast::insight histograms (ms-scale RPC/IO timers in Telemetry.cpp) are 100%
under 5s, so they keep the original buckets — this is collector-only.
Applies on collector restart (cumulative series reset once, handled by
rate()). Runbook and regression-threshold bucket notes updated to match.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a scope tag to every panel keyword — (per node), (network-wide),
(network event), or (cluster-wide) — so a reader can tell whether a term
describes this server's own state, a protocol-shared fact, or a
network-wide consensus process the node participates in. Add a matching
'Scope:' line to each glossary entry.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a Keywords definition-list and a References link line to 178 Grafana
panel descriptions across all 14 dashboards, defining the XRPL/rippled
domain terms each panel uses (tiers 1-9: ledger, consensus, transaction
pipeline, fees/queue, node state, peer/overlay, storage, validator,
RPC/pathfinding). Cross-cutting chart terms and job-queue internals are
intentionally excluded.
Add docs/telemetry-glossary.md (86 terms, 9 categories) as the deeper
reference, linked from each panel's References line and from the runbook.
Keywords are injected only where a term appears in that panel's prose
(matched over description text, not source-file citations).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The roundSpan_ / roundSpanContext_ comments still described the old
'detached, Scope stripped' model. roundSpan_ is now a thread-free SpanGuard
(no Scope); children link via the captured roundSpanContext_.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
acquireSpan_ is emplaced on the acquiring thread and reset() on a JtLedgerData
worker. SpanGuard is now thread-free (owns no thread-local Scope), so it can be
created here and destroyed on the worker with no scope to strip — dropping the
.detached() call the old scoped SpanGuard required.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reflects the C1 fix: rpc.command.* stays unscoped (its dispatch wraps
doRipplePathFind which yields), so pathfind.request nests under rpc.process.
The request -> compute -> discover sub-tree nests correctly via ScopedSpanGuard.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ledger.build was left as unscoped SpanGuard after the type split, so tx.apply
(created synchronously during applyTxs on the same JtAccept worker) no longer
nested under it. buildLedgerImpl runs synchronously with no yield, so
ScopedSpanGuard is safe and restores the ledger.build -> tx.apply edge.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
pathfind.request (PathFind.cpp) and pathfind.compute (PathRequest.cpp doUpdate)
were left as unscoped SpanGuard after the type split, so pathfind.compute and
pathfind.discover no longer nested under them. Both handlers run synchronously
with no coroutine yield, so ScopedSpanGuard is safe and restores the
request -> compute -> discover sub-tree. rpc.command.* stays unscoped (callMethod
wraps doRipplePathFind which holds across a yield), so pathfind.request parents
to rpc.process.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The "Ledger Close Duration" panel measured the consensus.ledger_close span
duration, which only wraps the sub-millisecond onClose() prologue — not the
ledger close. Repoint it to the consensus.round span (full round, open to
accept) so it reflects actual close time (~3-5s on mainnet). The consensus_mode
filter is preserved (consensus.round carries that attribute, verified live).
The sibling rate panels (Consensus Mode Over Time, Accept vs Close Rate,
Validation vs Close Rate) keep using consensus.ledger_close: a rate of that
span is a valid per-close event counter, only its duration was wrong.
Runbook updated to match.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The "ledger close time" was mis-derived and partly un-queryable:
- Build vs Close Duration derived close time from the consensus.ledger_close
span, which only wraps the onClose() prologue (~0.8ms live) — not the close.
Repoint the close series to consensus.round (full round, live P95 ~4.8s).
- The network close-time value (close_time) lived only as a span attribute,
un-queryable in Prometheus and unfit as a spanmetrics label (monotonic
timestamp -> unbounded cardinality). Expose it as last_close_time on the
existing server_info observable gauge (native gauge, no new instrument).
- Add a "Ledger Close Interval & Age" panel to ledger-operations and
node-health: interval = 1/rate(ledgers_closed_total) (counter-based,
scrape-independent); age = time() - (last_close_time + epoch_offset).
A gauge delta is deliberately NOT used for the interval — a timestamp gauge's
delta aliases to the scrape period, not the close cadence (verified live).
Guardrail comments in both collector configs record why close_time must never
become a spanmetrics dimension. Docs (09-reference) and the operator runbook
updated to match.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Document that deterministic-trace_id spans (tx.* apply pipeline, tx.process,
tx.receive, consensus.round) are now genuine trace roots with empty
parent_span_id via the custom DeterministicIdGenerator, superseding the old
synthetic-parent behavior that showed 'root span not yet received' in Tempo.
Also update the fresh-root note: peer entry spans use ScopedSpanGuard::freshRoot().
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
These peer-level entry spans (introduced on phase-6) used the removed
SpanGuard::rootSpan(). They are scoped on the peer thread and end there, so
use ScopedSpanGuard::freshRoot() — a scoped fresh trace root that does not
inherit any span left active on the peer thread.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
With unscoped SpanGuard + spanContext() (own-span capture), the consensus
round/establish/accept sites just capture and drop the detach dance;
openSpan + peer proposal/validation receive are thread-free handoffs (no
detach); rootSpan->freshRoot.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
OTel distinguishes a span's own context from the thread's
current/ambient context; SpanGuard is unscoped so spanContext()
(own span, default) is what cross-thread childSpan parenting needs,
and threadLocalContext() (static) snapshots RuntimeContext::GetCurrent()
for propagation. Renames captureContext.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
SpanGuard is unscoped (never pushed onto the thread-local context stack), so
RuntimeContext::GetCurrent() returned an unrelated ambient span. Build the
captured context from impl_->span so captureContext() is correct for every
caller regardless of scoping — childSpan(name, ctx) then parents explicitly to
this span across threads. Matches getTraceBytes()'s own-context behavior.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The generator's doc said it was 'dormant, caller lands on a later branch'.
Reworded to branch-agnostic language: hashSpan() is the primary caller that
forces the root branch to mint deterministic trace roots.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
SpanGuard is now thread-free (holds no Scope), so the tx.receive (PeerImp)
and tx.process (NetworkOPs) handoff sites no longer need .detached() before
being stored and ended on a worker thread — just construct the guard. The
stale "Scope leak" comments are replaced accordingly.
Make the six txq.* spans ScopedSpanGuard so their sub-spans nest via the
ambient context: txq.apply_direct/batch_clear under txq.enqueue, and
txq.accept_tx under txq.accept. All six are verified synchronous, ended at
scope, and never moved/handed off, so scoping is safe. applyDirect's span
is pure RAII (no method calls), so it is declared const.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The single-arg hashSpan fabricated a synthetic DefaultSpan parent (random
span_id) to carry the deterministic trace_id. That left every standalone
hash span with a phantom parent_span_id pointing at a span that is never
exported, so Tempo reported "root span not yet received".
Inject the deterministic trace_id through the DeterministicIdGenerator
instead: start the span on the SDK root branch (Context{kIsRootSpanKey,true})
with a PendingTraceId pinning hashData[0:16]. The SDK's no-valid-parent
branch calls GenerateTraceId(), which returns the pinned id, yielding a
TRUE root (empty parent_span_id) whose trace_id == hashData[0:16].
The deterministic bytes are unchanged (still raw hashData[0:16]). The
cross-node overload (real remote parent, remote=true) is untouched and must
keep producing a child of the sender's span. Drop the now-unused
<xrpl/basics/random.h> include; add <array> and DeterministicIdGenerator.h.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rewrite SpanGuardScope.cpp for the unscoped SpanGuard / scoped
ScopedSpanGuard split and the DeterministicIdGenerator: install the
generator in the TestTelemetry mock (4-arg TracerProviderFactory), drop the
obsolete detached()/detachInPlace() tests, and add freshRoot, scoped-ambient,
scope-handoff, ends-once, deterministic-root, and cross-thread death tests
with exact assertions.
Fix the last detached() caller: RipplePathFind now uses SpanGuard::freshRoot
(SpanGuard is thread-free, so it is held across the coroutine yield and ended
on resume with no scope to strip).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add MetricMacros.h exposing 13 call-site metric macros covering all 7 OTel
instrument kinds (Counter INC/ADD, UpDownCounter, Histogram, sync Gauge, and
Observable Counter/UpDownCounter/Gauge), each with a no-op branch when
XRPL_ENABLE_TELEMETRY is undefined. Sync macros lazily create-once via
std::call_once and cache the instrument at the call site; observable macros
register a callback eagerly. Sync Gauge is ABI-v2-gated with an actionable
static_assert under this repo's ABI v1 build.
Add GTest coverage (8 tests) driving the macros against a bare SDK
MeterProvider via a duck-typed FakeApp/FakeMetricsRegistry, asserting exact
create-once and disabled-gating behavior. No CMakeLists change needed: the
telemetry test-module glob already discovers the new source.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Entry points (http_request/ws_message/ws_upgrade) were inheriting a leaked
ambient span on reused coro workers; now scoped fresh roots via
ScopedSpanGuard::freshRoot. rpc.process stays a scoped child.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a custom OTel IdGenerator that returns a thread-local pending trace_id on
the SDK no-parent (root) branch and a random one otherwise, plus a PendingTraceId
RAII guard that pins that id for the next forced-root span and asserts on
destruction that it was consumed. Wire the generator into
TracerProviderFactory::Create via its 4-arg overload.
This lets hash-derived spans become true trace roots so they line up into one
trace across nodes. It is installed but dormant on this branch: the caller
(hashSpan) arrives on a later branch (phase-3). GenerateSpanId is always random
and is_random_ is false so the W3C random-trace-id flag is not set on
deterministic ids.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Separate the two responsibilities the old SpanGuard fused: span ownership
(thread-free) and scope activation (thread-bound TLS push).
- SpanGuard now owns ONLY the span. Its Impl drops the optional<Scope>,
the owner thread-id, the Detached tag and the scope-less ctor; ~Impl is
just `if (span) span->End()`. The guard never binds a thread-local
context stack, so it may be moved to and destroyed on any thread.
- ScopedSpanGuard is a new pimpl type that wraps a SpanGuard plus an
optional<Scope>. Member order (guard first, scope second) pops the
scope before the span ends. It is non-copyable and non-movable;
factories return unnamed temporaries so guaranteed copy elision covers
`auto s = ScopedSpanGuard::freshRoot(...)`.
- `operator SpanGuard() &&` replaces detached()/detachInPlace(): it pops
the scope eagerly on the origin thread and yields a thread-free
SpanGuard for handoff to a job or another thread. detached(),
detachInPlace() (both overloads) and the Detached apparatus are deleted.
- Move-assignment is re-enabled on SpanGuard (no Scope to re-bind), in
both the real class and the no-op stub.
- rootSpan() renamed to freshRoot() in both classes (behavior unchanged:
still forces kIsRootSpanKey).
- hashSpan behavior is intentionally unchanged (only builds the now
unscoped Impl); the true-root IdGenerator fix is a separate later task.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add two free functions in xrpl::telemetry that wrap the required
"detach a live SpanGuard held in a container" idiom so call sites write
one line instead of the make_shared/emplace rebuild by hand:
void detachInPlace(std::optional<SpanGuard>&);
std::shared_ptr<SpanGuard> detachInPlace(std::shared_ptr<SpanGuard>);
Both are no-ops on an empty/null/inactive guard. The #else branch adds
matching inline no-op stubs so callers compile with telemetry disabled.
Pure API addition, no behavior change; consumer call sites are rewritten
on phase4 in a follow-up.
The unit tests for these helpers land on phase2, where the telemetry
test module (in-memory-exporter harness + SpanGuardScope.cpp) exists —
mirroring how the detached()/rootSpan() tests were placed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
An earlier Doxygen-style normalization pass on this file introduced a
stray xrpld_ prefix into 5 comment references to the validation
counters/gauge. The actual instrument names registered in
MetricsRegistry.cpp have no such prefix (validation_agreements_total,
validation_missed_total, validation_agreement). Comment-only fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
xrpl_branch and xrpl_node_role are stamped by the perf-iac harness --
infrastructure outside this repo's OTel code -- so they have no L1
(*SpanNames.h), L2 (collector config), or L6 (MetricsRegistry.cpp) source
Rule D can derive them from. Unlike the generic builtins set (Prometheus/
Grafana mechanics every OTel setup has), these are repo-specific and
narrow, so they get their own EXTERNAL_INFRA_LABELS constant: a visible,
documented, deliberately narrow exception to the 'no hardcoded allowlist'
design principle, not a silent workaround. Add test coverage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
xrpl_branch and xrpl_node_role are collector/infra-injected resource
labels used by perf-iac dashboards to identify the build under test and
its role in the perf cluster. They have no L1 span-attribute source or
L6 native-metric-label source (like the existing job/instance/job_type
builtins), so Rule D flagged them across every perf-iac dashboard.
Register them in the builtins set, matching the existing pattern for
infra-level dashboard labels. Add test coverage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Convert leading /// blocks in LedgerSpanNames.h and PeerSpanNames.h to
house-style /** */. Comment-only: code is byte-identical.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Convert leading /// blocks to house-style /** */ across the telemetry
files carried on this branch (using the updated fix_doxy.py). Also
regenerate levelization results. Comment-only: code is byte-identical.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Convert leading /// blocks in these 9 phase3-originating files to house
style /** */ (using the updated fix_doxy.py that now handles /// ->
block conversion). Comment-only: code is byte-identical.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bring the 5 telemetry files introduced on this branch to the enforced
house Doxygen style (/** alone, ' * ' continuation prefix, no single-line
blocks) so the check-doxygen-style hook passes under CI's --all-files run.
Comment-only: code is byte-identical after comment stripping (verified).
Also regenerate levelization results: the committed ordering.txt carried
stale 'xrpl.telemetry > xrpld.consensus/rpc' edges that the current include
graph no longer produces.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
getMeter() became a pure virtual on THIS branch (native-metrics), which
makes the TestTelemetry mock abstract and would break a telemetry=ON
build of phase7/phase8 in isolation. The override was previously only on
phase9 (where it was mis-attributed); relocate it to phase7 where the
pure virtual is introduced, so every branch from here forward builds.
Mirrors NullTelemetry: an inert meter from a process-wide noop provider.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The opentelemetry-cpp::exporter_in_memory component declares no libs (the
Conan recipe ships libopentelemetry_exporter_in_memory.a but does not
associate it with the component), so linking that target put nothing on
the link line and xrpl_tests still failed with an undefined reference to
InMemorySpanExporterFactory::Create. Locate the archive with find_library
(HINTS the OTel package lib dir) and link it by path, ahead of the umbrella
so its undefined SDK references resolve. Verified: the archive now appears
on the generated link line and xrpl_tests links + the SpanGuardScope tests
pass. Supersedes the no-op component-target link in the prior commit.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The umbrella opentelemetry-cpp target's link set omits the in-memory
exporter library, so xrpl_tests failed to link with an undefined
reference to InMemorySpanExporterFactory::Create. The SpanGuardScope
tests use an InMemorySpanExporter to read exported spans. Link the
opentelemetry-cpp::exporter_in_memory component explicitly. Test-only:
production code never uses this exporter.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Fees::accountReserve gained a second accountCount parameter (merged from
develop). MetricsRegistry called the old one-arg form accountReserve(0),
which no longer compiles. Use accountReserve(0, 1) -- one account, zero
owned objects -- which returns the base reserve, matching the metric's
intent and the value the one-arg form previously returned.
Also normalize three pre-existing Doxygen blocks in this file to house
style (opener/continuation prefixes) so the check-doxygen-style hook passes.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
getMeter() became a pure virtual on this branch (metric gap-fill), which
made the TestTelemetry mock abstract and broke the scope-leak test build.
Mirror NullTelemetry: serve an inert meter from a process-wide noop
provider. The scope tests exercise only tracing, so the meter is unused.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Post-review fixes for the SpanGuard scope-leak change:
- SpanGuard.h: restore the injectCurrentContextToProtobuf real-class
declaration, its #else no-op stub, and the protocol::TraceContext
forward-declaration dropped during the 3->4 union merge (compile break).
- Re-parent consensus.phase.open and consensus.establish under the round
span via a new RCLConsensus::Adaptor::roundSpanContext() accessor: the
round span is now detached, so ambient parenting no longer works; the
phase spans link explicitly to the captured round context. csf::Peer
gains a matching no-op accessor so the generic engine still compiles.
- Switch five childSpan(op::X, ctx) sites to the full consensus::span::X
constants: childSpan(name, ctx) takes the name verbatim, so the suffix-
only op:: constants emitted short, non-dotted span names.
- TestTelemetry mock: add getConsensusTraceStrategy() override (base pure
virtual introduced on this branch) so the mock is not abstract.
- validate(): rewrite the trace_context comment to drop attack-surface
framing and the PR-discussion reference.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
roundSpan_, establishSpan_, openSpan_ and the accept span are emplaced on
one job worker and reset()/moved to another, leaking their thread-local OTel
Scope. Each is now detached AFTER its context is captured. Spans that
previously nested under them via the ambient thread context are re-pointed to
the captured SpanContext explicitly:
- round children (proposal.send, ledger_close, mode_change) -> roundSpanContext_
- establish children (update_positions, check) -> establishSpanContext_
- accept.apply -> acceptSpanContext_
onForceAccept (synchronous) keeps the accept span's behaviour; only onAccept
(moved into the JtAccept job) needs the detach. Member move-assignment is
deleted, so re-assignment uses emplace().
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The consensus.proposal.receive / consensus.validation.receive guards are
moved into checkPropose/checkValidation jobs, so detach their Scope on the
peer thread before the job hand-off. The standalone (no-propagated-context)
fallbacks in ConsensusReceiveTracing now use rootSpan() so they start a
fresh trace root instead of inheriting an ambient span; the hashSpan
propagated-context branches are unchanged (cross-node linking preserved).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
doRipplePathFind holds its pathfind.request span across context.coro->yield();
the coroutine resumes on a different JobQueue worker. rootSpan() gives it a
fresh trace root (no stale ambient parent), and detached() strips the
thread-local Scope so the guard neither lingers on the worker's context stack
across the yield nor pops the wrong stack on resume.
Adds SpanGuardScope.cpp: an in-memory-exporter test for rootSpan()/detached()
including the rootSpan().detached() combination used here.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
acquireSpan_ is emplaced on the acquiring caller thread but reset() in
done() on a JtLedgerData worker. Detaching strips the thread-local OTel
Scope on the origin thread so destruction on the worker does not corrupt
the origin thread's context stack.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
These inbound peer-message entry points (kConsumer) used span(), which
inherits whatever span is active on the peer thread — including a leaked
tx.receive scope — so validations/proposals were wrongly nested under
unrelated transaction traces. rootSpan() starts a fresh trace root.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Both spans are moved into job-queue lambdas and destroyed on a worker
thread. Detaching on the origin thread pops the thread-local OTel Scope
there, so later spans on the peer/RPC thread no longer inherit these as a
leaked ambient parent. Trace_id/parent are unchanged (both are hashSpan).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
rootSpan has the same (cat, prefix, name) signature as span(), so its
prefix/name arguments must be *SpanNames.h constants under Rule F/H. Add
rootSpan to the CALLSITE regex and CONSTANT_ARG_POSITIONS ({1,2}), update
the enumerating docstrings/README, and add two regression tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The otel-naming Rule F check scans @code doc examples as call sites; raw
string literals there trip the rule. Reference *SpanNames.h constants /
placeholder identifiers in the examples so comments follow the naming rules.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
detached() strips the thread-local OTel Scope so a guard can be safely
moved to and destroyed on another thread; it pops the Scope on the origin
thread and moves the span into a scope-less guard. rootSpan() starts a span
as a fresh trace root (kIsRootSpanKey) so inbound entry points never inherit
an ambient span left active on the thread.
Impl now holds an optional<Scope> (nullopt for detached guards). Updated the
SpanGuard class docs and docs/build/telemetry.md with the cross-thread rules.
The unit test lands on phase2 where the telemetry test module exists.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sync the Node Health dashboard with panels and field settings refined in
Grafana, and document the ledger-convergence panels:
- Add descriptions for "Validated Ledger Seq — Current (Stat)" (per-node lag
behind the network tip) and "Validated Ledger Seq — Convergence (Max − Min)"
(seq spread across nodes), in the standard 7-field panel format.
- Peer Count: threshold colours (red < 5, yellow < 10, green >= 10) so a low
peer count reads red rather than using a value gradient.
- Minor field-config fixes on Last Close and NuDB Storage panels (axis label,
threshold cleanup) as adjusted in Grafana.
Add the Grafana duration unit "dthms" to the cspell dictionary. Panel content
stays at parity with the Grafana Cloud copy.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Regroup all 47 panels under correct subsystem rows and fix the broken row
structure. Previously the "NodeStore I/O" row had no gridPos and collided at
y=32 with an appended "Extended Metrics" catch-all row, so NodeStore I/O
rendered empty and its panels landed under the wrong header.
- Rows: Overview (top), Operating Mode, NodeStore I/O, Jobs, Caches, Server
Info, Complete Ledgers & DB, Ledger Economy. Drop the stray "Extended
Metrics" row; each panel now sits under its subsystem.
- Normalize panel heights to three buckets: 12 (standard), 16 (emphasis /
detail plots), 24 (Operating Mode state timeline). Widths stay half (12) or
full (24).
- Contiguous, non-overlapping gridPos; all 47 panels preserved.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rework the legend identity bracket across all 14 dashboards so it reads
"<series> [instance, branch, work_item]" (comma-separated, Node Role dropped)
and collapses cleanly when perf-iac labels are absent -- "<series> [instance]"
with no empty "[, ]" gaps.
- Build one xrpl_ident label per query via label_join + a single label_replace
that trims leading/trailing empty segments and brackets the result. Done at
the outermost (post-aggregation) level so it survives each panel's by() clause.
- Add service_instance_id to the by() of 6 panels that grouped only by node_role,
so same-branch validator/peer series stay distinguishable after dropping the
node_role legend token.
- Fold the two legendFormat two-bracket panels (Transaction Results by Type,
Tx Apply Pipeline Latency by Type and Stage) into the same single-bracket form.
- Rename the annotation query label "Perf runs (perf-iac)" to
"Annotate perf-iac runs".
- Revert dashboard-level descriptions to plain text (the folder-list info tooltip
escapes HTML, so <br>/markdown render literally).
Verified: output byte-identical to the prior multi-call cleanup over a 2-day
live perf window; all 14 files valid JSON.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Group the node + perf-iac identity labels inside one pair of brackets in the
legend Display name so a series reads e.g. "Queue Depth [validator-0
test:pr:abc123 validator RIPD-7455]" on a perf node and "Queue Depth
[aws-dev-xrpl-1]" on a plain node. Absent perf-iac labels collapse to a single
space before the closing bracket; no empty "[]", no data impact. Covers the
eleven dashboards this branch owns.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Group the node + perf-iac identity labels inside one pair of brackets in the
legend Display name:
${__field.labels.series} [${__field.labels.service_instance_id} ${__field.labels.xrpl_branch} ${__field.labels.xrpl_node_role} ${__field.labels.xrpl_work_item}]
so a series reads e.g. "Queue Depth [validator-0 test:pr:abc123 validator
RIPD-7455]" on a perf node and "Queue Depth [aws-dev-xrpl-1]" on a plain node.
Absent perf-iac labels render empty and collapse to a single space before the
closing bracket (HTML/SVG whitespace collapse) — the only cosmetic artifact on
non-perf nodes; no empty "[]" and no data impact. Verified on a live local
Grafana render.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The legend now uses raw labels in the dashboard Display name (see the paired
dashboard commit), so the collector no longer needs to build resource_group /
iac_group. Those derived labels also could not survive the dashboards' sum by()
aggregation, which is why the grouped approach was dropped.
Revert transform/legendgroups and its transform/cloudlabels datapoint copies
from the Grafana Cloud collector config and the in-repo Alloy config; both files
are now identical to their pre-change state.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The collector-built resource_group / iac_group labels did not survive the
dashboards' own PromQL aggregation: panels that sum by(service_instance_id,
xrpl_branch, xrpl_node_role, ...) drop any label not in the by() clause, so the
derived group labels were aggregated away and the Display name rendered empty.
Switch the legend Display name to the raw labels the panels already group by,
which are always present by construction:
${__field.labels.series} ${__field.labels.service_instance_id}
${__field.labels.xrpl_branch} ${__field.labels.xrpl_node_role}
${__field.labels.xrpl_work_item}
Absent labels render empty (no brackets, no gaps). Covers the eleven dashboards
this branch owns. The matching collector-config revert lands on the branches
that introduced those configs (phase-7 local, phase-10 cloud/alloy).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The collector-built resource_group / iac_group labels did not survive the
dashboards' own PromQL aggregation: panels that sum by(service_instance_id,
xrpl_branch, xrpl_node_role, ...) drop any label not named in the by() clause,
so the derived group labels were aggregated away and the Display name rendered
empty ("Observing" with no node). Building them in the collector also coupled
every dashboard to a lockstep collector redeploy on every node.
Switch the legend Display name to reference the raw labels the panels already
group by, which are always present by construction and need no collector
support:
${__field.labels.series} ${__field.labels.service_instance_id}
${__field.labels.xrpl_branch} ${__field.labels.xrpl_node_role}
${__field.labels.xrpl_work_item}
${__field.labels.X} renders empty for a label a series lacks (verified in
Grafana), so dev/mainnet nodes show just "<name> <node>" and perf nodes add the
branch/role/work-item, with no empty brackets and no double spaces. Verified
live across timeseries, stat, bargauge, piechart, gauge and cross-node
aggregation panels.
Revert the transform/legendgroups processor from the local collector config;
this file is now identical to its pre-change state.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Update the config.alloy metric_statements comment to describe the current
bracket-baked, ", "-joined group-label values ("[node, mainnet]"), matching the
implementation. Comment-only; no behavior change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mirror the local collector change on the Grafana Cloud collector config and the
in-repo Alloy config: build the group-label VALUES with brackets and comma
separators (resource_group="[node, mainnet]",
iac_group="[branch, role, work-item]") in both trace and metric resource
statements, so span-derived and native metrics carry the same bracketed values.
This lets the dashboards use a bracket-free Display name and avoids the empty
"[]" that a templated bracket would leave when a group is absent (Grafana's
renameByRegex runs pre-interpolation and cannot strip it). Separator changes
from "-" to ", ".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Display name templated the brackets: "${series} [${resource_group}]
[${iac_group}]". When a group label was absent the "${}" rendered empty,
leaving a literal "[]" in the legend, and the cleanup renameByRegex could not
remove it (Grafana runs that transform on the pre-interpolation template, which
has no "[]" -- it only appears after interpolation at render time).
Bake the brackets and comma separators into the group-label VALUES in the
collector instead (resource_group="[node, mainnet]",
iac_group="[branch, role, work-item]"), so the Display name is bracket-free
("${series} ${resource_group} ${iac_group}") and an absent group renders as
nothing. Separator changes from "-" to ", ". Drop the dead cleanup transform.
Covers this branch's local collector config and its three dashboards
(ledger-data-sync, network-traffic, overlay-traffic-detail).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Display name templated the brackets: "${series} [${resource_group}]
[${iac_group}]". When a group label was absent (e.g. iac_group on a non-perf
node) the "${}" rendered empty, leaving a literal "[]" in the legend. The
empty-bracket cleanup renameByRegex could not remove it: Grafana runs that
transform on the pre-interpolation Display name template, which contains no
"[]" -- the "[]" only appears after variable interpolation, at render time.
Fix it at the source: the collector now bakes the brackets and comma separators
into the group-label VALUES (e.g. resource_group="[node, mainnet]",
iac_group="[branch, role, work-item]"). The Display name becomes bracket-free --
"${series} ${resource_group} ${iac_group}" -- so an absent group is simply an
absent label and renders as nothing. Also switches the intra-group separator
from "-" to ", " (e.g. "[node, mainnet]"). Drop the now-dead cleanup transform
from every panel; value-mapping renameByRegex rules are preserved.
Collector-side label building lives in the paired commits on the collector
config branches. This commit is the dashboard half for the eleven dashboards
this branch owns.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Panels that group by a distinguishing dimension (consensus_mode, tx_type,
command, method, stage, ...) already fold that dimension's value into the
curated series name via label_replace(..., "series", "<prefix> $1", "<dim>",
"(.*)"). The migration also added a separate [${__field.labels.<dim>}] bracket,
so those legends rendered the value twice, e.g. "proposing [proposing] [...]"
on Consensus Mode Over Time.
Remove the dimension bracket from all 29 affected panel Display names, leaving
${__field.labels.series} [resource_group] [iac_group]; the series name already
conveys the dimension. The migration script no longer emits the bracket either.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
In the Alloy config the spanmetrics connector derives span_calls_total /
span_duration_* from the tagged TRACES and sends them straight to the batch
processor, bypassing the metric resource statements. The grouped legend labels
were built only in metric_statements, so span-derived RED metrics carried no
resource_group / iac_group and span-based dashboard panels rendered empty group
legends.
Build the same labels in trace_statements too (mirroring how the tier attrs are
already set in both), so spans carry them before spanmetrics derives from them.
The reference otel collector is unaffected: there spanmetrics is a connector
feeding the metrics pipeline, which runs transform/legendgroups on all inputs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The docker/telemetry/alloy/config.alloy node config feeds Grafana Cloud via the
OTLP gateway (which promotes resource attributes to labels server-side). Add the
resource_group / iac_group build to its metric resource statements so nodes
using this Alloy path get the same grouped legend labels the dashboards expect.
This is a generic node config (no perf-iac attrs), so iac_group's guard never
fires here and only resource_group is built; the iac_group statement is kept for
parity and activates automatically on a perf-comparison node. Guards check != ""
as well as != nil since OTTL Concat emits a stray separator for an empty element.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Eighteen timeseries panels in consensus-health and transaction-overview
carried tooltip maxHeight 500; the dashboard guideline specifies 600. Normalize
them so every panel's tooltip matches Mode:All / Sort:Desc / Max Height:600.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The "Overlay Traffic by Category (Bytes In)" bargauge names each bar by
category via 22 byName field-override displayNames (Transactions, Proposals,
...). The legend migration had also added a generic
${__field.labels.series} [...] Display name, which competed with those
curated names. Remove the generic Display name (and its now-unneeded
empty-bracket cleanup transform) from this panel so the category names stand;
this panel's identity is the traffic category, not the node/perf-iac grouping.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mirror the local-config guard hardening on the Grafana Cloud collector config:
add != "" to the transform/legendgroups presence guards so an empty-string
perf-iac attribute (work_item | default('') on a no-work-item run) does not
produce iac_group="--". Empty-string sources now leave the group unset.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The transform/legendgroups guards checked only != nil. The perf-iac collector
stamps xrpl.branch / xrpl.node.role / xrpl.work.item with an empty-string
default (work_item | default('')) when a run has no work item, and an empty
string is not nil. OTTL Concat renders a stray separator for an empty element,
so an all-empty perf-iac stamp would have produced iac_group="--".
Add != "" to every presence guard (iac_group sources, service.instance.id,
xrpl.network.type) so a group is built only from genuinely-present values and
left unset otherwise. Verified on a live collector: empty-string perf-iac attrs
leave iac_group absent.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add the transform/legendgroups processor (introduced on phase-7 for the local
config) to the dual-export grafanacloud collector config, and copy the two
group labels onto datapoint attributes in transform/cloudlabels.
The local Prometheus scrape promotes resource attributes to labels via
resource_to_telemetry_conversion, but Grafana Cloud's OTLP ingest does not, so
resource_group / iac_group must be copied onto datapoint attributes the same
way service_instance_id / deployment_environment / xrpl_network_type already
are. legendgroups runs before cloudlabels in both metrics pipelines so the
built resource attributes are available to copy; the copies are presence-guarded
so an absent group is not created as an empty datapoint label.
Verified on a live collector (debug exporter): a perf node emits datapoint
attributes resource_group and iac_group; a non-perf node emits resource_group
only, with iac_group correctly absent.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Apply the collector-built grouped-label legend approach (introduced on
phase-7) to the eleven dashboards this branch and its ancestors own:
consensus-health, fee-market, job-queue, ledger-operations, node-health,
peer-network, peer-quality, rpc-pathfinding, rpc-performance,
transaction-overview, validator-health.
Per legend panel: drop the 11-rule renameByRegex chain and legendFormat
"__auto"; set a Standard-option Display name referencing the pre-built
resource_group / iac_group labels plus, when the panel groups by a
distinguishing dimension (e.g. command, tx_type, consensus_mode, stage,
method), that dimension as a leading bracket; keep a single generic
empty-bracket cleanup rule; set tooltip maxHeight to 600.
Value-mapping renameByRegex rules (e.g. consensus close-time
increased->Coarser) are preserved. Only labels a panel groups by appear in
its legend; filter-only selectors ($command with no by(command)) narrow data
but do not become legend groups. All 14 dashboards validate; no legend
renameByRegex chain remains.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The per-panel legend used legendFormat "__auto" reshaped by an 11-rule
renameByRegex chain. __auto sorts labels alphabetically and cannot group,
so the chain existed only to bolt grouping onto its output -- unmaintainable
and non-compliant with the dashboard legend guideline.
Build the grouped labels once at ingestion instead. A new transform/legendgroups
processor derives two resource attributes, promoted to Prometheus labels by
resource_to_telemetry_conversion:
- resource_group: service_instance_id[-xrpl_network_type]
- iac_group: xrpl_branch-xrpl_node_role-xrpl_work_item (perf runs only)
Each is left unset when its sources are absent, so Prometheus drops the empty
label and the legend shows nothing for it. Concat is presence-guarded because
OTTL Concat renders "<nil>" for an absent attribute.
Dashboards now use a Standard-option Display name that references the pre-built
labels (${__field.labels.series} [resource_group] [iac_group]); ${__field.labels.X}
renders empty for a missing label, so absent groups vanish. A single generic
renameByRegex strips the resulting empty "[]" bracket. legendFormat is dropped
(Display name supersedes it). Tooltip maxHeight set to 600 per guideline.
Applied here to the three dashboards this branch introduced (ledger-data-sync,
network-traffic, overlay-traffic-detail); value-mapping renameByRegex rules are
preserved. Verified end-to-end on a live local stack (collector -> Prometheus ->
Grafana): perf and dev nodes render correct, clutter-free legends.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The docker/telemetry/grafanacloud/*.json files are local deploy variants
(they pin the Grafana Cloud datasource UID) and are regenerated from the
canonical grafana/dashboards/*.json. Untrack them and gitignore so they
stay local-only and never drift in version control.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Panel legends only showed the node (service_instance_id); the perf-iac
filter dimensions (xrpl_branch, xrpl_node_role) never appeared, so
baseline-vs-test series were indistinguishable.
Each node-referencing target injects its curated name via
label_replace(...,"series",...) and uses the "__auto" legend, which shows
only labels that vary across the displayed series. A word-boundary-anchored
renameByRegex chain reshapes the auto output into
"Name [node, role]-[branch, work_item]", collapsing empty groups so a
dimension appears only when the filter widens it. Aggregating queries gain
xrpl_branch/xrpl_node_role in their by() so those labels survive; multi-
dimension legends keep their rich label set and append branch/role.
Quantile panels surface the quantile via the panel title (q$quantile),
since {{quantile}} is a dashboard variable, not a series label.
Timeseries panels connect only null gaps shorter than 30m
(spanNulls=1800000): brief scrape gaps bridge, genuine outages stay broken.
Non-node panels (heatmaps, rows, tables, state timelines, category/bucket
series) are left unchanged. Applied to all 14 dashboards.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Panel legends only showed the node (service_instance_id); the perf-iac
filter dimensions (xrpl_branch, xrpl_node_role) never appeared, so
baseline-vs-test series were indistinguishable.
Each node-referencing target injects its curated name via
label_replace(...,"series",...) and uses the "__auto" legend, which shows
only labels that vary across the displayed series. A word-boundary-anchored
renameByRegex chain reshapes the auto output into
"Name [node, role]-[branch, work_item]", collapsing empty groups so a
dimension appears only when the filter widens it. Aggregating queries gain
xrpl_branch/xrpl_node_role in their by() so those labels survive; multi-
dimension legends keep their rich label set and append branch/role.
Quantile panels surface the quantile via the panel title (q$quantile),
since {{quantile}} is a dashboard variable, not a series label.
Timeseries panels connect only null gaps shorter than 30m
(spanNulls=1800000): brief scrape gaps bridge, genuine outages stay broken.
Non-node panels (heatmaps, rows, tables, state timelines, category/bucket
series) are left unchanged. Applied to all 14 dashboards.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Panel legends only showed the node (service_instance_id); the perf-iac
filter dimensions (xrpl_branch, xrpl_node_role) never appeared, so
baseline-vs-test series were indistinguishable.
Each node-referencing target injects its curated name via
label_replace(...,"series",...) and uses the "__auto" legend, which shows
only labels that vary across the displayed series. A word-boundary-anchored
renameByRegex chain reshapes the auto output into
"Name [node, role]-[branch, work_item]", collapsing empty groups so a
dimension appears only when the filter widens it.
Timeseries panels connect only null gaps shorter than 30m
(spanNulls=1800000): brief scrape gaps bridge, genuine outages stay broken.
Dashboards (stable panel set): ledger-data-sync, network-traffic,
overlay-traffic-detail.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Grafana Cloud variants of the 14 xrpld dashboards. Identical to the local
grafana/dashboards/ originals except the DS_PROMETHEUS / DS_TEMPO datasource
template variables have their `current` value pinned to the Grafana Cloud
datasource UIDs (grafanacloud-ripplex-prometheus / -traces).
Includes the ledger-data-sync data-type split and node-health mixed-unit
split + Cache Hit Ratio fix merged forward from phases 7 and 9.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The Cache Hit Ratio panel divided two nodestore_state series that carry
different `metric` label values (node_reads_hit / node_reads_total). Without
ignoring(metric) the two vectors share no common label set, so the division
matched nothing and the panel showed "No data".
Add `ignoring(metric)` to the division. Verified live against mainnet: now
returns a valid ratio (~0.11-0.23). Pre-existing bug, surfaced when the panel
was split out from NodeStore Read/Write Totals.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Three panels mixed incompatible units on one Y-axis (count vs ms, ops vs
percent), causing the small-magnitude series to flatline to zero.
- Last Close Info -> Last Close — Proposers (short) + Converge Time (ms)
- NodeStore Read/Write Totals -> Throughput (ops) + Cache Hit Ratio (%)
- NodeStore Read Threads & Duration -> Read Threads (short) + Busy Ratio (%)
Each new panel has a single unit; per-series unit overrides removed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Each family panel (7-8 series) was overloaded. Split by Ledger/Transaction/
Account State so each panel shows only related series. Get+Share stay
together within a type.
- Ledger Data Exchange: 8 targets -> 3 panels
- Ledger Share/Get Traffic: 8 targets -> 3 panels
- GetObject Traffic by Type: 8 targets -> 3 panels
- GetObject Messages by Type: 7 targets -> 3 + specials panel
- GetObject Aggregate & Special: renamed, kept as one
- Overlay Traffic Heatmap: unchanged
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Span-derived metrics (span_calls_total, span_duration_milliseconds_*) were
emitted at the spanmetrics connector default of 60s. Grafana Cloud stores one
datapoint per flush, but dashboard panels use rate(...[$__rate_interval]),
whose window at short ranges (<=1h) falls below 60s — leaving <2 samples per
window, so those panels rendered "No data" on the Cloud datasource. (Local
Prometheus scrapes at 15s, so local dashboards were unaffected.)
Set metrics_flush_interval: 15s so every rate window holds >=2 samples at any
dashboard range. Trade-off: ~4x span-metric datapoints pushed to Cloud. Native
insight metrics (ledgermaster_*, jobq_*, peer_quality) already push faster and
were never affected.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Brings dashboard panel units, validator-health full-width layout, and filter
description polish into phase 10. The validator-health conflict was resolved
by keeping phase 10's per-target datasource uids and re-applying the unit,
layout, and description changes on top.
Document the three perf-iac filter template variables (xrpl_work_item,
xrpl_branch, xrpl_node_role) in the telemetry runbook and the data-collection
reference: what they filter, their example values, and that perf-iac stamps
them as resource attributes from its own alloy pipeline (absent outside
perf comparison runs, so the filters default to All). Also record perf as a
network value.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Set display units per panel so values render with correct scaling:
- counts/levels/depths → short
- per-minute rates (rate()*60) → cpm; genuine per-second → cps
- per-hour rate (rate()*3600) → suffix:/hr
- fee amounts → si:drops; UNL expiry → d (days)
Value-mapped and identity-number panels (server state, ledger seq, build
version) are left unitless so they are not abbreviated.
Lay out every validator-health panel at full width for readability.
Polish the shared filter descriptions (deployment tier, network values +
perf) across all eleven dashboards.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Set display units on native-metric dashboard panels so counts render
SI-scaled (Active Peers, Peer Disconnects → short). Polish the shared
filter descriptions: spell out the deployment-tier and network values and
add the perf network value.
Dashboards: network-traffic, ledger-data-sync, overlay-traffic-detail.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Brings the perf-iac dashboard filters (work item / branch / node role) into
phase 10. Conflicts in network-traffic and node-health were resolved by
keeping phase 10's query fixes (jobq_job_count rename; per-metric
label_replace identity for bytes-in breakdown) and re-applying the three
filter matchers on top.
Add three template variables (Work Item, Branch, Node Role) and their
label matchers to the remaining OTel dashboards so perf-iac comparison
runs can be filtered by ticket, comparison side (baseline/test), and node
role. The labels (xrpl_work_item, xrpl_branch, xrpl_node_role) are stamped
by perf-iac's alloy config as OTLP resource attributes and promoted to
Prometheus labels by Grafana Cloud; no repo collector change is needed.
Each variable mirrors the existing xrpl_network_type config (multi-select,
includeAll, allValue ".*") so local and non-perf runs — where these labels
are absent — continue to render (an absent PromQL label matches ".*").
Dashboards: consensus-health, ledger-operations, peer-network,
rpc-performance, transaction-overview, node-health, rpc-pathfinding,
fee-market, job-queue, peer-quality, validator-health.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add three template variables (Work Item, Branch, Node Role) and their
label matchers to the native-metric dashboards so perf-iac comparison
runs can be filtered by ticket, comparison side (baseline/test), and node
role. The labels (xrpl_work_item, xrpl_branch, xrpl_node_role) are stamped
by perf-iac's alloy config as OTLP resource attributes and promoted to
Prometheus labels by Grafana Cloud; no repo collector change is needed.
Each variable mirrors the existing xrpl_network_type config (multi-select,
includeAll, allValue ".*") so local and non-perf runs — where these labels
are absent — continue to render (an absent PromQL label matches ".*").
Dashboards: ledger-data-sync, network-traffic, overlay-traffic-detail.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Dashboard audit against the live datasource surfaced two broken panel queries
and applied UX polish across all 14 dashboards.
- node-health "Job Queue Depth": job_count -> jobq_job_count. The JobQueue
collector is wrapped in group("jobq") (Application.cpp), so the registered
job_count gauge is emitted with the jobq_ prefix; the panel queried the
unprefixed name and returned nothing.
- network-traffic "Overlay Traffic by Category" + "All Traffic Categories":
topk(N, rate({__name__=~".*_bytes_in"}[...])) errors on Mimir ("vector
cannot contain metrics with the same labelset") because rate() drops
__name__ and the many counters collapse. Replaced with an enumerated
label_replace form that re-attaches __name__ per metric, preserving the
{{__name__}} legend and per-series display-name overrides.
- All 14 dashboards: refresh set to 10s.
- peer-quality: each panel full screen width.
- validator-health: at most two panels per row (row headers preserved).
- docs: telemetry-runbook and 06-implementation-phases updated for the
jobq_ prefix and the network-traffic query pattern.
Verified end-to-end against a local mainnet xrpld node feeding the local
stack: jobq_job_count returns data (old job_count empty), both network-traffic
exprs execute (old form reproduces the labelset error), and panels render
through the Grafana proxy. All 14 pass validate_dashboards.py.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
libstdc++ ships <format> starting with gcc-13, so std::format in
SpanGuard::span() fails to compile on gcc-12. Build the "<prefix>.<name>"
span name via std::string::append instead and drop the <format> include.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The phase-6 change replaced this piechart with the applied-ratio state
timeline, but phase-9's independently-diverged dashboard kept its own
copy (id 17) which survived the merge-forward. Its query filtered on the
$txq_status template variable, which was removed alongside the swap, so
the panel resolved to txq_status=~"" and rendered empty. Delete it so
only the state timeline remains.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace the TxQ Accept Status piechart on the Transaction Overview
dashboard with a state-timeline showing each node's applied fraction of
TxQ accepts (applied / applied+failed) over time, colored by threshold
(green >=0.9, yellow >=0.7, red below). Remove the now-orphaned
txq_status template variable (the piechart was its only consumer) and
document the panel in the telemetry runbook.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The phase8->phase9 merge resolution took phase9's side of transaction-overview,
which reverted the Transaction Apply Failed Rate panel to the old
tx.apply/STATUS_CODE_ERROR query (constant zero). Re-apply the tx.transactor
stage=apply / ter_result!=tesSUCCESS repoint and raised thresholds.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add an Alloy-based collector config as an alternative to the reference
OpenTelemetry Collector. It forwards xrpld OTLP traces and native metrics
to the Grafana Cloud OTLP gateway and, critically, runs a spanmetrics
connector so span-derived RED metrics (span_calls_total,
span_duration_milliseconds_*) are produced. Those metrics are not emitted
by xrpld; they are derived from spans, so an Alloy node without this
connector is missing from every span-based dashboard.
All credentials, endpoints, usernames and the per-node host label are read
from the environment (no secrets in the file). Adds a matching
.env.grafanacloud-alloy.example and gitignore exception, and the otelcol
term to the cspell dictionary.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Seed the $node template variable from target_info instead of a
signal-specific metric. Signal metrics only list nodes that emit that
specific signal, so nodes (e.g. Alloy-collected ones) were missing from
the dropdown and every panel. target_info is emitted for every node, so
all nodes are always selectable. Panels already filter on
service_instance_id=~"$node", so no panel changes are needed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Seed the $node template variable from target_info instead of a
signal-specific metric. Signal metrics only list nodes that emit that
specific signal, so nodes (e.g. Alloy-collected ones) were missing from
the dropdown and every panel. target_info is emitted for every node, so
all nodes are always selectable. Panels already filter on
service_instance_id=~"$node", so no panel changes are needed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Seed the $node template variable from target_info instead of a
signal-specific metric. Signal metrics only list nodes that emit that
specific signal, so nodes (e.g. Alloy-collected ones) were missing from
the dropdown and every panel. target_info is emitted for every node, so
all nodes are always selectable. Panels already filter on
service_instance_id=~"$node", so no panel changes are needed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Observe the overflow total from Overlay::getJqTransOverflow() via an
ObservableCounter instead of the unused push helper (which had no call
site, so the metric read flat zero). Keeps the increment path in the
overlay-owned code and removes the dead incrementJqTransOverflow() helper.
Update the job-queue panel to reflect the now-live metric, and refresh
stale xrpld_-prefixed metric names in the header comments.
NOTE: committed without a local build per instruction; needs compile +
clang-tidy verification before push.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace hardcoded [5m] rate windows with [$__rate_interval] inside
rate()/increase() on plain-counter panels so ops/s panels resample
correctly on zoom-out instead of under-sampling or flat-lining on
short windows. Histogram _bucket windows keep their fixed [5m] span.
Set Min step ("interval": "15s") on every rate target to match the
Prometheus 15s scrape interval.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The spanmetrics connector had no namespace, so it emitted traces_span_metrics_*
metric names by default. The span dashboards and docs are renamed to query
span_* names; this is only correct if the connector emits them too, so add
namespace: "span" to the spanmetrics connector. Both sides change together:
renaming the dashboards without the namespace (or vice versa) would break the
pipeline. Matches the phase9 collector config.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The phase7 OTelCollector::formatName lowercases and strips names, emitting
snake_case metrics with no xrpld_ prefix. The native Grafana dashboards and
the telemetry docs still queried the old xrpld_CamelCase names, so they were
broken against their own pipeline. Rename every metric name to match what the
code emits: drop the xrpld_ prefix and lowercase the remainder. The two job
histograms also drop the redundant 'duration' word (job_queued_us,
job_running_us) to match the phase9 forms. Add havetxset to the cspell
dictionary since the lowercased metric name no longer word-splits.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Completes the default change applied on phase-6-statsd and
phase-7-native-metrics. The four dashboards introduced on phase-9
(validator-health, peer-quality, fee-market, job-queue) now also
pre-select 'xrpld' rather than 'All' for the service_name filter.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Continuation of the same default change applied on phase-6-statsd
(50d7047abe): the 5 dashboards introduced on phase-7 (formerly
system-*, renamed to bare names on this branch) also get their
service_name filter pre-selected to 'xrpld' instead of 'All'.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Change the pre-selected value of the service_name template variable from
'All' to 'xrpld'. On first load, dashboards now filter to service.name=xrpld
directly rather than showing every service the collector sees. The 'All'
option is still selectable — this only changes the default the widget
starts on.
Reduces the chance of a fresh dashboard link showing unrelated services
(e.g. non-xrpld exporters sharing the same Grafana Cloud stack) before
the user narrows the filter.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Grafana Cloud's OTLP ingest already promotes the `service.name` resource
attribute to a `service_name` metric label. The `transform/cloudlabels`
processor was ALSO setting `service_name` as a datapoint attribute,
which Mimir treats as a duplicate label and concatenates as
`service_name="xrpld;xrpld"` on every series. Dashboards filtering on
`service_name=xrpld` then returned no data.
Remove the redundant `set(attributes["service_name"], ...)` line and
add a comment documenting the doubling. `service.instance.id` does not
exhibit the same behaviour, so that copy is retained for the $node
filter to keep resolving.
Reintroduces the pre-e76eed5296 shape (which was correct on this label)
without reverting that commit's other cloud-label fixes.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Completes the rename applied on phase-6-statsd and phase-7-native-metrics.
These four dashboards were introduced on phase-9: peer-quality and
validator-health via b92354715d, fee-market and job-queue via 25868f2740.
Same reasoning: the underlying Prometheus label is `service_name`,
so the UI display label should say 'Service Name' rather than the
ambiguous 'Service'.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Continuation of the same rename applied on phase-6-statsd (commit
91c6582bd5): 5 more dashboards introduced on phase-7 (formerly the
system-* set, renamed to bare names by 2c590a47c5) still carried the
older 'Service' UI label. Aligning them with the rest.
Same reasoning as the phase-6 commit: the underlying Prometheus label
is `service_name`, and the UI label 'Service' misled users into typing
`service=xrpld` in Grafana Metrics Drilldown.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The template variable's Prometheus label name is `service_name` but the
UI display label read 'Service', which mismatched what users typed into
Grafana's Metrics Drilldown UI (which defaults to filter on a bare
`service` label). Rename the display label to 'Service Name' so it
unambiguously points at the underlying `service_name` Prometheus label.
Behaviour unchanged; template-variable name (`service_name`) and PromQL
label are the same. Only the widget label shown at the top of the
dashboard changes from 'Service' to 'Service Name'.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- OTelCollector::formatName made static (no member state); braces added;
<cctype> added for std::tolower.
- Telemetry.cpp getMeter(): braces around single-statement if.
- GetMeter.cpp: drop unused <opentelemetry/metrics/meter.h> and
sync_instruments.h; add <xrpl/beast/utility/Journal.h> and <memory>;
make sdkProvider const.
- ValidationTracker.h: add <cstddef> for std::size_t.
- detail/ValidationTracker.cpp: use std::views::reverse range-based loop;
add <ranges>.
- test ValidationTracker.cpp: use auto for cast result.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Strip xrpld_ prefix, lowercase beast::insight names, and replace
traces_span_metrics_ with span_ in all remaining tracked files:
alert rules, integration tests, workload validation, TESTING.md,
OpenTelemetryPlan docs, code comments, and config templates.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove redundant xrpld_ prefix from all 34 MetricsRegistry instrument
names — service identity comes from the OTel resource, not the metric
name. Duration histograms retain their unit suffix (job_queued_duration_us,
job_running_duration_us, rpc_method_duration_us).
Set spanmetrics connector namespace to "span" so derived metrics become
span_calls_total and span_duration_milliseconds_bucket instead of the
verbose traces_span_metrics_ prefix.
Update all 14 dashboard JSONs: strip xrpld_ prefix, lowercase
beast::insight metric names, replace traces_span_metrics_ with span_.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add service_name to transform/cloudlabels so Grafana Cloud dashboards
can filter by service.name (fixes empty panels for native metrics).
Set spanmetrics namespace to "span" in the cloud collector config and
update dashboards + runbook: traces_span_metrics_ → span_.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Drop the xrpld_ prefix from beast::insight metric names — the OTel
resource (service.name=xrpld) already identifies the service. formatName()
now lowercases and sanitizes spaces/dots to underscores for
Prometheus-compatible names.
Also fix 4 source strings containing spaces that produced invalid OTel
instrument names (silently rejected by the SDK).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Raise TX error rate threshold from 50% to 95%. Short-lived CI test
environments lack pre-funded accounts for complex transactions
(AMMCreate, EscrowFinish, NFTokenCreateOffer, etc.), causing expected
failures that do not indicate an instrumentation problem.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replace unused observer_result.h with nostd/shared_ptr.h and
nostd/unique_ptr.h in MetricsRegistry.h (misc-include-cleaner)
- Use auto for cast initializations in MetricsRegistry.cpp and
ValidationTracker test (modernize-use-auto)
- Run rename scripts: rippled -> xrpld in comments and docs
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add job_type to otel-naming builtins (standard Prometheus label for
job-queue metrics used in node-health dashboard)
- Add OpenTelemetry SDK to UBSAN suppressions (intentional unsigned
integer overflow in attributemap_hash.h hash computation)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add missing <string_view> include in Consensus.h and RCLConsensus.h.
Remove unused SHAMap.h include from RCLConsensus.h.
Add braces around single-line if/else bodies in RCLConsensus.cpp.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove unused SpanNames.h from Transactor.cpp (transitively included
via TxApplySpanNames.h). Remove unused <ranges> from
TraceContextValidation.h (std::ranges::any_of lives in <algorithm>).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace std::transform with std::ranges::transform in Redaction.cpp.
Replace std::all_of with std::ranges::all_of in Redaction test.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add missing <cstdint> include for std::uint32_t in Telemetry.h.
Add braces around single-line if bodies in Telemetry.cpp.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update the transform/cloudlabels processor to set service_instance_id
instead of exported_instance, matching the local Prometheus path where
resource_to_telemetry_conversion promotes service.instance.id with that
exact label name. Also update source comments and documentation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Merge forward from phase6/7 and apply the same fix to phase9's own
dashboards and alerting rules. exported_instance is a Prometheus scrape
artifact absent in push-based pipelines; service_instance_id is always
present via resource_to_telemetry_conversion.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
exported_instance is a Prometheus scrape artifact that only exists when
Prometheus scrapes a target with a conflicting instance label. Users
running push-based pipelines (remote_write, OTLP) never get this label.
service_instance_id is always present via resource_to_telemetry_conversion
regardless of ingestion path.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
exported_instance is a Prometheus scrape artifact that only exists when
Prometheus scrapes a target with a conflicting instance label. Users
running push-based pipelines (remote_write, OTLP) never get this label.
service_instance_id is always present via resource_to_telemetry_conversion
regardless of ingestion path.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The workload validator queried each expected metric once, immediately after
a fixed post-workload propagation wait. Several beast::insight metrics
(ledger-age and peer-finder gauges, overlay-traffic and rpc-request counters)
only populate after the node validates ledgers and sustains peer traffic,
then travel a 1s OTLP export + 15s Prometheus scrape before they are
queryable. On a slower CI runner that pipeline can settle after the wait
ends, so the single query raced and reported "0 series", failing 12 checks
that pass locally with the same config and binary.
Poll each metric on the /api/v1/series endpoint until it appears or a 45s
window (two scrape cycles) elapses. Present metrics still return on the first
query with no added delay; a genuinely-absent metric still fails after the
timeout. Makes the check robust to runner speed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
xrpld resolves a relative [debug_logfile] against the config file's own
directory (docker/telemetry), not the working directory. The value
`docker/telemetry/data/logs/devnet/debug.log` therefore doubled the
prefix to docker/telemetry/docker/telemetry/... — the collector's mount
never saw the log, so no lines reached Loki.
Use the config-dir-relative form `data/logs/devnet/debug.log`, which
resolves to docker/telemetry/data/logs/devnet/debug.log — exactly the
dir the compose stack bind-mounts as /var/log/xrpld.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Extend the log-mount consolidation to the phase-10 workload-validation
assets, which still used the old `rippled` container path:
- docker-compose.workload.yaml: mount target renamed to /var/log/xrpld,
source made overridable via XRPLD_LOG_DIR (default /tmp/xrpld-validation)
- otel-collector-config.grafanacloud.yaml: single glob /var/log/xrpld/*/debug.log
- run-full-validation.sh sets XRPLD_LOG_DIR to its workdir
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The collector's log ingestion used a container path named after the old
`rippled` binary and defaulted its host mount to a location that either
needed root or diverged from where the telemetry configs actually write.
Consolidate on one accessible log root:
- container mount target renamed /var/log/rippled -> /var/log/xrpld
- filelog glob -> /var/log/xrpld/*/debug.log
- compose mount source defaults to the repo-relative ./data/logs
(user-owned, no root), overridable via XRPLD_LOG_DIR
- devnet telemetry cfg writes to data/logs/devnet/debug.log so it lands
one subdir below the mount root and matches the glob
- integration-test.sh sets XRPLD_LOG_DIR to its own workdir
- docs/runbook/task-list updated to match
The default xrpld config (cfg/xrpld-example.cfg) is intentionally left
untouched.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- PerfLogImp::rpcEnd(): return after the requestId-not-found UNREACHABLE
so a stale (now - epoch) duration is no longer recorded to the counter
and histogram in release builds.
- MetricsRegistry peer-version gauge: compare versions numerically via
BuildInfo::encodeSoftwareVersion() instead of a lexicographic string
compare, stripping the non-digit prefix so peer 'rippled-X.Y.Z' lines
up with our bare 'X.Y.Z'. Fixes every peer counting as higher-version.
- MetricsRegistry::stop(): call Shutdown() before ForceFlush() before
reset() so the reader thread stops before teardown and no gauge
callback fires during shutdown.
- MetricsRegistry::start(): extract initExporterAndProvider() and
initSyncInstruments() helpers to keep each function under the line
limit; no behavior change.
- time_in_current_state_seconds: read NetworkOPs::getServerStateDurationUs()
(a lightweight accessor over StateAccounting) and convert microseconds
to seconds, replacing the hardcoded 0.0.
- Correct log-timestamp examples to real Logs::format() output
(2024-Jan-15 10:30:45.123456 UTC) in docs, TESTING.md, and reference.
- Update Loki to v3.4.2 and switch the collector exporter to
otlphttp/loki on the native /otlp endpoint (docs + task list).
- Make the collector log-mount path configurable via XRPLD_LOG_DIR.
- Remove implementation-phase references from shipped config/script
comments, keeping the functional descriptions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reviewer noted implementation-phase references should not appear in
shipped config comments. Drop the phase-8-owned 'Phase 8' tokens from
loki.yaml and tempo.yaml, keeping the functional descriptions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The getYays()/getNays() accessors already exist on the phase-7 branch.
A cleanup+restore pair on this branch moved them, producing a no-op
diff against phase-7 that mixed consensus churn into the log-correlation
PR. Restore DisputedTx.h to the phase-7 state so this branch carries no
unrelated consensus change.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reviewer noted implementation-phase references should not appear in
shipped config comments. Drop the "Phase 3/4" token, keep the description.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reviewer noted implementation-phase references should not appear in
shipped config comments. Drop the "Phase 4" tokens, keep descriptions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reviewer noted implementation-phase references should not appear in
shipped config comments. Drop the "Phase 3" tokens, keep descriptions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reviewer noted implementation-phase references should not appear in
shipped config comments. Drop the "Phase 2" tokens, keep descriptions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reviewer noted implementation-phase references should not appear in
shipped config comments. Drop the "Phase 1b" token, keep the filter
description.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- node-health.json: repoint the two "All Jobs" detail panels from the
__name__ regex over old xrpld_<job>_milliseconds_bucket names to the
native xrpld_job_running/queued_duration_us_bucket histograms, grouping
by the job_type label and legending on {{job_type}}.
- OTelCollector::callHooks: copy the hook list under mutex_ and invoke
handlers outside the lock. A handler can drop the last reference to an
OTelHookImpl, whose destructor calls removeHook() and re-acquires the
non-recursive mutex_ — invoking handlers under the lock could deadlock.
- ~OTelGaugeImpl: document that the SDK ObservableRegistry serializes
RemoveCallback against the Observe() callback pass with one mutex, so
callback removal is synchronous and gaugeCallback cannot run on a
dangling pointer after the destructor.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Key Jobs execution/dequeue panels queried StatsD-era per-job metric
names (xrpld_<job>_milliseconds_bucket, xrpld_<job>_q_milliseconds_bucket)
that the native OTel path no longer emits. Phase 9's MetricsRegistry emits
job timings as two label-dimensioned histograms instead:
xrpld_job_running_duration_us_bucket{job_type="<job>"}
xrpld_job_queued_duration_us_bucket{job_type="<job>"}
Rewrite the 22 affected panel queries (11 job types x running/queued) to
the label-dimensioned form, preserving the histogram_quantile / rate /
sum-by structure and all template-variable filters.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- MetricsRegistry: service.name was recorded as boolean true because the
string literal "xrpld" bound to the OTel AttributeValue variant's
char-const* -> bool overload. Assign std::string so it selects the
string alternative; Prometheus now shows service_name=xrpld (and
exported_job round-trips correctly) on every MetricsRegistry series.
- Grafana alerting contact points: the Slack url / email addresses used
${ENV_VAR} references that expand to empty when unset, and Grafana's
provisioning validator (which lacks ${VAR:-default} support) then
crashes the whole stack on startup. Use non-empty disabled placeholders
(unroutable webhook host, .invalid email) so the stack boots with zero
configuration; delivery stays off until a real destination is supplied.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- ValidationTracker: remove dead hard-trim Pass 1. The preceding
eviction loop already erases every reconciled entry older than the
late-repair cutoff using the same condition, so Pass 1 could never
match. The "drop any reconciled entry" fallback is retained.
- integration-test.sh: probe the StatsD port with a UDP-aware check
(ss -ulnp) instead of curl. StatsD is UDP-only on 8125, so the TCP
curl probe always reported "refused" — a false negative. Guarded by
a command -v ss check that logs a skip when ss is unavailable.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Give developers direct access to the full OpenTelemetry metrics API,
symmetric with getTracer(), so all seven OTel instrument types
(including UpDownCounter, sync Gauge, and the observable variants) are
reachable — not just the four beast::insight models.
- Telemetry: build and own the metrics pipeline (OTLP HTTP metric
exporter + PeriodicExportingMetricReader + SpanMetrics histogram view)
alongside the tracer, sharing the same resource attributes and TLS
config. Register it globally via metrics::Provider::SetMeterProvider
and expose Telemetry::getMeter(). Metrics enable with [telemetry];
the metrics endpoint is derived from the trace endpoint.
- beast OTelCollector: no longer owns a pipeline. It fetches the global
Meter, becoming a thin shim over the shared provider (legacy path
during beast deprecation). This also resolves the review note that the
metric exporter ignored [telemetry] use_tls — TLS now comes from the
shared telemetry pipeline.
- Add a libxrpl unit test covering getMeter() on the enabled (global
provider) and disabled (noop) paths, exercising an UpDownCounter.
Design: docs/superpowers/specs/2026-07-07-direct-otel-metrics-api-design.md
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- cmake: keep the opentelemetry-cpp umbrella target for the beast metrics
link and document why. The reviewer suggested linking individual
component targets to avoid over-linking, but the OTel Conan package
under-declares inter-component dependencies (the OTLP client references
sdk::common symbols without a declared edge), so naming components
directly reorders the static link into an unresolvable state. Verified
by building xrpl_tests both ways.
- OTelCollector.h: add usage examples, thread-safety and limitations
@note blocks to the class doc.
- OTelCollector.cpp: correct the @param name docs on the instrument
Impl constructors to describe the already-formatName()'d value.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Phase 11 task list still linked the moved External Dashboard Parity spec
at its old docs/superpowers/ path. Point them at the appendix in
06-implementation-phases.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The 09-data-collection-reference and Phase9 task list still linked the moved
External Dashboard Parity spec at its old docs/superpowers/ path. Point them
at the appendix in 06-implementation-phases.md where the content now lives.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Move the phase-9 alerting design spec out of docs/superpowers/ (which the
project guidelines say not to create) into an appendix of
OpenTelemetryPlan/Phase9_taskList.md, keeping the phase plan self-contained.
Drops the stale one-off branch/PR-restructure section from the spec.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Move the standalone External Dashboard Parity design spec out of
docs/superpowers/ (which the project guidelines say not to create) and into
an appendix of OpenTelemetryPlan/06-implementation-phases.md, so the phase
plan is self-contained. Repoint the 10 "Source" links in the phase 3/4/7
task lists to the new appendix anchor.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Update the phase-9 task list and the alerting design spec to reference the
Alerting section of docs/telemetry-runbook.md, now that the standalone
ALERTING.md has been folded into the runbook.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the placeholder webhook contact point with two env-driven tiers:
- xrpld-default (Slack) receives warning-severity alerts
- xrpld-critical (Slack + email) receives critical-severity alerts, via a
severity=critical child route in the notification policy
Secrets stay out of git: contactpoints.yaml references ${SLACK_WEBHOOK_URL},
${SLACK_CHANNEL}, and ${ALERT_EMAIL_TO}, which Grafana expands from a
gitignored .env.alerting (loaded via an optional env_file on the grafana
service). .env.alerting.example documents the variables; email additionally
needs GF_SMTP_* configured.
Fold the ALERTING.md runbook content into the Alerting section of
docs/telemetry-runbook.md (operator docs live in one place) and delete
ALERTING.md. Move the .env.* ignore rule into docker/telemetry/.gitignore.
Verified live: Grafana 11.5.2 provisioned both contact points, the
severity-routed policy, and all 6 rules with no errors.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Add PathFinding Spans section to the runbook Span Reference and
remove the incorrect "not emitted / later-phase" annotations:
pathfind.request/compute/discover/update_all are emitted since
Phase 2 (upstream of Phase 5).
- Runbook label names use the bare spanmetrics dimensions
(command, rpc_status, local, consensus_mode); fix stale docs.
- 05-configuration-reference: note tls_client_cert/key require use_tls=1.
- TelemetryConfig: reject mTLS cert paths when use_tls=0 so telemetry
is not silently sent in plaintext.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Soften the "non-reversible" wording in Redaction.h. XRP account
addresses are a public, enumerable set, so an unsalted hash is
obfuscation (keeps plaintext out of traces/dashboards), not a
secrecy guarantee. Document why a salt is intentionally omitted:
it would break cross-node/restart correlation, the reason for
hashing rather than dropping the attribute.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Resolve the doCommand error-path span suffix and command attribute against
the handler registry instead of using raw request input. A recognized command
yields its canonical handler name; anything else (unregistered command, or
neither command/method present) collapses to the literal "unknown". This bounds
telemetry cardinality to the finite set of handler names plus "unknown" while
preserving per-command error attribution for known-but-rejected commands
(e.g. a submit that hit rpcTOO_BUSY stays rpc.command.submit). The command
attribute is promoted to a Prometheus label by the spanmetrics connector, so
attacker-controlled input previously enabled unbounded label cardinality.
Also add the missing LCOV_EXCL_STOP in the first SpanGuard::linkedSpan overload
so its coverage exclusion no longer bleeds into the second overload's guards.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Change the six remaining `type: static` search filters in tempo.yaml to
`type: dynamic` so Grafana queries Tempo for available tag values instead
of requiring manual free-text entry. Aligns service-name, node-id,
network-type, span-name, span-status, and span-duration with the existing
node-version and network-id filters.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Final four dashboards introduced in this phase: replace the "pratik-xrpld"
$service_name example with "xrpld-validator".
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same cleanup for the five system-metric dashboards introduced in this phase:
replace the "pratik-xrpld" $service_name example with "xrpld-validator".
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The $service_name template variable's description used a personal-name example
value. Replace "pratik-xrpld" with the generic "xrpld-validator" in the five
dashboards that introduced it.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Follow the config change that replaced the personal home directory: the
runbook prose now points at /var/log/xrpld/mainnet/debug.log, matching the
mainnet config and the collector's filelog receiver.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the hardcoded /home/pratik/xrpld-logs path with the conventional
/var/log/xrpld in the mainnet config's [debug_logfile], the docker-compose
log volume mount, and the collector-config comment. Removes a personal home
directory from shipped config.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Add an operator-facing Alerting section to the runbook covering the six
provisioned Grafana rules and how to point them at a real receiver.
- Strip the "(Phase 9)" suffix from the MetricsRegistry Gauges heading.
- ALERTING.md: drop the internal dev-phase reference from the intro.
- Phase9_taskList.md: expand Task 9.9 to describe the provisioned alerting
(rules, contact point, policy, ALERTING.md runbook).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Operator-facing cleanup: strip the "(Phase 8)" suffix from the Log-Trace
Correlation section heading in the runbook.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same operator-facing cleanup as the RPC/transaction/consensus headings: the
runbook documents the running system, not the development phase that added
each span. Strip the "(Phase 6)" suffixes from the Ledger and Peer span
reference headings.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The operator runbook is for people running OpenTelemetry against a node, not
a record of the internal development phases. Strip the "(Phase N)" suffixes
from the span-reference headings and replace the personal-name example
service.name value with a generic one.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add operator-facing alerting on the health-critical metrics phase 9 emits,
completing the "alerting rules" deliverable already listed in the phase-9
task list and Jira story.
Provisioned YAML under grafana/provisioning/alerting/ (auto-loaded by the
Grafana container; no docker-compose change needed):
- rules.yaml: 6 alerts in 3 groups — consensus/ledger health, validator
health, job queue. Each is a query(A, 5m rate/quantile by exported_instance)
-> reduce(B, last) -> threshold(C) on the Prometheus datasource.
- contactpoints.yaml: xrpld-default webhook receiver (local-dev placeholder).
- policies.yaml: flat routing to xrpld-default, grouped by
alertname + exported_instance.
Also adds docker/telemetry/ALERTING.md runbook (per-alert meaning, causes,
threshold tuning, how to point alerts at a real receiver) and the design spec.
Verified live: Grafana 11.5.2 provisioned all 6 rules with no rule in error
health; contact point loaded.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The test Peer adaptor's getTelemetry() returns a shared function-local
NullTelemetry instance and does not read any per-peer state, so
clang-tidy's readability-convert-member-functions-to-static flags it as
an error under -warnings-as-errors. Mark it static; it has no callers
that depend on it being a non-static member.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Filter dropdowns were empty on Grafana Cloud: template query-variables
referenced uid ${DS_PROMETHEUS} (a datasource variable that had no value
and was hidden), while panels used the bare default datasource. So panels
rendered but Service/Environment/Network/Node variables resolved against
nothing.
Wire DS_PROMETHEUS (and DS_TEMPO where used) through every panel, target,
annotation, and query-variable, and make the pickers visible. Empty
current value means each instance auto-selects its default datasource of
that type, so one dashboard set works on both the local stack and Grafana
Cloud without per-instance uids.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Grafana Cloud ingests metrics via OTLP with no Prometheus scrape, so the
tier/instance resource attributes never became series labels the way the
local prometheus exporter promotes them. Dashboard $node,
$deployment_environment, and $xrpl_network_type filters therefore matched
nothing on Cloud (only service_name existed).
Split the metrics pipeline: metrics/local keeps the Prometheus exporter
(unchanged label promotion), metrics/cloud adds a transform that copies
service.instance.id -> exported_instance, deployment.environment ->
deployment_environment, and xrpl.network.type -> xrpl_network_type onto
datapoint labels so the same dashboards filter correctly on both backends.
Also strip telemetry.sdk.* on all pipelines. Validated against
otel-collector-contrib 0.121.0.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add a resource/stripsdk processor that deletes telemetry.sdk.language,
telemetry.sdk.name, and telemetry.sdk.version (auto-added by the OTel SDK
to every Resource). Wire it into the traces pipeline; the metrics and
logs pipelines pick it up on the branches that introduce them.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Point the workload validation uid list and implementation-phases doc at
the renamed bare dashboard uids.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Rename the 4 xrpld-* dashboard files to bare names and set every
dashboard uid to its bare filename stem; update doc/plan references.
Removes the redundant xrpld-/system- prefixes now that titles are clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Rename the 5 system-* dashboard files to bare domain names (uid ==
filename stem) and update all doc references. The system- prefix was
redundant now that titles no longer carry a recorder-pipeline suffix.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Make dashboard uids bare domain names (uid == filename stem) and update
doc references. Grafana links/bookmarks to the old uids will need
refreshing.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Grafana renders the dashboard-level description as plain text (not
markdown), so the h6 headings/italics showed as literal characters.
Convert to plain 'What this shows:' / 'Use it to:' labeled lines. Panel
descriptions (which DO render markdown) are unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Grafana renders the dashboard-level description as plain text (not
markdown), so the h6 headings/italics showed as literal characters.
Convert to plain 'What this shows:' / 'Use it to:' labeled lines. Panel
descriptions (which DO render markdown) are unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The separate xrpld-rpc-perf-otel dashboard was merged into xrpld-rpc-perf,
so the validation harness must no longer expect it to exist.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Match the reader-facing dashboard catalog to the actual dashboard titles,
which no longer carry the recorder-pipeline suffix.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The separate RPC Performance (OTel) dashboard was folded into RPC
Performance as a per-method section, so describe it as an addition to the
existing dashboard rather than a new one.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fold the 7 panels from the separate RPC Performance (OTel) dashboard into
RPC Performance as a labelled 'Per-Method Detail' section, and retire the
duplicate dashboard. The two showed the same subject (RPC) from two data
sources; one dashboard with two sections is clearer and loses no panel.
Adds the $method template variable for per-method filtering.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Consensus Health, Ledger Operations, and Transaction Overview had no
top-level description, so Grafana showed no info button. Add a one-line
summary of what each dashboard shows and its data source.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The integration-test.sh trace queries used resource.service.name="rippled",
but the OTel resource attribute is "xrpld" (TelemetryConfig.cpp, MetricsRegistry.cpp),
so every span check and the service-registration check returned no results.
Also restore .github/scripts/rename/README.md, which an overzealous
find-and-replace had corrupted into circular no-ops (e.g. rename "xrpld" to
"xrpld"). The file documents the rename process and must retain the source
names (rippled/ripple); restored to match develop.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Every dashboard already lives in the rippled/xrpld dashboard folder and
carries domain-specific tags (consensus, rpc, statsd, etc.), so the
leading "rippled" tag added no filtering value and only cluttered the
tag list.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Every dashboard already lives in the xrpld folder and carries a
domain-specific tag (rpc, consensus, statsd, etc.), so the leading
"xrpld" tag added no filtering value and only cluttered the tag list.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Repoint the Grafana Cloud collector config comment at the "Deployment
Tiers" section of docs/telemetry-runbook.md now that the standalone tier
doc has been folded into the runbook.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Remove the standalone docker/telemetry/DEPLOYMENT_TIERS.md; its content now
lives in the "Deployment Tiers" section of docs/telemetry-runbook.md, the
established operator doc. Repoint the Grafana Cloud collector config comment
at that section.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- 09-data-collection-reference: add $service_name, $deployment_environment,
$xrpl_network_type to the template-variable table; note they appear on all
dashboards and point to the runbook's Deployment Tiers section.
- TESTING.md: add a Prometheus check verifying the tier labels
(deployment_environment, xrpl_network_type, service_name) are present on
metric series.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add a Deployment Tiers section covering the four filtering dimensions
(node, service, network, environment), who owns each attribute, the
collector's upsert-vs-insert rule, per-tier collector values, and how the
tier labels reach metrics. This is the operator-facing home for the
deployment-tier feature.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- 05-configuration-reference: service_name now applies to metrics as well
as traces (service.name resource attribute).
- 07-observability-backends: document that collector enrichment includes
deployment-tier tagging (deployment.environment + fallback
xrpl.network.type) for filtering one Grafana stack by tier.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Apply the deployment-tier tagging to the Grafana Cloud dual-export config
so cloud data is filterable by tier just like the local stack:
- add the resource/tier processor (deployment.environment upsert,
xrpl.network.type insert) and run it on all three pipelines.
- enable resource_to_telemetry_conversion on the prometheus exporter and
add the tier keys to spanmetrics for the local scrape surface.
Add docker/telemetry/DEPLOYMENT_TIERS.md documenting the four filtering
dimensions, the upsert-vs-insert ownership rule, per-tier collector
values, and how the labels reach each signal.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add $service_name, $deployment_environment, $xrpl_network_type template
variables to the metric-gap-fill dashboards and wire them into every
panel query that filters by $node, so these dashboards can be sliced by
tier like the rest of the set.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add resource/tier to the logs pipeline so log records carry the same
deployment.environment and xrpl.network.type attributes as traces and
metrics. This lets a single Grafana stack filter logs by tier alongside
the other signals. resource/logs still runs first to set service.name
and the job label; resource/tier then adds the tier attributes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The native OTel metrics path hard-coded service.name="xrpld" and stamped
no network attribute, while traces stamped a configurable service.name
and xrpl.network.type. Metrics therefore could not be filtered by service
or network. Align the two paths:
- OTelCollector::New / OTelCollectorImp gain serviceName + networkType
params. service.name uses the configured value (default "xrpld" when
unset, preserving today's behavior); xrpl.network.type is stamped when
provided. The key is a string literal because beast/insight sits below
the telemetry module and cannot include its SpanNames const.
- CollectorManager reads service_name from [insight], falling back to the
[telemetry] value, and receives the network type from the caller.
- Application derives the network type once via the shared
telemetry::networkTypeFromId, now declared in Telemetry.h and moved out
of an anonymous namespace so the trace and metric paths reuse a single
0/1/2 -> mainnet/testnet/devnet mapping (no duplication).
Dashboards (5 system-* files): add $service_name, $deployment_environment,
$xrpl_network_type template variables and wire them into every panel query
that filters by $node.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Extend the deployment-tier tagging to the metrics path and expose it in
the dashboards so a single Grafana stack can be sliced by environment,
network, and service:
Collector (otel-collector-config.yaml):
- prometheus exporter: enable resource_to_telemetry_conversion so
resource attributes (deployment.environment, xrpl.network.type,
service.name, service.instance.id) become Prometheus labels.
- metrics pipeline: run resource/tier so statsd metrics are tagged too.
- spanmetrics: add the tier keys to the resource-metrics grouping key
so per-tier series stay grouped separately.
Dashboards (5 files):
- add $service_name, $deployment_environment, $xrpl_network_type
template variables, named to match their Prometheus labels.
- inject the matching label filters into every panel query that
already filters by $node.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add a resource/tier processor so each collector stamps the deployment
tier onto the signals it forwards, enabling a single Grafana stack to
hold data from many collectors and filter by tier:
- deployment.environment (upsert): the collector is authoritative for
its environment (local/ci/test/prod).
- xrpl.network.type (insert): the xrpld node already stamps its own
chain, so the collector only fills this when absent, never
overwriting a node's real network.
Wired into the traces pipeline (the only pipeline on this branch);
later phases extend it to metrics and logs as those pipelines appear.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Apply the same width rule from the earlier transaction-overview fix to the
job-queue, rpc-perf-otel, and rpc-performance dashboards. Timeseries panels
whose legend is a right-side table need the horizontal room, so they go to
full width (w=24); each stacks onto its own row and panels below shift down.
Pie charts, heatmaps, and bottom-legend panels stay half width.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CI enabled more clang-tidy checks via the upstream merge:
- DiscardScope discardScope -> const (misc-const-correctness)
- drop the (::max)() Windows macro-guard parens; NOMINMAX is defined
project-wide so the bare form is correct (readability-redundant-parentheses)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add an opt-in collector config + compose override that forward traces,
metrics, and logs to a Grafana Cloud OTLP gateway alongside the local
backends. Enabling it is a runtime choice (layer the override compose
file) — no xrpld rebuild, base stack unchanged. Credentials come from a
gitignored .env file. Documented in the telemetry runbook and TESTING.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 5 added 4 more unit-test cases calling the old setupTelemetry name.
Align them with the renamed makeTelemetrySetup free function so xrpl_tests
builds.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The free function was renamed makeTelemetrySetup but this unit test still
called the old name, breaking the xrpl_tests build on arm64.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The followsFrom linkedSpan overload referenced attr::linkType /
attr_val::followsFrom, which don't exist in SpanNames.h — the other two
link sites already use the local kLinkTypeKey / kLinkTypeFollowsFrom
constants. Unify the third site. Fixes arm64 build (-Werror).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The network-type label is derived from [network_id] in TelemetryConfig.cpp;
unmapped/unset IDs fall through to "unknown", not "standalone". Align the
design-doc resource-attribute table with the code and cfg example.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The follows-from link attribute used the dotted xrpl.link.type form, which
is reserved for resource attributes. Span attributes use the <domain>_<field>
underscore form, so the key becomes link_type.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Tempo datasource search filter used the stale dotted xrpl.consensus.ledger_id
tag, but the consensus.round span emits the bare consensus_ledger_id key
(ConsensusSpanNames.h). The dotted tag matched no spans and failed the Rule C
naming check. Align it with L1.
The TxQSpanNames.h span-hierarchy comment and the Phase 3 task-list attribute
table showed txq.accept.tx, but the constant emits txq.accept_tx
(op::acceptTx = "accept_tx"). Correct both.
- peer.validation.receive now asserts the shared bare ledger_hash /
full_validation keys (was the dotted xrpl.ledger.hash and validation_full);
PARITY_SPAN_ATTRS checks both on the peer span too.
- Fix a span-name drift: the per-transaction accept span is txq.accept_tx
(op::acceptTx = "accept_tx"), not txq.accept.tx — the old assertion never
matched and was silently skipped as optional.
- Drop the "intentionally dotted" notes; there is no dotted span attribute.
The peer.validation.receive span now records the ledger hash and full-validation
flag under the shared bare keys ledger_hash / full_validation (same keys the
consensus validation spans use), so there is no longer a dotted span attribute.
Remove the "only dotted span attribute" note and update the peer attribute table.
The peer.validation.receive span recorded the ledger hash under the dotted
xrpl.ledger.hash form — the only dotted span attribute, and inconsistent with
the bare ledger_hash the consensus validation spans use for the same value.
Make the base SpanNames.h ledgerHash bare (ledger_hash) and add the shared
fullValidation key, so the peer and consensus validation spans share one key
per concept (told apart by span name, not an emitter prefix). PeerSpanNames now
re-exports both from the base; the peer validation_full key folds into the
shared full_validation. The peer trust attrs (proposal_trusted /
validation_trusted) already follow the convention and are unchanged.
Make consensus span-attribute keys consistent with the peer spans that record
the same concepts, using ONE shared name per concept (told apart by span name),
not an emitter prefix:
- Add the shared ledger_hash and full_validation keys to the base SpanNames.h
(a ledger-object property and an is-full-validation flag, both shared with the
peer validation spans); consensus re-exports them via `using` instead of
defining local copies.
- trusted (bare) -> proposal_trusted on consensus.proposal.receive and
validation_trusted on consensus.validation.receive — the same message-type
qualification the peer.{proposal,validation}.receive spans use.
Also collapse a same-emitter duplicate: establishCounter_ was recorded under
both establish_counter and establish_count on two consensus spans; both now use
establish_count (the establishCounter span-name constant is removed; the
establishCounter_ member variable is unchanged).
The consensus.proposal.receive and consensus.validation.receive spans record
the trust flag under proposal_trusted / validation_trusted (shared with the
peer spans of the same message type), not a bare `trusted`. Update the span
reference table to match.
State the rules so they stay consistent across code, collector, Tempo,
dashboards, and docs:
- Per-span-unique field -> bare name (the span name carries the domain).
- Same concept on more than one span -> ONE shared key, reused verbatim and
distinguished by span name, never tagged with the emitting workflow
(e.g. ledger_hash, full_validation, proposal_trusted/validation_trusted).
Defined once in the base SpanNames.h and re-exported by each domain header.
- Collision qualifier <domain>_<field> only to separate DIFFERENT concepts that
share a word, or the OTel-reserved status key (rpc_status, consensus_state).
- Dotted xrpl.<...> is reserved for resource attributes (xrpl.network.*).
Updates CONTRIBUTING.md (permanent home) and OpenTelemetryPlan §2.3.3.
Rule A silently missed a dotted span attribute (xrpl.ledger.hash) because of
two interacting bugs:
1. attr_keys_from_header resolved each constant via a flat global symbol table
keyed by bare name, so a later header defining a same-named constant (e.g.
consensus attr::ledgerHash = "ledger_hash") clobbered the base header's
attr::ledgerHash = "xrpl.ledger.hash", erasing the real dotted key from L1.
Now each constant is resolved against its own header (the global table only
seeds seg::/join() cross-file references); using-re-exports still resolve
globally.
2. derive_dotted_resource_keys allowlisted any dotted key declared in the base
SpanNames.h. Now it allowlists only the keys actually passed to
Resource::Create() in Telemetry.cpp (semconv service.* + the attr:: constants
set there, e.g. xrpl.network.*). A dotted key declared in a header but never
set as a resource attr is a Rule-A violation.
Adds 4 regression tests (collision, using-re-export, allowlist scope, brace
matching). No allowlist exception is added — the check now catches the
violation so the offending code can be fixed.
Forward-merge the OTel naming-check Rule C/D fixes and the phase 2-9 chain.
Conflict resolutions:
- 02-design-decisions.md TxQ Attributes: take the convention-correct markdown
table (txq_* keys) over the stale planned-attrs code block.
- SpanGuardFactory.cpp: keep the explanatory comment for the literal attribute
keys in the libxrpl test.
Rule C was reading docker/telemetry/tempo.yaml (the Tempo server config), which
has no filter tags, so it always SKIPped — L3 was silently unenforced. The
trace-search filter tags actually live in the Grafana datasource provisioning
file (docker/telemetry/grafana/provisioning/datasources/tempo.yaml) as
search.filters[].{tag,scope}. Point Rule C there (server file as fallback),
pair each tag with its scope, validate only span-scope tags against L1 (resource/
intrinsic tags like service.*/name/status/duration are exempt), and strip the
TraceQL span. prefix.
On phase-9 this turns "SKIP: C" into "OK: C: 24 tempo span-filter tags all in
L1" — L3 is now genuinely guarded. Adds a RuleCTempo test class (4 cases:
span-tag-not-in-L1 flagged, span-tags-pass, resource/intrinsic ignored, skip
when datasource absent). 83 tests total.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 9 surfaced two Rule D gaps (false positives, not data errors):
- TraceQL `span.<attr>` / `resource.<attr>` references: the bare attribute is
in L1, but the scope-prefixed form was flagged. Now strip the
span./resource./event./link/instrumentation_scope. prefix before the L1
lookup.
- Native OTel metric labels (e.g. `job_type`, `reason`) emitted by
MetricsRegistry are valid dashboard labels but are not span attributes. Add
an L6 source: parse `Add(.., {{"label", ...}})` instrument calls and accept
those label keys alongside L1 and builtins.
Verified against phase-9's real dashboards: 6 prior false positives -> 0.
79 tests (7 new for span-prefix stripping and metric-label extraction).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring phase-8 (log correlation) and the naming-check + BasicConfig fixes forward
into phase 9 (native-metrics gap fill). Phase 9 adds MetricsRegistry and
LedgerSpanNames.h.
Conflict resolution:
- presentation.md: took phase-9's restructured deck — sampling now lives in
Slide 8 (Head vs Tail) and the Phase 10/11/12 future-phase slides; phase-8's
older standalone sampling section is superseded (no content lost).
- ordering.txt/loops.txt: regenerated via generate.py.
Naming check: A/B/E/F/G green. Rule D reports 6 findings that are all check
gaps, not data errors (verified every dashboard label traces to code):
- 3 are TraceQL `span.<attr>` references (bare attr is in L1); Rule D should
strip the span./resource. scope prefix.
- 2 (`reason`, `job_type`) are native OTel metric labels emitted by
MetricsRegistry.cpp — a new source of truth the check does not yet model.
The Rule D enhancement lands on phase-1c and merges forward.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Forward-merge the phase-2/3 BasicConfig.h include-path fix to keep the chain
consistent. (phases 4+ already had the correct include; no file change here.)
Forward-merge the phase-2/3 BasicConfig.h include-path fix to keep the chain
consistent. (phases 4+ already had the correct include; no file change here.)
Forward-merge the BasicConfig.h include-path fix (basics -> config) so the
phase-3 build-test job compiles. phase-3 had the same broken include as phase-2.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
src/tests/libxrpl/telemetry/TelemetryConfig.cpp included
<xrpl/basics/BasicConfig.h>, but the header lives at
<xrpl/config/BasicConfig.h> (the phase-1b layout). This broke the build-test
job on phase-2 (and phase-3) across all platforms with a "file not found"
fatal error. The fix already exists on phase-4+; backport it to the branch
that introduced the test so each PR builds on its own.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring native-metrics (phase 7) and the naming-check work forward into phase 8
(log correlation).
Conflict resolution — tempo.yaml datasource provisioning (search.filters):
- Took phase-7's complete 30-filter list as the base (it is the comprehensive
set), preserved phase-8's tracesToLogs Loki-correlation block, and added
phase-8's two unique filters (tx-type, ledger-hash). Verified the result is
the full union (32 filters) with no filter dropped and no duplicate ids.
- Fixed a pre-existing dotted tag carried in from phase-7: the
consensus-ledger-id filter used `xrpl.consensus.ledger_id`; the code emits
the underscore key `consensus_ledger_id` (ConsensusSpanNames.h), so the
dotted form was a dead filter. Now corrected.
Every span-scope filter tag verified to exist in the L1 *SpanNames.h key set.
Naming check green; 72 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Rule D (dashboard PromQL labels must exist in L1) flagged `__name__` once the
phase-7 system-*.json dashboards started using `sum by (le, __name__)`.
`__name__` is the Prometheus reserved label for the metric name itself — a
builtin, not a span attribute. Add it to the builtin allowlist and cover it
with a test. (Earlier dashboards only used `__name__` inside `{__name__=~...}`
matchers, which the label regex did not extract, so this surfaced only now.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring the hardened OTel naming check (Rule E fix + 71 tests) and the phase 1-6
convention work forward into phase 7 (native metrics).
Conflict resolution:
- 05-configuration-reference.md: kept phase-7's native-OTLP metrics story
(server=otel, /v1/metrics) over phase-6's superseded StatsD scrape job, in
prose form (no code block).
- ordering.txt/loops.txt: regenerated via generate.py (not hand-edited).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring phase-4 forward into phase 5 (docs & deployment). Phase 5 owns the
operator runbook (docs/telemetry-runbook.md).
Conflict resolution:
- 05-configuration-reference.md: took phase-4's code-block-free prose for the
config-parser section; the mTLS options (tls_client_cert/tls_client_key)
remain documented in the §5.1.2 options table.
Phase-5-owned naming fix:
- docs/telemetry-runbook.md: converted 20 dotted xrpl.<domain>. attribute keys
in the Span Reference tables to the underscore convention (tx_hash, peer_id,
ledger_seq, consensus_mode/round/round_id/ledger_id, tx_id). Span NAMES stay
dotted (rule 5). `trusted` left as-is — verified it matches the code constant
ConsensusSpanNames.h (the proposal_trusted/validation_trusted split is a
separate, not-yet-applied code change).
End-to-end reconciliation: every span-attribute token in the runbook (L5) now
matches an L1 *SpanNames.h constant. The xrpl_<domain>_<field> tokens are the
spanmetrics-derived Prometheus labels (documented mapping), not span attrs.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three entries in Phase4_taskList.md annotated dotted attribute keys
(`xrpl.consensus.round`, `round_id`, `ledger_id`) as "(kept — rule 5)". That
is wrong: rule 5 keeps SPAN NAMES dotted, not attribute keys — the code emits
the underscore form (consensus_round, consensus_round_id, consensus_ledger_id)
per ConsensusSpanNames.h. Remove the false justification; the dotted form in
this task list remains as the illustrative-readability form documented in the
note at the top of the file.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The two "Spans Produced" tables in 06-implementation-phases.md (added by the
Phase 4 consensus work) listed span ATTRIBUTE keys in the dotted
xrpl.consensus.* / ledger.seq / mode.old form. Convert them to the underscore
convention, matching the authoritative ConsensusSpanNames.h constants
(consensus_round, ledger_seq, consensus_mode, consensus_round_id,
consensus_ledger_id, consensus_result, mode_old/new, agree_count/disagree_count,
etc.). Span NAMES in column 1 stay dotted (rule 5). Every key verified to exist
as a constant in ConsensusSpanNames.h / SpanNames.h.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring phase-3 forward into phase 4 (consensus tracing). Phase 4 introduces
ConsensusSpanNames.h (and already did the trusted→proposal/validation rename
and underscore-attr conversion in ffc197b914).
Conflict resolution (reviewed by a code-review agent before commit):
- 02-design-decisions.md §2.4 Consensus Attributes: merge-both — kept phase-3's
underscore table form AND folded in phase-4's richer "Phase 4a" attribute set
(round_id, ledger_id, trace_strategy, converge/establish/disputes counts,
agree/disagree counts, threshold_percent, consensus_result, mode_old/new),
each mapped to its authoritative ConsensusSpanNames.h constant. Dropped the
planned-but-unimplemented proposers_agreed/proposers_total (no code constant;
agree_count/disagree_count serve that role).
- SpanGuardFactory.cpp: kept phase-4's explanatory comment about why a libxrpl
test uses literal keys, plus the converted command/rpc_status keys.
Naming check passes (100 keys across 8 *SpanNames.h headers, including consensus).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring phase-2 forward into phase 3 (transaction tracing). Phase 3 introduces
TxSpanNames.h, TxQSpanNames.h, and TxApplySpanNames.h.
Conflict resolution:
- TxQ.cpp: kept phase-3's txq_span-based instrumentation (phase-2 had none).
Dropped the orphaned `NumberSO{... fixUniversalNumber}` line — develop's
#5962 (Retire fixUniversalNumber) removed that symbol repo-wide; the
conflict block had carried one stale copy that would not compile.
- 05/08/OpenTelemetryPlan.md: dropped the deleted 04-code-samples / POC_taskList
references (carried from phase-2), kept phase-3's new secure-OTel.md doc rows,
section, and Mermaid node/edge/style. Config code block -> prose; merged the
secure-OTel hardening pointer with the authoritative-config prose.
- Phase3_taskList.md: removed the "dotted keys for readability" note that came
from phase-2 — phase 3 already uses the underscore keys.
Reviewed by code-review agents: telemetry instrumentation intact, naming check
green (47 keys across 7 *SpanNames.h headers), no conflict markers.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Phase3/4/5 task lists show attribute keys in the older dotted
`xrpl.<domain>.<field>` form because that mirrors how a fully qualified
attribute reads in a Tempo trace view, which keeps the planning prose readable.
Add a note to each clarifying that the implemented keys follow the underscore
convention in CONTRIBUTING.md and that the *SpanNames.h constants (enforced by
the CI naming check) are the single source of truth — so the dotted form here
is illustrative, not the literal key.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
SpanGuardFactory.cpp set dotted "xrpl.rpc.command"/"xrpl.rpc.status" attribute
keys. Test files are exempt from the naming check's Rule F (they pass arbitrary
literals to exercise the API), so this slipped through, but the keys should
still illustrate the underscore convention. Use "command"/"rpc_status".
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring the hardened OTel naming check forward from phase-1c: unconditional
Rule F, test-file exemption, and the Rule H in-place-constant warning. The
check passes clean on phase 2 (24 keys across 4 *SpanNames.h headers including
PathFind; the SpanGuardFactory.cpp test is correctly exempt from Rule F).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three robustness fixes to check_otel_naming.py, all on phase-1c where the
script lives:
- Rule F now runs UNCONDITIONALLY. It is a purely syntactic check on the
call-sites and does not need the L1 key set, so code that calls
SpanGuard::span/setAttribute directly without ever defining a *SpanNames.h
is still caught (previously it was silently skipped when no header existed).
- Exempt test files from Rule F (tests pass arbitrary literal keys to exercise
the API). The call-site matcher now requires a SpanGuard/`.`/`->` receiver,
so std::span and bare declarations no longer false-positive.
- Add Rule H (warning, non-fatal): a namespace-qualified constant used at a
telemetry call-site but not defined in any *SpanNames.h is flagged, catching
constants defined in-place instead of in the proper header. Bare locals and
std:: names are not warned to avoid noise.
SpanGuard.h / Telemetry.h @code examples updated to reference constants that
exist on this branch. README documents the new behavior.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The script and its README live on phase-1c (where check_otel_naming.py was
introduced). The test-file Rule-F exemption was mistakenly applied here on
phase-2; revert to phase-1c's version verbatim. The exemption and further
script improvements will land on phase-1c and merge forward, keeping the
script's logic on the branch that owns it.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring the naming convention, code-sample cleanup, and CI naming check into
phase 2 (RPC tracing). Phase 2 introduces PathFindSpanNames.h.
Conflict resolution:
- 04-code-samples.md, POC_taskList.md: deletion wins.
- 02-design-decisions.md: took the convention-applied tables, but kept phase-2's
accurate PathFinding summary row (pathfind_fast/search_level/num_paths/...,
matching the implemented PathFindSpanNames.h).
- 05/08: took the code-block-free prose; kept phase-2's Phase2-5_taskList.md
index rows (dropping only the deleted POC row). Fixed stale setup_Telemetry/
make_Telemetry doc references to the code-correct setupTelemetry/makeTelemetry.
- Telemetry.h auto-merged to the constant-based @code examples.
check_otel_naming.py change: exempt test files from Rule F (tests pass
arbitrary literal keys to exercise the API). The check passes clean on the
merged tree (24 keys across 4 *SpanNames.h headers, including PathFind).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add Rule G to check_otel_naming.py: every span-attribute key must be
lower_snake_case (^[a-z][a-z0-9_]*$ per dot-separated segment). This catches
camelCase, UPPERCASE, and spaces in keys, which the structural (dotted) and
source (literal) rules did not. Document it in the script README and
CONTRIBUTING.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add check_otel_naming.py and wire it into on-pr.yml so every PR validates
that span-attribute names stay consistent across the code, collector, Tempo,
dashboards, and docs.
- The valid key set is derived dynamically from the *SpanNames.h constants and
the resource attributes the code registers in Telemetry.cpp — no hardcoded
allowlist to drift.
- Each rule is presence-gated: it runs only when the file it needs is in the
tree, so the check is correct whether telemetry changes land in one PR or
several (the collector/Tempo/dashboard/runbook layers arrive in later phases).
- Rule A flags dotted span-attribute keys; Rule F flags string-literal
attribute keys and span-name arguments (values may be runtime data).
- stdlib-only, mirroring the levelization check (bare `python`, no pip step).
- Telemetry.h / SpanGuard.h @code examples now use *SpanNames.h constants so
the strict literal check passes.
- CONTRIBUTING.md documents the check and how to run it locally.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring the phase-1a/1b naming-convention and code-sample cleanup into 1c.
Conflict resolution:
- 04-code-samples.md, POC_taskList.md: deletion wins.
- OpenTelemetryPlan docs (01/02/03/05): took the convention-applied,
code-block-free versions; verified no attribute category, table row, or
section header was lost (the differences were dotted->underscore renames).
- Telemetry.h: kept 1c's RpcSpanNames.h constant-based example
(rpc_span::attr::command) over the string literal.
- 31 non-telemetry files are clean develop carry-forward (identical to 1b).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring the phase-1a code-sample cleanup into phase-1b.
Phase 1a deleted the two code-sample docs (04-code-samples.md,
POC_taskList.md) and stripped C++/config code blocks from the remaining
plan docs, replacing them with prose summaries and pointers to the real
source/config files.
Conflict resolution:
- 04-code-samples.md, POC_taskList.md: deletion wins (phase-1b's SpanGuard
rewrites of these files are intentionally dropped).
- 03/05/OpenTelemetryPlan.md: took phase-1a's code-block-free prose, but
kept phase-1b's accurate descriptions of the real telemetry code
(DiscardFlag.h, FilteringSpanProcessor, SpanGuard factory methods, and
the corrected file-count/file-list tables).
- presentation.md left untouched.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bring the span attribute naming convention (phase 1a) into phase 1b.
Conflict resolution kept phase-1b's SpanGuard-based workflow and applied
the underscore naming convention to all non-code-sample text:
- Converted prose, tables, Mermaid labels, and TraceQL/PromQL query
references across the plan docs to the underscore form.
- Converted the two @code attribute-key examples in Telemetry.h
(command, tx_type).
- Left the code-sample files (04-code-samples.md, POC_taskList.md) and
03-implementation-strategy.md code blocks at the phase-1b version; the
code-sample docs are slated for removal on phase-1a.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Establish the single, authoritative naming convention for OpenTelemetry
span attribute keys so the code, collector, Tempo, dashboards, and docs
stay in sync.
- CONTRIBUTING.md: new "Telemetry span attribute naming" section under
the Style guide as the permanent, canonical home for the rules.
- OpenTelemetryPlan/02-design-decisions.md: new section 2.3.3 stating the
decided convention as design, and section 2.4 attribute schema realigned
to the underscore form (exact key spelling defers to the *SpanNames.h
constants).
- Sweep the remaining plan docs: convert dotted xrpl.<domain>.<field> span
attribute keys to the underscore form; leave span names and the
OTel-standard service.*/http.* and xrpl.network.* resource keys dotted.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The xrpl_tests target links only xrpl.libxrpl and cannot include the
xrpld-level ConsensusSpanNames.h / RpcSpanNames.h (CI build failure on
ubuntu-gcc-debug-arm64). Revert the test to underscore-form attribute
literals, restore the SpanNames.h include for seg::consensus, and drop
the now-unused tests.libxrpl -> xrpld.* levelization entries.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- tempo.yaml: align consensus filter tags with emitted keys
(consensus_mode, consensus_round, ledger_seq) instead of dotted form
- haveConsensus(): set span attributes before early-return paths so the
consensus.check span carries diagnostics even when consensus is not reached
- replace hardcoded consensus phase/result/vote literals with
ConsensusSpanNames.h val constants; add val::phaseOpen/Establish/Accepted
- ConsensusReceiveTracing.h: use canonical consensus::span constants instead
of duplicate inline detail:: names
- SpanGuardFactory test: use rpc_span / consensus::span constants now that
levelization permits the dependency
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
BasicConfig.h moved from xrpl/basics/ to xrpl/config/ on develop
(PR #7095 / reorg). Phase 1b's telemetry headers still referenced the
old path, breaking a fresh compile with "BasicConfig.h: No such file or
directory". Point both telemetry includes at the new location.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Head sampling is fixed at 1.0 and not configurable; remove the
sampling_ratio line from the benchmark, full-validation, and validator
config-template sample blocks.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Head sampling is fixed at 1.0 and not configurable. Remove the
sampling_ratio line and its stale tuning comment; point volume reduction
at collector-side tail sampling.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Head sampling is fixed at 1.0 and not configurable. Drop sampling_ratio
from the docker sample/test configs and the data-collection reference;
rewrite the span-metric sampling caveat and volume guidance to point at
collector-side tail sampling.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Head sampling is now fixed at 1.0 and not configurable. Drop the
sampling_ratio config row and redirect volume-reduction guidance to
collector-side tail sampling. trace_peer=0 stays as the high-volume opt-out.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The telemetry test module compiles ValidationTracker.cpp (from src/xrpld/)
into xrpl_tests and includes <xrpld/telemetry/ValidationTracker.h>. The
single-binary test scheme dropped the src/ include directory that the old
per-module target carried, breaking the build. Restore it on the telemetry
branch of the test CMakeLists.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
sampling_ratio is no longer parsed (head sampling fixed at 1.0), so remove
the parse/clamp assertions and the sampling_ratio_clamped test. Update the
default-tracePeer expectations to TRUE to match the enabled-by-default change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Document that head sampling is intentionally fixed at 100% and no longer
exposes a sampling_ratio config knob. A per-node ratio let nodes make
divergent keep/drop decisions for the same distributed trace, producing
broken/partial traces; pinning at 1.0 with a ParentBased sampler keeps
decisions coherent across the network. Volume reduction is delegated to
collector-side tail sampling.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Update the docker/telemetry sample config and testing guide to set
trace_peer=1, matching the new on-by-default behavior and the other
trace categories already enabled in these all-categories sample profiles.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Update the Peer Spans note, the known-gaps caveat, and the trace category
toggle table to show trace_peer enabled by default. The Production Setup
example keeps trace_peer=0 as a deliberate high-volume opt-out alongside
sampling_ratio=0.01.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reflect the new on-by-default behavior in the config defaults table. The
volume-reduction guidance (High memory usage, Production mainnet tuning)
intentionally keeps trace_peer=0 as an opt-out recommendation for
high-traffic deployments.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Flip the tracePeer Setup default and the trace_peer config parser
default from off to on, and update the example config and build doc to
match. Peer spans record only peer_id (a node-local numeric connection
id) plus trust/ledger metadata — no IP addresses or public keys — so
the privacy concern behind disabling it does not apply. The high-volume
characteristic is retained in the docs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Flip trace_peer default false->true across the Phase-1a plan docs and
correct the rationale: peer spans record only peer_id (numeric local
connection id) plus trust/ledger metadata, never IP addresses or public
keys, so the 'includes addresses' caveat was inaccurate. The high-volume
note is retained.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The txq.enqueue span sets tx_hash, tx_type, and txq_status on every
code path (TxQ.cpp:746/748/751), but fee_level_paid and
required_fee_level are set only on the fee-evaluated path
(TxQ.cpp:895-898), which is reached after the rejected and
applied_direct early exits. They are therefore not guaranteed on every
txq.enqueue span, so requiring them caused the validation harness to
fail whenever a txq.enqueue span took an early-exit path.
Remove the two conditional attributes from required_attributes and
document why in the span note.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Several close-time panels showed raw codes/booleans and unreadable
time-bar-charts. Relabel and re-visualize them so each reads clearly:
- Close-Time Proposal Spread (was "Close Time Bin Distribution"):
corrected a wrong description (it claimed per-node proposals; the
metric is the number of DISTINCT close-time positions peers proposed
per round, rawCloseTimes.peers.size()). Converted the time-barchart
(unreadable timestamp axis) to a horizontal bar gauge summing rounds
per distinct-position count.
- Consensus Outcome Distribution: renameByRegex maps the raw
consensus_state codes to human labels (yes->Agreed,
moved_on->Moved On (partial), expired->Expired (timeout), no->No
Consensus); value mappings alone do not relabel pie legends.
- Close-Time Agreement Rate (was "Close Time Agreement"): legend
relabelled from "close_time_correct=true/false" to Agreed/Disagreed.
- Close-Time Resolution Change (was "Close Time Resolution Direction"):
converted to a bar gauge; increased/decreased/unchanged relabelled to
Coarser (more disagreement) / Finer (better agreement) / Steady.
All four verified by rendering the panels to PNG via the Grafana
image renderer.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
"Close Time: Raw Proposals" and "Close Time: Effective / Quantized"
plotted the absolute close time (Ripple-epoch seconds), which is an
ever-rising line with no analytical value. There is no clean way to make
them useful: the values are Ripple-epoch seconds (not Unix ms, so date
units misrender), TraceQL metrics cannot do the epoch offset or an
inter-ledger-gap subtraction in-query, and Grafana calculateField
transforms break on these grouped Tempo metric frames (verified by
render: "No data").
Remove both. The useful consensus-timing signals are already covered:
time-to-consensus (round_time_ms), rounds per ledger (establish_count),
previous round time, and the close-time vote-bins / resolution-direction
/ bin-distribution panels remain. Re-tile and re-id (22 panels).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
With validation now passing 133/133, the only remaining job failure was the
regression gate flagging 4 timing "regressions". Two compounding causes:
1. Stale baseline: the committed baseline was captured (2026-04-24) under the
old, lighter workload — before the new txq-burst phase (60 TPS) existed. The
heavier per-ledger work genuinely raises ledger.build / tx.apply /
ledger.validate / acceptLedger timings, so every run regressed against it.
Refreshed the baseline from the latest CI-measured timings (same workload).
2. Histogram quantization: SpanMetrics latency buckets are
[1,5,10,25,...]ms, so a sub-millisecond quantile near a low-end boundary can
jump a full bucket (1ms->5ms) between runs with no real change. The old
absolute bounds (2-5ms) were narrower than one bucket width, so that jitter
tripped the gate. Widened the default span bounds to 10-15ms (~2 low-end
buckets) and pct to 50%, and the job_queue running bound to 20ms, to tolerate
quantization noise while still catching genuine multi-bucket regressions. The
consensus.* overrides (tight pct, large abs) are unchanged.
The refreshed baseline also picks up real rpc.ws_message timings (previously null
under the phantom rpc.request key).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Number of rounds is an integer, but avg_over_time(establish_count)
produced fractional values (2.1). Switch the Rounds panel to
count_over_time() by (span.establish_count): one integer series per
round count (1/2/3...), showing how many ledgers needed that many
establish rounds — the meaningful distribution, inherently integer
(decimals=0).
Apply dashboard rule 9: panels with a right-side table legend take full
width. Widen "Consensus Rounds per Ledger" and "Consensus Outcome
Distribution" to w=24 and re-tile the dashboard.
Verified via the Grafana proxy: rounds=2 dominates (~11-12 ledgers),
rounds=3 occasional.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 9 introduces the ledger.acquire span (InboundLedger fetch) that phases 7-8
do not have, so the forward-merged 09-data-collection-reference inventory is
extended here:
- §1.1: add ledger.acquire to the Ledger span table.
- §1.2: add its attributes (acquire_reason, timeouts, peer_count, outcome) and
note it also sets ledger_seq; bump the span count.
Also fix two stale StatsD metric references in the Peer Quality dashboard
(xrpld-peer-quality.json): rippled_Peer_Finder_Active_{Inbound,Outbound}_Peers
-> xrpld_Peer_Finder_* to match the xrpld_ metric prefix the rest of the stack uses.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The consensus duration panels plotted span wall-clock
(traces_span_metrics_duration_milliseconds), which is ~3-8 ms of
instrumentation overhead, not the real consensus time (~3000 ms). And
the close-time value panels plotted an ever-rising absolute epoch line.
Rework them to answer the actual operational questions, all from
attributes that already exist on the consensus spans:
- Time to Reach Consensus (p50/p95) and Average Time to Reach Consensus:
round_time_ms on consensus.accept — the wall-clock to agree a ledger.
- Consensus Rounds per Ledger (Establish Count): avg and max of
establish_count on consensus.establish — how many proposal rounds it
took to converge (1 = first proposal).
- Previous Round Time per Ledger: previous_round_time_ms on
consensus.round.
Reorder the dashboard into an investigation flow: health/throughput ->
time-to-consensus and rounds -> ledger close/apply timing -> close-time
detail -> failures/mode/mismatch. Assign stable sequential panel ids.
Verified each query returns data via the Grafana datasource proxy
(p95 ~4096 ms, avg ~2825 ms, rounds ~2).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The §1 span and attribute inventory had regressed to an older 16-span snapshot
that uses the pre-2026-05-13 dotted attribute keys, while phase-7's code emits
~36 spans with bare/underscore attribute keys. The §Data Flow Overview and §2
System Metrics sections (native OTLP transport — phase-7's migration) were already
correct and are left unchanged.
- §1.1: expand the span inventory to the full surface — add gRPC (grpc.<MethodName>),
TxQ (txq.*), PathFind (pathfind.*), and the full consensus set (round/phase.open/
establish/update_positions/check/mode_change/proposal.receive/validation.receive).
Fix the phantom rpc.request -> rpc.http_request, add rpc.ws_upgrade. No grpc.request,
no pathfind.rank, no ledger.acquire (the latter is added in phase-9, not yet present here).
- §1.2: convert every span-attribute key from dotted xrpl.<domain>.<field> to the
bare/underscore form. The sole span-attr dotted exception is xrpl.ledger.hash on
peer.validation.receive (shared constant); consensus.validation.send uses bare
ledger_hash. Resource attrs xrpl.network.id/type stay dotted. Fix tx_count/tx_failed
placement (on tx.apply, not ledger.build). Add attribute tables for the new families.
- §1.3: list the full set of spanmetrics dimension labels (bare keys, from the collector
config) instead of the stale xrpl_rpc_command-style names.
- §4/§5: convert Tempo TraceQL and PromQL examples to the bare attribute/label forms.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The print-env CI fix let the Telemetry Stack Validation job build and run the
workload harness end-to-end for the first time. It reported 129/136 checks
passing; this commit fixes the 7 real failures plus a latent regression-gate bug.
Validation-suite fixes (verified against the CI run's actual emission + live node):
- expected_metrics.json: the beast::insight job-depth gauge is `xrpld_jobq_job_count`,
not `xrpld_job_count` (the latter is a Phase 9 OTel counter). Reverted the prior
rename. Removed the statsd_histograms block (`xrpld_rpc_time`/`xrpld_rpc_size`):
these RPC timers do not emit under the WS workload (0 series in CI).
- expected_spans.json: `tx_status` is only set on suppressed/known-bad receives, so
it is no longer a required attribute of every `tx.receive`. Marked `pathfind.compute`
and `pathfind.discover` optional and the `pathfind.request -> pathfind.compute`
hierarchy as skip — the self-to-self XRP probe returns before computing paths in a
fresh cluster with no liquidity, so only `pathfind.request` fires.
Regression-gate bug (telemetry-validation.yml "Print regression summary"):
- `jq -e` exits non-zero when its filter result is boolean false — the normal case
for a populated (non-placeholder) baseline — which was misreported as
"Failed to parse baseline JSON" and failed the job. Dropped `-e` (kept `-r`) so a
non-zero exit genuinely means malformed JSON.
The optional-span handling and regression comparison both worked correctly in the
CI run (txq.* / pathfind.update_all skipped-when-absent, 0 regressions detected).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Apply the dashboard guidelines to the five close-time panels:
- Axis labels (Title Case) on every panel: "Close Time (Ripple
Seconds)" for the value panels, "Count / Milliseconds" for vote
bins/resolution, "Rounds in Window" for the count panels.
- Human-readable legends with the dimension in brackets per the legend
convention: "Raw Close Time [{{resource.service.instance.id}}]",
"Effective Close Time [...]", "Resolution Direction
[{{span.resolution_direction}}]", "{{span.close_time_vote_bins}} Vote
Bins" — replacing the bare label tokens.
- Unit "none" (plain number): the close-time values are Ripple-epoch
seconds and TraceQL metrics cannot offset them to a wall-clock unit,
and the others are counts/ms on a shared axis.
Verified rendered values against raw spans: close times ~833,998,8xx,
resolution 10000 ms, vote bins 1/2/3 — all correct.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
TraceQL metrics queries default to a 3h max range
(query_frontend.metrics.max_duration), so a dashboard set to a longer
window failed with "range ... exceeds 3h0m0s". Add a query_frontend
block raising it to 168h, matching the search max_duration, so the
consensus close-time panels work at 6h/12h/24h ranges.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Raw Proposals and Effective/Quantized panels rendered nameless
series: their legendFormat used {{service.instance.id}}, but the
TraceQL metrics query groups by resource.service.instance.id and Tempo
returns that full key as the series label. The legend token did not
match any label, so each series showed blank. Use the matching
{{resource.service.instance.id}} token.
Verified via the Grafana datasource proxy that all six close-time panels
now return correctly-labelled series.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Raw Proposals and Effective/Quantized panels showed wrong values
(e.g. 759M, 852M, even 0) against a true value of ~834M. Cause:
quantile_over_time bucketizes into an exponential histogram tuned for
duration distributions, so it cannot represent large absolute integers
(Ripple-epoch seconds) accurately.
Switch both panels to avg_over_time, which returns the correct value
(verified ~833,996,7xx matching the raw span attribute). Average is also
the semantically right aggregation here: close time is a single agreed
value per consensus round, not a latency distribution, so a median was
never meaningful.
Set the unit to none rather than seconds: the value is Ripple-epoch
seconds (Unix = value + 946684800) and TraceQL metrics cannot do the
offset arithmetic in-query, so a duration unit would misrender it.
Clarify in the description that the absolute level tracks wall-clock and
the useful signal is per-node spread / raw-vs-effective gap.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The five Close Time panels still rendered "No Data" after the metrics
rewrite. Root cause: each query carried
`span.close_time_correct=~"$close_time_correct"`, but close_time_correct
is a boolean span attribute and TraceQL's regex match (=~) against a
bool matches nothing in a metrics query, so every panel returned an
empty series set (HTTP 200, {"series":[]}).
Remove that filter clause. The panels do not break down by
close_time_correct, so dropping it restores data without losing any
dimension. The $node filter (a string attribute) is unaffected and
stays.
Verified via the Grafana datasource proxy that all six targets now
return series.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Wire the two previously-registered-but-never-incremented validation
counters to ValidationTracker's gross lifetime tallies, exported as
monotonic ObservableCounters. New gross atomics count each ledger once at
first classification and are never adjusted on late repair, keeping the
_total counters monotonic and additive (agreements_total + missed_total ==
ledgers reconciled); the repair-aware windowed view stays on the existing
xrpld_validation_agreement gauge. The validator-health dashboard panels
that already query these names now render data instead of "No data".
Also de-stale 09-data-collection-reference.md: §5b documented flat metric
names (xrpld_cache_SLE_hit_rate, ...) that the code never emits — it emits
labeled gauges (xrpld_cache_metrics{metric="SLE_hit_rate"}). Replace the
stale flat-name tables with a pointer to the canonical labeled section,
reconcile the contradictory headline counts, and correct xrpld_job_count
to its real exported name xrpld_jobq_job_count.
Adds two GTests asserting gross tallies stay frozen on repair while net
totals move, plus the additive invariant.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Phase 10 validation harness had drifted from the code's recording surface
and the telemetry-validation CI job was failing before it could build.
CI fix (telemetry-validation.yml):
- Replace nonexistent local action ./.github/actions/print-env with the remote
XRPLF/actions/print-build-env (the build-xrpld job failed in 56s on this).
- Sync prepare-runner and upload-artifact action SHAs to the canonical workflow.
Recording-surface reconciliation (docker/telemetry/workload/):
- Migrate span attributes from dotted xrpl.<domain>.<field> to the bare/underscore
form introduced by the 2026-05-13 span-attr naming redesign (tx_hash, peer_id,
ledger_seq, consensus_mode, consensus_round, full_validation, quorum, ...).
Dotted xrpl.ledger.hash is retained only on peer.validation.receive (shared
constant), while consensus.validation.send uses bare ledger_hash.
- Fix attribute placement: tx.apply carries tx_count/tx_failed (not ledger_seq);
ledger.build carries ledger_seq/close_* (not tx_count/tx_failed).
- Replace the phantom rpc.request span with the real WS root rpc.ws_message; drop
the never-emitted duration_ms; rebuild the parent-child map accordingly.
- Add the new spans the code emits: apply-pipeline stage spans
(tx.preflight/preclaim/transactor with stage/tx_type/ter_result), txq.*,
consensus sub-spans (round/establish/update_positions/check/phase.open),
ledger.acquire, grpc.*, pathfind.*. Conditional spans are marked optional so
they are skipped (not failed) when the workload does not exercise them.
- validate_telemetry.py: service.name and Loki job label rippled -> xrpld; fix
PARITY_SPAN_ATTRS (rename the 4 real attrs, drop the 3 that are metrics not span
attrs); add optional-span handling that skips missing optional spans while still
validating attributes when present.
- expected_metrics.json: rippled_ -> xrpld_ on all beast::insight/overlay metrics,
xrpld_job_count, the 15 on-disk xrpld-* dashboard UIDs, and the real bare
spanmetrics dimension labels.
- regression-metrics.json + baseline-timings.json: rpc.request -> rpc.ws_message.
Metrics pipeline fix:
- Switch node [insight] config from server=statsd/prefix=rippled to server=otel +
/v1/metrics endpoint + prefix=xrpld across run-full-validation.sh,
xrpld-validator.cfg.template, benchmark.sh and the workload compose. The
collector has no StatsD receiver, so system metrics only reach Prometheus over
OTLP.
Synthetic load for new spans:
- Add ripple_path_find to the RPC load generator (drives pathfind.* spans).
- Add a high-TPS txq-burst workload phase to force fee escalation (drives txq.*).
All facts verified against the *SpanNames.h headers and a live xrpld node +
collector (Tempo service.name=xrpld, tx.preflight attrs [stage,ter_result,tx_type],
279 xrpld_ Prometheus metrics and zero rippled_).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The five Close Time panels (Raw Proposals, Effective/Quantized, Vote
Bins & Resolution, Resolution Direction, Bin Distribution) rendered
empty: they used TraceQL `| select(attr)`, which returns a trace list
that a timeseries/barchart panel cannot plot.
Enable TraceQL metrics in Tempo and rewrite the panels to use it:
- tempo.yaml: add the local-blocks processor to the metrics generator so
recent blocks are queryable via /api/metrics/query_range. Set
filter_server_spans=false because the consensus spans are
SPAN_KIND_INTERNAL (the default keeps only server spans, so attribute
aggregations over internal spans returned nothing), and
flush_to_storage=true with a traces_storage path so query_range can
read the flushed blocks.
- consensus-health.json: replace each panel's select() with a metrics
query — quantile_over_time on the integer close-time attributes,
avg_over_time for vote bins / resolution, and count_over_time by the
resolution_direction and vote-bin dimensions. Set the raw/effective
panels' unit to seconds (the values are Ripple-epoch seconds, which
dateTimeFromNow rendered with the wrong epoch).
Verified the query forms compile and return series against live internal
spans; the close-time series populate once the node reaches full sync.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Current Job Latency gauge sat at the bottom of the Job Queue
Analysis dashboard; per the dashboard guideline gauges belong at the
top. Move it to the first row and reflow the remaining panels below it.
Also assign explicit sequential panel ids so deep links stay stable.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Correct the width rule from the previous layout commit. Full width
(w=24) is now applied ONLY to timeseries panels whose legend is a
right-side table, since those legends need the horizontal room. Panels
with default/bottom legends, pie charts, and the heatmap return to half
width. This narrows "Transaction Receive vs Suppressed" and "TxQ
Enqueue Rate by Transaction Type", which were wrongly widened.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Make transaction-overview deep-links stable and improve readability:
- Assign explicit sequential panel ids (1..20) so viewPanel=panel-N
URLs stay pinned to the same chart across edits. Previously ids were
unset and Grafana auto-assigned them by array position, so any
reorder silently repointed bookmarks.
- Move the single-value stat panel (Transaction Apply Failed Rate) to
the top row.
- Lay out in three topic sections (Processing, Apply Pipeline, Queue).
Within each, timeseries with a breakdown dimension (tx_type, stage,
ter_result, suppressed) take full width so their right-side table
legends are readable; single-series panels, pie charts, and the
heatmap stay half-width and pair up.
All six template variables already default to All (includeAll + multi);
no change needed there.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Apply the dashboard legend convention across all panels now that the
P50 series have been removed (P95-only):
- Drop the redundant "P95 " / "P50 " prefix; the panel title already
states the percentile.
- Put every filter/dimension value inside [] comma-separated, ending
with exported_instance, e.g. "AMMDeposit [Preclaim, xrpld-mainnet]".
- Add exported_instance to the by() clause and legend of the three
panels that filtered on $node but omitted it (Transaction Rate by
Type, Transaction Results by Type, TxQ Accept Status), so per-node
series are produced.
- Title-case the stage value for display via label_replace in the four
apply-pipeline panels; the span attribute stays lowercase
(preflight/preclaim/apply) since legendFormat cannot change case.
- Cap tooltip maxHeight at 500 on every panel.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The existing apply-pipeline panels show latency by stage (all types
combined) or by type (single span). Neither answers "for a given
transaction type, which stage dominates its latency". Add a p95 panel
grouped by both tx_type and stage, filterable via the $tx_type and
$stage variables. Both dimensions already exist in spanmetrics, so no
collector change is needed. Reflow the section so the full-width
failure panel sits below the new full-width panel.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Wire the apply-pipeline stage spans (tx.preflight, tx.preclaim,
tx.transactor) added on phase-3 through the observability stack so the
spanmetrics connector produces per-stage RED metrics without any native
instruments.
- collector: add the `stage` dimension to the spanmetrics connector so
the three stages split into separate metric series (3 bounded values).
- dashboard: add a "Tx Apply Pipeline" section to transaction-overview
with rate, p95 latency, and failure-rate panels grouped by stage, plus
a `stage` template variable. Panels follow the existing config (node
filter, exported_instance legends, Title Case, axis labels).
- The failure panel filters ter_result != tesSUCCESS rather than span
status, because a failing ter code completes the span normally — only
thrown exceptions set an error status. This matches the existing
"Transaction Results by Type" panel convention.
- docs: document the spans, attributes, and stage dimension in the data
collection reference and runbook, including the sampling caveat that
span-derived metrics inherit tracer head-sampling and undercount at
sampling_ratio < 1.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The tx.transactor span covered only the apply stage; preflight and
preclaim had no telemetry, so a transaction that hard-failed those
stages produced no apply-pipeline span and per-stage latency/failure
was invisible.
Add tx.preflight and tx.preclaim spans in applySteps.cpp via a
makeStageSpan() helper using SpanGuard::hashSpan, so all three stages
share a deterministic trace_id derived from txID[0:16] even though they
run sequentially and often cross-thread. Each span carries stage,
tx_type, and ter_result; exceptions are recorded as tefEXCEPTION before
the public wrappers map them. The type lookup is guarded behind the
span-active check so it costs nothing when tracing is off.
Add a stage="apply" attribute to the tx.transactor span and move its
three hardcoded attribute strings to a new library-safe header
include/xrpl/tx/detail/TxApplySpanNames.h, which mirrors the daemon-side
TxSpanNames.h strings so the collector spanmetrics connector aggregates
both span sets under one dimension set.
A constants-contract test pins the span-name, attribute-key, and
stage-value strings; span content stays covered by the docker
integration test, as the rest of the telemetry suite is.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reduced from 30 to 7 filters: service.instance.id, name, status, command,
tx_hash, tx_type, ledger_hash. Full attribute inventory is in
OpenTelemetryPlan/09-data-collection-reference.md §4; TraceQL autocomplete
covers the rest.
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
From the code-review pass:
- transaction-overview.json: the tx.process and tx.transactor latency-by-type
panels used lowercase legends (p95/p50) without the per-node dimension. Use
Title Case (P95/P50), add exported_instance to the by() clause, and include
[{{exported_instance}}] in the legend, per the dashboard legend convention.
- consensus-health.json: panel descriptions still referenced the old dotted
attribute names (xrpl.consensus.mode, xrpl.ledger.seq) after the A1 rename;
update them to the bare emitted names (consensus_mode, ledger_seq).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
handleMismatch() recorded the "consensus_txset" reason but then fell through
to the transaction-level comparison, which also recorded a reason
("same_txset_diff_result" / "different_txset"). A single mismatch with
disagreeing consensus tx-set hashes therefore incremented
xrpld_ledger_history_mismatch_total twice across two reason labels, so the
sum over reason exceeded the real mismatch count.
The consensus tx-set hash disagreement is the root cause; return after
recording it so each mismatch contributes exactly one reason.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The A1 fix (xrpl_consensus_mode -> consensus_mode) was applied on phase-6, but
the phase-8->phase-9 merge conflict resolution for consensus-health.json took
phase-9's pre-fix panel base, silently reintroducing all 11 stale
xrpl_consensus_mode label references (the spanmetrics label that is never
populated — see the original A1 commit).
Re-apply the label fix on phase-9: xrpl_consensus_mode -> consensus_mode in
every panel expr, legendFormat, and the $consensus_mode template variable's
label_values() query. The Grafana variable name $consensus_mode is unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Visualise the metrics added in this series:
- consensus-health: "Ledger History Mismatch Rate by Reason"
(xrpld_ledger_history_mismatch_total by reason — fork diagnostics)
- fee-market: "Queue Abandonment Rate (Expired)" and "Queue Admission
Rejections (Dropped)" (xrpld_txq_expired_total / dropped_total)
- peer-network: "Reduce-Relay Peer Selection" and "Reduce-Relay Missing-Tx
Frequency" (xrpld_reduce_relay_metrics)
- system-node-health: "Ledger Acquire Duration" and "Ledger Acquire Rate by
Outcome" (ledger.acquire span)
otel-collector-config.yaml: add outcome and acquire_reason spanmetrics
dimensions so the ledger.acquire outcome breakdown populates.
All panels follow the existing template: $node filter, exported_instance in
legends, Title Case, axis labels.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add the new synchronous counters (ledger_history_mismatch_total{reason},
txq_expired_total, txq_dropped_total{reason}) and the reduce-relay observable
gauge to the ASCII ownership diagram in the MetricsRegistry header so the
documented instrument inventory matches the code.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The transaction reduce-relay subsystem (selected vs suppressed peers,
feature-disabled peers, missing-tx frequency) was computed in OverlayImpl's
TxMetrics but only surfaced via the get_counts JSON RPC — invisible to
Prometheus/Grafana, despite being the central efficiency KPI for the feature.
Add an observable gauge xrpld_reduce_relay_metrics{metric} that reads
Overlay::txMetrics() and parses its rolling-average fields:
- selected_peers (txr_selected_cnt)
- suppressed_peers (txr_suppressed_cnt)
- not_enabled_peers (txr_not_enabled_cnt)
- missing_tx_freq (txr_missing_tx_freq)
The JSON values are decimal strings (std::to_string), parsed via std::stoll —
the same JSON-reading pattern as registerNodeStoreGauge. No new Overlay
accessor or core-interface change required.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
InboundLedger drives ledger back-fill and fork recovery with timeout/retry
logic (kLedgerTimeoutRetriesMax = 6), but emitted only a global ledger_fetches
counter — sync/recovery cost was a telemetry blind spot.
Add a ledger.acquire span that wraps the acquisition lifecycle:
- Started in InboundLedger::init() with ledger_seq and acquire_reason
(history / consensus / generic, mirroring InboundLedger::Reason).
- Finalized in InboundLedger::done() with outcome (complete / failed),
timeouts, and peer_count, then reset so the span duration is exported.
Held as a std::optional<SpanGuard> member (same pattern as RCLConsensus
roundSpan_). New op/attr/val constants added to LedgerSpanNames.h. Compiles to
a no-op when telemetry is disabled via the SpanGuard fallback.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The transaction queue had no metric for demand that leaves or never enters the
queue, so fee-underpayment abandonment and admission-control rejection were
invisible (distinct from jq_trans_overflow, which is the job queue).
Add two synchronous counters via MetricsRegistry:
- xrpld_txq_expired_total — incremented in TxQ::processClosedLedger() for each
queued transaction removed because its LastLedgerSequence passed (submitters
who under-bid the escalating fee and were never included).
- xrpld_txq_dropped_total{reason} — incremented in TxQ::apply() at the
queue-full admission-control returns (reason="queue_full").
Both reach MetricsRegistry via the Application& parameter already passed to
these methods; calls are null-guarded so they no-op when telemetry is disabled.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
LedgerHistory::handleMismatch() already classifies a built-vs-validated ledger
mismatch (prior ledger, close time, consensus tx set, same/different tx set),
but only bumped a single untyped beast::insight counter — the reason was
dropped. Fork diagnosis was therefore a log-grep exercise.
Add a labeled OTel counter so the mismatch reason is a queryable time series:
- MetricsRegistry: new ledgerHistoryMismatchCounter_ + incrementLedgerHistoryMismatch(reason)
- LedgerHistory: record one reason per classification branch (unknown,
prior_ledger, close_time, consensus_txset, same_txset_diff_result,
different_txset). Reaches MetricsRegistry via the existing app_ reference.
The existing beast::insight mismatchCounter_ is left intact.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The xrpld_peer_quality{metric="peers_insane_count"} gauge was hardcoded to 0.0
with a TODO, leaving the "Insane/Diverged Peers" panel permanently empty.
PeerImp::json() already exposes the peer's tracking state via the "track"
field (set to "diverged" when tracking_ == Tracking::Diverged). The peer-quality
callback already iterates peer->json() for latency and version, so count peers
whose "track" field equals "diverged" in the same loop — no change to the
abstract Peer interface required.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A phase-8->phase-9 merge (a675897aaf) duplicated the "Consensus Outcome
Distribution" and "Consensus Failures Over Time" panels: both appeared twice
with byte-identical queries (verified ignoring gridPos). The pair existed once
on phase-6/7/8 and became two on phase-9 only, so the duplication originated
in phase-9's own merge history.
Remove the second (lower) copy of each and re-stack panel y-positions with no
gaps. The single retained copy keeps the original y=64 row.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Both metrics are already emitted and live in Prometheus but were not fully
visualised.
- Fee Market (xrpld-fee-market.json): "Load Factor Attribution (Stacked
Components)" — stacks load_factor_fee_escalation / fee_queue / local / net /
cluster so an operator can see which component drives the effective fee. The
existing panels showed the aggregate only.
- Validator Health (xrpld-validator-health.json): "Agreement % (7d)" and
"Agreements vs Missed (7d)" — the xrpld_validation_agreement gauge already
observes agreement_pct_7d / agreements_7d / missed_7d, but the dashboard only
plotted 1h and 24h windows.
Panels follow the existing template: $node filter, exported_instance in legends,
Title Case, axis labels.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The grpc.{Method} spans (GRPCServer.cpp) and pathfind.* spans (PathRequest.cpp)
are emitted but had no dashboard coverage. The existing RPC & Pathfinding
dashboard only plotted StatsD timers. Add span-derived rows:
- gRPC Request Rate by Method (grpc.* by method)
- gRPC Latency P95 by Method
- gRPC Error Rate by Status (by grpc_status)
- Pathfinding Compute Duration (pathfind.compute p95/p50)
- Pathfinding Request & Discovery Rate (pathfind.request / pathfind.discover)
otel-collector-config.yaml: add method, grpc_role, grpc_status spanmetrics
dimensions (bounded value sets). Add a $grpc_method template variable so the
gRPC panels can be filtered by method, consistent with the dashboard filter
conventions.
Note: these spans populate only when the node serves gRPC / pathfinding
traffic; they are correct but not exercised by the current health-check
workload (they will be covered by the Phase 10 workload generator).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The consensus state-machine and TxQ lifecycle spans are emitted by the code
and present in Prometheus, but no panel visualised them. Add panels keyed on
those span_names (verified live) plus the low-cardinality dimensions needed to
break them down.
Consensus Health (consensus-health.json) — new rows:
- Consensus Round Duration (full round, p95/p50, mode-filterable)
- Consensus Phase Duration (open vs establish breakdown)
- Position Update Duration (update_positions p95/p50)
- Consensus Stall Rate (consensus.check by consensus_stalled)
- Consensus Mode-Change Rate by Target Mode (mode_change by mode_new)
Transaction Overview (transaction-overview.json) — new rows:
- TxQ Enqueue Rate by Transaction Type (txq.enqueue by tx_type)
- Queue Bypass Ratio (txq.apply_direct vs txq.enqueue)
- Queue Accept (Drain) Duration per Ledger (txq.accept p95/p50)
- Queue Cleanup Rate (txq.cleanup expired entries)
otel-collector-config.yaml — add spanmetrics dimensions for the lifecycle
breakdowns: mode_new, consensus_stalled, consensus_phase, consensus_result
(all bounded value sets, safe as Prometheus labels).
All new panels follow the existing dashboard template: $node filter,
exported_instance in every legend, Title Case, axis labels, row layout.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The spanmetrics connector dimension was `xrpl.consensus.mode`, but the code
emits the span attribute under the bare key `consensus_mode` (matching every
other dimension after the Phase 6 rename). The mismatch left the
`xrpl_consensus_mode` Prometheus label empty, so the Consensus Health
"Consensus Mode Over Time" panel and the `$consensus_mode` template variable
(which filters every panel) matched no live series.
- otel-collector-config.yaml: dimension `xrpl.consensus.mode` -> `consensus_mode`
- consensus-health.json: 11 label refs `xrpl_consensus_mode` -> `consensus_mode`
(the `$consensus_mode` Grafana variable name is unchanged)
- telemetry-runbook.md: refresh the stale spanmetrics label table to the bare
names actually emitted (command/rpc_status/consensus_mode/local/
proposal_trusted/validation_trusted), fix dotted->bare attribute names in
span tables and TraceQL examples (tx_hash, ledger_seq, consensus_round_id,
consensus_ledger_id, consensus_round, tx_id event attr), correct the
consensus_round_id query to int (not quoted string), and fix the
load_type value query ("exception_rpc" -> "exceptioned RPC").
Verified against the live stack: Tempo span tags confirm bare attribute keys
(consensus_mode, ledger_seq, tx_hash, ...); the populated xrpl_consensus_mode
series in Prometheus is stale retained data from an older build.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
beast::insight metrics exported via OTLP carried no exported_instance
label because [insight] omitted service_instance_id (only [telemetry]
set it). Every system-* dashboard filters insight metrics with
exported_instance=~"$node", and the $node template variable is sourced
from label_values(..., exported_instance) — so with the label absent,
$node was empty and all insight-backed panels showed no data.
Add service_instance_id to [insight] in both telemetry configs, matching
the [telemetry] value (xrpld-mainnet / xrpld-devnet). CollectorManager
already reads this key and passes it to OTelCollector, which sets the
service.instance.id resource attribute.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The phase-9 NodeStore I/O totals, write-load/read-queue, read-threads,
and object instance-count panels rendered large cumulative values with
unit "none". Switch to "short" for readable abbreviation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Count and message-volume panels (operating-mode transitions, job queue
depth, network/overlay message totals, getobject message counts) used
unit "none", rendering large values as raw unscaled numbers. Switch to
"short" so Grafana abbreviates (e.g. 1.5 Mil) for readability.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The system-* dashboards queried the legacy StatsD rippled_ prefix, but the
node now emits beast::insight metrics via native OTLP under the xrpld_
prefix (config: [insight] server=otel, prefix=xrpld). All queries returned
no data.
Migration (names derived from C++ beast::insight registrations, not live
Prometheus, since a syncing node does not emit every metric yet):
- rippled_ -> xrpld_ prefix across all panel queries and template variables
(including the $node variable query, which broke the whole dashboard filter)
- Histogram Event instruments export with unit ms, so bare _bucket becomes
_milliseconds_bucket: ios_latency, rpc_time, rpc_size, pathfind_fast/full
- Job-type metrics were StatsD summaries (label quantile="$quantile"); on the
OTLP path they are histograms. Converted those queries to
histogram_quantile($quantile, rate(xrpld_<job>_milliseconds_bucket[5m]))
and added the previously-undefined $quantile template variable
- Per-job-type detail panels: __name__ regex now matches _milliseconds_bucket
No panels removed. Panels for metrics not yet emitted (e.g. warn/drop, or
job types the syncing node has not run) show no data until the path executes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
P100 from a histogram is degenerate — it always returns the upper bound of
the highest populated bucket (a single slow outlier pins it to the top
boundary), producing a flat line. Revert to meaningful quantiles:
- Job Queue Wait Time / Job Execution Time: p75 (typical) + p99 (tail)
- Per-Job-Type / Per-Method: p99
- Added gauge panels showing current p99 with green/yellow/red thresholds
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The three duration heatmaps (transaction, consensus accept, RPC latency)
had an axisLabel of "Duration (ms)" but no unit code, so y-axis tick
values rendered unscaled. Set unit=ms on both the yAxis options and
panel defaults so buckets display as proper millisecond values.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Grafana does not recognize 'us' as a unit code, so microsecond values
rendered as raw numbers with a plain 'us' suffix (no scaling). The
correct code is 'µs'. Affects job-queue and OTel RPC latency panels
backed by *_duration_us histograms.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds template variables $tx_type, $ter_result, $txq_status to the
Transaction Overview dashboard. All relevant panels now respect these
filters, enabling operators to drill into specific transaction types
or result codes.
Changes:
- Panel 2 renamed to "Transaction Processing Latency by Type" (now
shows p95/p50 per tx_type instead of aggregate)
- Panels 1,3,4,5,7,9,12 filter by $tx_type
- Panel 10 filters by $tx_type and $ter_result
- Panel 11 filters by $txq_status
- Removed redundant "TX Processing Latency by Type (p95)" panel
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds template variables $tx_type, $ter_result, $txq_status to the
Transaction Overview dashboard. All relevant panels now respect these
filters, enabling operators to drill into specific transaction types
or result codes.
Changes:
- Panel 2 renamed to "Transaction Processing Latency by Type" (now
shows p95/p50 per tx_type instead of aggregate)
- Panels 1,3,4,5,7,9,12 filter by $tx_type
- Panel 10 filters by $tx_type and $ter_result
- Panel 11 filters by $txq_status
- Removed redundant "TX Processing Latency by Type (p95)" panel
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Shows p95 latency of tx.process span broken down by tx_type. Works for
both received and locally-processed transactions, unlike the tx.transactor
panel which requires the node to be synced and applying.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The OTel C++ SDK's SetAttribute appends rather than overwrites on
in-flight spans. Setting suppressed=false as a default then overriding
to true resulted in both values appearing in the exported span.
Fix: remove the default-false set, place suppressed=false once after
the HashRouter check passes (non-suppressed path), and suppressed=true
remains only in the suppressed path.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Resolve consensus dashboard conflict and remove duplicate
consensus_state dimension in collector config.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Collector config: add tx_type, ter_result, txq_status, consensus_state,
load_type, is_batch as spanmetrics dimensions so they appear as
Prometheus labels for dashboard queries.
New dashboard panels:
- Transaction Overview: Rate by Type, Results by Type, TxQ Status (pie),
Transactor Duration p95 by Type
- Consensus Health: Outcome Distribution (pie), Failures Over Time
- RPC Performance: Resource Cost by Command, Batch vs Single
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Resolve XrplCore.cmake conflict: keep telemetry module before tx module
(correct ordering from the cascade).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Wraps Transactor::operator() with a span that captures tx_type,
ter_result, and applied. This is the universal dispatch point — every
transaction flows through it, giving per-type latency breakdown.
Adds libxrpl.tx > xrpl.telemetry levelization dependency.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Wraps Transactor::operator() with a span that captures tx_type,
ter_result, and applied. This is the universal dispatch point — every
transaction flows through it, giving per-type latency breakdown.
Adds libxrpl.tx > xrpl.telemetry levelization dependency.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Resolve runbook conflict: keep both phase 6 ledger/peer span tables
AND new insights/sample queries section from the enrichment work.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds comprehensive "Insights and Sample Queries" section showing operators
what questions they can answer with the newly-added span attributes:
- Transaction workflow analysis (filter by tx_type, fee, ter_result)
- TxQ health (txq_status, ledger_changed)
- RPC debugging (is_batch, request_payload_size, load_type)
- PathFinding performance (dest_currency, num_source_assets)
- Consensus health (consensus_state, is_bow_out, disputes_count)
- Cross-subsystem correlation examples
Also updates all span reference tables with the new attributes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds workflow-critical attributes to consensus spans:
- consensus.proposal.send: is_bow_out (identifies resignation proposals)
- consensus.accept: consensus_state (yes/moved_on/expired), disputes_count
- consensus.validation.send: ledger_hash (correlates validation to ledger)
Enables answering: "Did we reach consensus or time out?", "How many
disputes existed at acceptance?", "Which ledger did we validate?"
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The apply() function doesn't have a `using namespace telemetry` directive
(unlike processTransaction), so tx_span attrs need explicit qualification.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Enriches the tx.process span with final outcome after batch application:
- ter_result: the TER code string (e.g., "tesSUCCESS", "tecPATH_DRY")
- applied: boolean whether the transaction was included in the ledger
These attributes complete the tx.process span lifecycle — it now captures
identity (tx_type, tx_hash), intent (fee, sequence), and outcome
(ter_result, applied) for full workflow traceability.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Wire up span attributes that enable filtering/grouping traces by request
characteristics: batch detection, payload size, resource cost category,
command name on WS spans, and pathfinding search parameters (destination
amount/currency, source asset count).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The check-rename CI job requires all rename scripts to have been run.
The telemetry config files had 'prefix=rippled' which should be 'xrpld'.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The project-wide rename check changed the factory function name but
missed this call site in the consensus simulation test framework.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ServiceRegistry gained the pure virtual getMetricsRegistry() in phase 7
but TestServiceRegistry was never updated. Returns nullptr since tests
don't need a real MetricsRegistry.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
clang-tidy misc-include-cleaner requires each symbol to be provided by
a directly-included header. Replace the convenience trace/context.h
(which only provides GetSpan/SetSpan) with the specific headers for
kSpanKey, holds_alternative, get, shared_ptr, Span, and span.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two stray "rippled" tokens introduced by 43258e8d ("docs(telemetry):
add secure-OTel pipeline analysis…") were caught by check-rename in
CI. Re-run docs.sh to convert them to xrpld so the rename check
passes on PR #6425 (and downstream PR #6426 once merged up).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Brings in pratik/otel-phase3-tx-tracing's two commits that move all
consensus-tracing content off phase-3:
- c9521b97fe refactor(telemetry): pull consensus-tracing scope-leak
out of phase-3
- d6b101069e refactor(telemetry): remove consensus tracing from
phase-3
Phase-4 already owns this content (with the consensus_span -> cons_span
namespace rename, round/accept/proposalSend/validationSend span
construction, and the relocated ConsensusSpanNames.h under
src/xrpld/consensus/), plus its own evolution on top — so the merge is
resolved as a tree-identity merge:
- src/xrpld/app/consensus/ConsensusSpanNames.h: keep phase-4's
renamed/expanded version (modify/delete conflict, resolved with
--ours).
- src/xrpld/app/consensus/RCLConsensus.cpp: keep phase-4's version
with cons_span attribute calls, the trace_context inject blocks
on broadcastPropose/validate, the secure-OTel TODO, and the
full validation/round span instrumentation (content conflict,
resolved with --ours).
- src/xrpld/overlay/detail/PeerImp.cpp: keep phase-4's version with
the proposalReceiveSpan/validationReceiveSpan calls, lambda span
captures, and cons_span::attr::* setAttribute calls (content
conflict, resolved with --ours).
- src/xrpld/telemetry/ConsensusReceiveTracing.h: phase-3 deleted it,
phase-4 still uses it. Restored from phase-4 HEAD (silent
auto-deletion otherwise).
- include/xrpl/telemetry/TraceContextPropagator.h: phase-3 stripped
consensus references and the secure-OTel TODO; phase-4 still has
both. Restored from phase-4 HEAD.
- src/xrpld/telemetry/PropagationHelpers.h: phase-3 swapped the
@see ConsensusReceiveTracing.h cross-reference for TxTracing.h;
phase-4 still wants the consensus reference. Restored from
phase-4 HEAD.
Net tree change on phase-4: zero. Verified via 'git diff <pre-merge-sha>
HEAD' returning empty.
Signed-off-by: Pratik Mankawde <3397372+pratikmankawde@users.noreply.github.com>
Phase-3 (PR #6425) is scoped to transaction tracing only; consensus
tracing belongs to phase-4 (PR #6426). The previous commit on this
branch removed the namespace/attribute scaffolding c6c019ed8b leaked
into phase-3, but phase-3 still carried the consensus span construction
and trace-context propagation introduced in earlier commits
(61cb1faf8f, 93bed03d8d). Move that out too so phase-3 creates and
propagates no consensus spans of any kind.
Removed:
- src/xrpld/telemetry/ConsensusReceiveTracing.h (deleted; phase-4
owns it).
- PeerImp.cpp: remove the std::make_shared<SpanGuard>(
proposalReceiveSpan(...))/validationReceiveSpan(...) constructions
in onMessage(TMProposeSet)/onMessage(TMValidation), drop the
sp = std::move(span) lambda captures, and drop the
#include <xrpld/telemetry/ConsensusReceiveTracing.h>.
- RCLConsensus.cpp: drop the two telemetry::injectToProtobuf() blocks
that injected the active trace context into TMProposeSet (in
Adaptor::propose, after addSuppression) and TMValidation (in
Adaptor::validate, around the broadcast call). Drop the now-unused
#include of TraceContextPropagator.h and the
XRPL_ENABLE_TELEMETRY-gated include of
opentelemetry/context/runtime_context.h.
- TraceContextPropagator.h: update file-level @see comment to drop
the ConsensusReceiveTracing.h reference and to scope the
"wired into the P2P message flow via PropagationHelpers.h"
sentence to TMTransaction only.
- PropagationHelpers.h: replace the
@see ConsensusReceiveTracing.h cross-reference with
@see TxTracing.h.
Inert consensus metadata (TraceCategory::Consensus enum value,
seg::consensus constant, isCategoryEnabled/categoryToSpanKind switch
arms, the SpanGuard.h doc-comment example) is intentionally preserved
on phase-3: nothing references it after this commit, but phase-4
needs it and removing it would widen the phase-3 -> phase-4 merge
surface for no benefit.
Verified via git grep: no remaining phase-3 references to
proposalReceiveSpan, validationReceiveSpan, ConsensusReceiveTracing,
consensus_span::, consensus.proposal, or consensus.validation.
Signed-off-by: Pratik Mankawde <3397372+pratikmankawde@users.noreply.github.com>
Commit c6c019ed8b ("addressed code review comments") bundled tx-tracing
review fixes with consensus-tracing scaffolding that belongs on
pratik/otel-phase4-consensus-tracing (PR #6426). This commit lifts the
consensus-only parts back out of phase-3 so PR #6425 stays scoped to
transaction tracing. Phase-4 already carries the same content (via prior
phase-3 -> phase-4 merges) plus its own evolution on top, so nothing is
moved across — only removed here.
Removed:
- src/xrpld/app/consensus/ConsensusSpanNames.h (deleted; phase-4 owns
it).
- PeerImp.cpp: revert onMessage(TMProposeSet)/onMessage(TMValidation)
consensus-attr setAttribute calls to the hardcoded
"xrpl.consensus.{trusted,round,ledger.seq}" strings used before
c6c019ed8b. Drop the now-unused
#include <xrpld/app/consensus/ConsensusSpanNames.h>.
- RCLConsensus::Adaptor::validate(): remove the
TODO(observability/secure-OTel) block on validation trace_context.
- TraceContextPropagator.h: remove the TODO(observability/secure-OTel)
block on injectToProtobuf().
Tx-tracing parts of c6c019ed8b are intentionally untouched.
No phase-3 caller of telemetry::consensus_span:: remains; verified via
git grep. No test on phase-3 references the removed header.
Signed-off-by: Pratik Mankawde <3397372+pratikmankawde@users.noreply.github.com>
Document the threat model and chosen hardening approach for the OTel
pipeline: mTLS to the collector as primary defense (across-network
deployment), NetworkPolicy as defense-in-depth, and source-side
validation plus per-peer rate limiting for protocol::TraceContext on
peer messages. Skips Basic Auth (wrong shape for multi-operator
fleet) and HTTP-gateway header stripping (rippled is P2P).
Wires the new doc into the master plan ToC, mermaid diagram, and
body section, plus cross-refs from the privacy section in
02-design-decisions.md and the collector config in
05-configuration-reference.md so readers reach it from natural
in-context entry points. Adds a backlink at the top of secure-OTel.md
to the master plan.
Adds 'exfiltration' and 'htpasswd' to cspell dictionary.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Drop xrpl.node.amendment_blocked / xrpl.node.server_state from telemetry
surface (constants in SpanNames.h, two filters in tempo.yaml). Operators
read the same data via server_info / server_state RPC; OTel SDK 1.18.0
cannot refresh resource attrs at runtime so resource-level emission was
not viable either.
- Namespace all pathfind span attributes under pathfind_* (underscore form
per Phase 1c rule 5). Renames in PathFindSpanNames.h and call sites in
PathRequest.cpp, PathRequestManager.cpp, plus the rule-5 retention
xrpl.pathfind.ledger_index -> pathfind_ledger_index.
- Wire pathfind_source_account / pathfind_dest_account on pathfind.request
in doPathFind / doRipplePathFind handlers (only when present + string).
- Collapse per-asset pathfind.discover / pathfind.rank spans into one
pathfind.discover hoisted around the per-source-asset loop in
PathRequest::findPaths. Span count goes from 2N to 1 per RPC call;
per-asset breakdown traded for bounded storage and cardinality. Trade-off
documented inline.
- Fix pathfind_num_paths semantics: now sums getBestPaths().size() across
the loop (paths actually returned) instead of the maxPaths input cap.
- PathRequestManager::updateAll: move span creation after the locked
requests_ snapshot, early-return when no active subscriptions exist
(avoids empty span on every ledger close), set pathfind_num_requests
= requests.size().
- Update Phase2_taskList.md and 02-design-decisions.md to match.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per PR #6438 review thread r3250432621: known-command errors
(rpcTOO_BUSY, rpcNO_PERMISSION, etc.) were collapsing into a
single rpc.command.unknown span name, hiding per-command error
rates in dashboards. Same anti-pattern existed for gRPC, where
every method was bucketed under grpc.request with the method
relegated to an attribute.
- RPCHandler.cpp: doCommand error path uses cmdName as the span
suffix; the rpc_span::val::unknownCommand fallback only applies
when the request truly omits both command and method fields.
- GRPCServer.cpp: gRPC span name is now grpc.<MethodName>
(e.g. grpc.GetLedger). Method also retained as an attribute.
- GrpcSpanNames.h: drop the unused op::request constant; update
the span-hierarchy comment.
- RpcSpanNames.h: update the gRPC span diagram to match.
Dashboards on downstream phases will benefit from per-command
breakdowns without needing TraceQL attribute filters.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three interrelated fixes in otel-collector-config.yaml; without them the
Phase 8 log-trace correlation pipeline is silently broken.
1. `resource/logs` processor now upserts `job: xrpld` alongside
`service.name: xrpld`. Loki 3.x OTLP ingestion renames
`service.name` to the label `service_name`, so the runbook /
integration-test queries (`{job="xrpld"} |= "trace_id="`) returned
empty. Upserting the `job` resource attribute at the collector lets
the canonical Loki label flow through unchanged.
2. `filelog` regex makes the `partition:` capture non-capturing-optional.
`Logs::format()` omits the `partition:` prefix when partition is
empty (common for framework-level log lines); the old regex required
it and silently dropped those records.
3. Timestamp parser now matches the real log format. `Logs::format()`
writes microsecond-precision timestamps like
`2026-04-15 10:30:45.123456 UTC`. The layout was
`%Y-%b-%d %H:%M:%S` — missing fractional seconds and timezone —
which failed strptime and dropped timestamps. New layout is
`%Y-%b-%d %H:%M:%S.%f` with `location: UTC`.
Also adds a block-comment documenting the real log format so the
next person to touch this doesn't re-introduce the same gaps.
- `src/test/telemetry/MetricsRegistry_test.cpp` (Beast `unit_test::suite`
format under `src/test/`) duplicates the GTest version already
maintained at `src/tests/libxrpl/telemetry/MetricsRegistry.cpp`.
Project rule (`tasks/lessons.md` §Test Format): all new tests use
GTest under `src/tests/libxrpl/`. The GTest version exercises the
same four cases (disabled construction, start/stop lifecycle, recording
no-op, destructor-calls-stop). Deleting the Beast duplicate eliminates
drift and keeps the test authoritative in one place.
- Drop the matching `test.telemetry > xrpl.basics/xrpl.core/xrpld.telemetry`
entries from `.github/scripts/levelization/results/ordering.txt`
because `xrpl.test.telemetry` (the GTest binary) retains its own
entries; the removed ones belonged to the deleted Beast suite.
- `.claude/instructions.md` was committed as a symlink to an
author-local absolute path (`/home/pratik/sourceCode/personal/Rippled/
instructions.md`) that does not exist for any other contributor or in
CI. Remove the symlink from git tracking and add `.claude/` to
`.gitignore` so future agent commits do not re-add per-developer
settings.
MetricsRegistry observable-gauge callbacks run on the OTel reader thread
and read live state from nodeStore_, overlay_, networkOPs_, ledgerMaster,
inboundLedgers, loadManager, and others. The old shutdown sequence called
metricsRegistry_->stop() AFTER all those services were already stopped,
which left a race window between each service's stop() and the final
provider_->ForceFlush() during which a callback could dereference
already-stopped service state. The try/catch guards in each callback
mitigated crashes but not reads from freed members.
- Add MetricsRegistry::detachCallbacks() that sets an atomic<bool>
callbacksDetached_ with release ordering. Idempotent.
- Guard every ObservableGauge callback entry with an acquire-load of the
same flag and return early if it is set. Covers all 15 registered
callbacks (cacheHitRate, txq, objectCount, loadFactor, nodeStore,
serverInfo, buildInfo, completeLedgers, dbMetrics, validatorHealth,
peerQuality, ledgerEconomy, stateTracking, storageDetail,
validationAgreement).
- Application::run() shutdown sequence now calls
metricsRegistry_->detachCallbacks() right after m_loadManager->stop()
and BEFORE m_shaMapStore, m_jobQueue, overlay_, grpcServer_,
m_networkOPs, serverHandler_, m_ledgerReplayer, m_inboundTransactions,
m_inboundLedgers, ledgerCleaner_, m_nodeStore, perfLog_ are stopped.
The acquire/release pair guarantees subsequent reader-thread ticks see
the detach before they dereference stopped services.
- metricsRegistry_->stop() keeps setting the flag as a belt-and-suspenders
defense in case a future caller forgets to detach first.
- Drop the misleading "No explicit RemoveCallback is needed" comment
from stop(); provider destruction alone does not beat the reader
thread to already-freed state.
The objectCountGauge callback previously discarded its state pointer
via `void* /* state */`; restore the state argument so it can access
self->callbacksDetached_ too.
File renames to match the post-docs.sh project-wide rename + the UID
rename applied in the previous commit. Five phase-9 dashboards are
affected:
- rippled-fee-market.json -> xrpld-fee-market.json
- rippled-job-queue.json -> xrpld-job-queue.json
- rippled-peer-quality.json -> xrpld-peer-quality.json
- rippled-rpc-perf.json -> xrpld-rpc-perf-otel.json
- rippled-validator-health.json-> xrpld-validator-health.json
`rippled-rpc-perf.json` is renamed to `xrpld-rpc-perf-otel.json` (rather
than `xrpld-rpc-perf.json`) to avoid colliding with the
phase-6 `rpc-performance.json` dashboard which also uses the
`xrpld-rpc-perf` UID. The new filename matches its now-unique
`xrpld-rpc-perf-otel` UID that was set in the merge commit.
Phase 4 added a span catalog in `06-implementation-phases.md` listing the
source location for each consensus span. Line numbers `Consensus.h:707`,
`RCLConsensus.cpp:232/341/492/541/900` drift on every refactor and would
become stale PR after PR. Filename alone is enough for operators to
grep — the RCLConsensus.cpp spans are already unambiguous from the span
name itself.
Follow-up to the phase-6 dashboard cleanup. The three dashboards
introduced by commit f6105ece98 (consensus-health, rpc-performance,
transaction-overview) were missed in the initial UID rename and still
carried `rippled-*` UIDs plus line-number refs in panel descriptions.
- UIDs: `rippled-consensus` -> `xrpld-consensus`,
`rippled-rpc-perf` -> `xrpld-rpc-perf`,
`rippled-transactions` -> `xrpld-transactions`, matching the
post-`docs.sh`-rename runbook and the other dashboards in this PR.
- Strip `:<line>` suffixes from `ServerHandler.cpp`, `RCLConsensus.cpp`,
`NetworkOPs.cpp`, etc. references in panel descriptions. Line numbers
drift on every refactor; the filename is enough to grep.
- Fix the Overall RPC Throughput panel: two targets filtered on
`span_name="rpc.request"` (never emitted) instead of
`span_name="rpc.http_request"` (the real emitted name). The panel
would have shown zero data until this fix.
Follow-up to the dashboard cleanup on this branch. Caught additional sites
in TESTING.md that still reference the never-emitted `rpc.request` span:
- TraceQL query examples in Step 5 "Verify traces in Tempo" now filter on
`name="rpc.http_request"` (the real emitted name).
- Expected-spans table replaces `rpc.request` with `rpc.http_request`.
- Query loop under the Prometheus verification section now iterates over
the full set of emitted RPC entry-point names
(`rpc.http_request`, `rpc.ws_upgrade`, `rpc.ws_message`, `rpc.process`).
Also drop `exporter=otlp_http` from the sample telemetry config block.
`TelemetryConfig.cpp` does not parse an `exporter` key in any phase through
Phase 8; only OTLP/HTTP is wired up, so the line is either a silently
ignored no-op or misleading documentation.
Phase-6 introduces ledger-operations, peer-network, and the five StatsD
dashboards. Align them with the rest of the chain:
- Rename dashboard UIDs from `rippled-*` to `xrpld-*` so the provisioned
UIDs match the post-rename-script documentation (`docs.sh` rewrites
.md but not .json, so the two drifted). Runbook references
`xrpld-rpc-perf`, `xrpld-transactions`, etc., now the JSON matches.
- Add the `$node` template variable + `exported_instance=~"$node"` filter
to every target in the five `statsd-*` dashboards. Mirrors the pattern
already used by consensus-health, ledger-operations, and peer-network
per the project rule that every dashboard must support per-node
filtering.
- Strip `:<line>` (and `:NN-NN` range) suffixes from C++ file references
in every dashboard panel description and in docker/telemetry/TESTING.md.
Line numbers drift on every refactor; the filename alone is enough to
grep.
- Replace stale `rpc.request` entries with the real emitted span names
(`rpc.http_request`, `rpc.ws_upgrade`, `rpc.ws_message`, `rpc.process`)
in TESTING.md so operators can copy-paste the filters and hit real
traces.
- Also drop the `:706` line ref from the `StatsDCollector.cpp` callout
in `06-implementation-phases.md`.
- RPC Spans table: `rpc.request` was documented but the code actually emits
`rpc.http_request`. Listed the actual emitted names
(`rpc.http_request`, `rpc.ws_upgrade`, `rpc.ws_message`, `rpc.process`)
and their parent/child relationship.
- Drop `:<line>` suffixes from Source File columns in both RPC and
Transaction span tables. Line numbers drift with every refactor; the
filename is enough for operators to grep.
- Summary table: replace the never-emitted `rpc.request` row with the real
entry points so `span_name=` filters in PromQL / TraceQL match.
Phase-1a plan documents advertised OTLP/gRPC on port 4317 as the default
exporter, four unparsed [telemetry] config keys, and "Phase 4a Complete"
status with exit-criteria checkboxes marked done. Every downstream branch
through Phase 5 ships only OTLP/HTTP on port 4318 via OtlpHttpExporterFactory,
never parses the advertised keys, and the Phase 4 work is not yet delivered.
Fixes:
- 02-design-decisions.md: flip §2.1.1 SDK dependency recommendations to
OTLP/HTTP (shipped) with OTLP/gRPC marked Future. Update §2.2 architecture
diagram and text from OTLP/gRPC:4317 to OTLP/HTTP:4318. Rewrite §2.2.1 as
"OTLP/HTTP (Shipped)" and §2.2.2 as "OTLP/gRPC (Future Work — Planned
Upgrade)" with a concrete checklist (Conan dep, config parsing, factory
branch, runbook/dashboard updates) for landing the gRPC transport later.
- 05-configuration-reference.md: drop the fabricated exporter/otlp_grpc key
and the :4317 default from the sample config block and the options-summary
table. Move trace_pathfind, trace_txq, trace_validator, trace_amendment
into a new "Planned (not yet implemented)" table citing the phase that will
add each one. Keep the example config minimal so copy-paste does not produce
a silently-ignored stanza.
- 06-implementation-phases.md: reset Phase 4 Exit Criteria checkboxes from
[x] to [ ] (Phase 4 is not shipped at Phase-1a time). Rename "Phase 4a
Complete" to "Phase 4a Plan" and describe the work as future. Replace the
broken forward link to Phase4_taskList.md (introduced in the Phase 2 PR)
with a sentence pointing readers to where that spec will land. Renumber
the final section 6.12 to 6.11 so it sits directly after 6.10; section 6.11
("Effort Summary") was intentionally removed in earlier edits.
SpanGuard::span() hardcoded SpanKind::kInternal for every span. Tempo's
service-graph and spanmetrics RED calculations rely on kServer /
kConsumer / kClient / kProducer to classify inbound vs outbound vs
internal operations. With kInternal everywhere, the service graph
collapses to a single self-loop and RED metrics attribute all latency
to internal work.
Add categoryToSpanKind() mapping:
- Rpc -> kServer (inbound synchronous request)
- Peer -> kConsumer (inbound async peer message)
- Transactions -> kInternal
- Consensus -> kInternal
- Ledger -> kInternal
Only the single-argument overload is affected; childSpan / linkedSpan
continue to default to kInternal because they represent in-process
continuations of an already-kinded parent.
Spanmetrics dimensions used xrpl.rpc.command etc. but C++ emits bare
"command". Tempo tags for phase6-added consensus/tx/peer filters used
qualified names but C++ uses bare names. Dashboard panel referenced
xrpl_tx_suppressed (never populated) instead of suppressed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Consensus span attributes use bare names (close_time_correct,
consensus_state, close_resolution_ms) and shared canonical attrs
(xrpl.ledger.seq) per SpanNames.h. xrpl.consensus.mode and
xrpl.consensus.round are correct (domain-qualified to avoid collision).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Transaction span attributes use bare names (local, tx_status) per
SpanNames.h convention, not xrpl.tx.* qualified names. xrpl.tx.hash
is correct (shared canonical attr defined in SpanNames.h).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
RPC span attributes use bare names (command, rpc_status, rpc_role) per
the naming convention in SpanNames.h, not xrpl.rpc.* qualified names.
Node health attributes (amendment_blocked, server_state) are resource
attributes set at Tracer init, not span attributes.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Panels 8-15 from statsd-node-health.json and panels 8-9 from
statsd-network-traffic.json were lost when Phase 7 renamed these files
to system-*. The merge (5cd71ed107) took Phase 7's smaller version
without the extra panels added by commit b933e8ae00 on Phase 6.
Recovered panels (system-node-health.json):
- Key Jobs Execution Time (11 job types)
- Key Jobs Dequeue Wait Time (11 job types)
- FullBelowCache Size
- FullBelowCache Hit Rate
- Ledger Publish Gap (validated - published age delta)
- State Duration Rate (Full vs Tracking)
- All Jobs Execution Time Detail (34 job types)
- All Jobs Dequeue Wait Detail (34 job types)
Recovered panels (system-network-traffic.json):
- Duplicate Traffic (Wasted Bandwidth)
- All Traffic Categories Detail (topk 15 by byte rate)
All recovered panels updated to include exported_instance=~"$node"
filter per project dashboard guidelines.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- MetricsRegistry.cpp: concatenate nested namespaces, add missing
direct includes (Journal.h, string, string_view, cstdint), suppress
readability-convert-member-functions-to-static in #else stubs by
referencing enabled_ member, void unused instanceId parameter.
- MetricsRegistry test: add missing direct includes (Log.h, Journal.h,
uint256.h, io_context.hpp, optional, stdexcept, string), make
throwUnimplemented() static, add [[nodiscard]] to getOpenLedger/
isStopping/getTrapTxID overrides, make const-eligible registry const.
- PerfLogImp.cpp: add braces around if/else body per
readability-braces-around-statements.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- MetricsRegistry.cpp: concatenate nested namespaces, add missing
direct includes (Journal.h, string, string_view, cstdint), suppress
readability-convert-member-functions-to-static in #else stubs by
referencing enabled_ member, void unused instanceId parameter.
- MetricsRegistry test: add missing direct includes (Log.h, Journal.h,
uint256.h, io_context.hpp, optional, stdexcept, string), make
throwUnimplemented() static, add [[nodiscard]] to getOpenLedger/
isStopping/getTrapTxID overrides, make const-eligible registry const.
- PerfLogImp.cpp: add braces around if/else body per
readability-braces-around-statements.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
NetworkOPs.h and SpanNames.h were only needed for per-span
nodeAmendmentBlocked/nodeServerState calls, which were removed
in the attr naming simplification. Fixes clang-tidy CI failure.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- otel-collector-config.yaml: spanmetrics dimensions use new bare names.
- tempo.yaml: TraceQL filter tags use new bare names.
- 02-design-decisions.md: strip xrpl.txq.* prefix from planned attrs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Phase2_taskList: update attr refs to bare names, note node-health
attrs moved to resource level.
- 02-design-decisions: strip xrpl.pathfind.* prefix from planned attrs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Update OpenTelemetryPlan docs and Telemetry.h doc example to reflect
the renamed per-span attributes: xrpl.rpc.command -> command,
xrpl.rpc.status -> rpc_status, xrpl.grpc.method -> method, etc.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Update 31 attribute references in telemetry-runbook.md to match the
simplified naming: drop xrpl.<domain>. prefix on per-span attrs, use
domain-qualified names for collisions (rpc_status, consensus_state,
etc.), and unify cross-domain refs (xrpl.ledger.seq, xrpl.tx.hash).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Drop xrpl.pathfind.* prefix from per-span attrs (source_account,
dest_account, fast, search_level, num_complete_paths, num_paths,
num_requests).
- Keep xrpl.pathfind.ledger_index qualified (rule 5: distinct from
xrpl.ledger.seq).
- Remove per-span nodeAmendmentBlocked/nodeServerState calls from
RPCHandler — promoted to resource-level attrs.
- Mark node-health attrs in SpanNames.h as RESOURCE-ONLY with doc.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Phase 10's workload validation configs (expected_metrics.json,
regression-metrics.json, validate_telemetry.py) queried the
MetricsRegistry metrics under the rippled_ prefix, but MetricsRegistry
emits them as xrpld_ (see MetricsRegistry.cpp). On a live run the
workload validator reported every MetricsRegistry metric as missing,
masking genuine regressions.
Rename the following to xrpld_ across the workload validator,
expected-metrics manifest, and regression-metrics template:
- nodestore_state, cache_metrics, txq_metrics, load_factor_metrics,
object_count
- rpc_method_started_total / _finished_total / _errored_total /
_duration_us
- job_queued_total / _started_total / _finished_total /
_queued_duration_us_bucket / _running_duration_us_bucket
- peer_quality, server_info, validator_health, ledger_economy,
db_metrics, complete_ledgers, build_info, state_tracking,
storage_detail
- ledgers_closed_total, validations_sent_total,
validations_checked_total, state_changes_total
- validation_agreement, validation_agreements_total,
validation_missed_total
Mirrors the phase-9 fix in commit 5601615952.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
MetricsRegistry emits OTel SDK metrics with the xrpld_ prefix
(MetricsRegistry.cpp defines "xrpld_nodestore_state",
"xrpld_cache_metrics", etc.), but the Phase 9 dashboards and the
Step 10c integration-test assertions introduced in 892fee638a
queried the rippled_ prefix. Every Phase 9 panel and assertion
therefore rendered "No data" or failed on a live run, even though
the underlying series were being exported correctly.
Rename the rippled_ prefix to xrpld_ for every MetricsRegistry
metric in dashboards and the integration test:
- nodestore_state, cache_metrics, txq_metrics, load_factor_metrics,
object_count
- rpc_method_started_total / _finished_total / _errored_total /
_duration_us_bucket
- job_queued_total / _started_total / _finished_total /
_queued_duration_us_bucket / _running_duration_us_bucket
- peer_quality, server_info, validator_health, ledger_economy,
db_metrics, complete_ledgers, build_info, state_tracking
- ledgers_closed_total, validations_sent_total,
validations_checked_total, state_changes_total
- validation_agreement (ValidationTracker 1h/24h/7d windows)
Also add ValidationTracker window-gauge assertions to Step 10c of
integration-test.sh so the 1h/24h/7d agreement and miss counts are
checked alongside the other Phase 9 gauges.
The rippled_ prefix is preserved for beast::insight metrics
(rippled_LedgerMaster_*, rippled_Peer_Finder_*, rippled_total_*,
rippled_Overlay_*, rippled_State_Accounting_*, rippled_transactions_*,
rippled_proposals_*, rippled_validations_Messages_*) because those
flow through the StatsD-style OTelCollector configured with
`[insight] prefix=rippled` and remain on that prefix by design.
Verified against a live 6-node consensus network: all 22 Phase 9 +
ValidationTracker assertions now report 6+ series per metric.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Two fixes so gauges register in Prometheus (via StatsD) even when their
initial/steady-state value is 0:
1. StatsDGaugeImpl m_dirty: default-init to true so the initial value
(0) is emitted on the first flush. Previously, gauges whose value
never changed from 0 were never flushed and never appeared
downstream.
2. io_latency_sampler firstSample_: new atomic<bool>, init true.
m_event.notify now fires when either firstSample_ is true (exchanged
to false) or lastSample >= 10 ms. This guarantees the io_latency
metric is registered on startup; subsequent sub-10 ms samples are
still suppressed to avoid flooding.
Set defaults for tx_span::attr::suppressed (false) and
tx_span::attr::status ("new") immediately after creating the txReceive
span. Without defaults, spans whose suppressed/status attributes would
only be set in the HashRouter-suppressed branch lacked these attributes
entirely, producing incomplete span data in downstream stores.
The suppressed branch still overrides these when the transaction has
already been seen via HashRouter.
Two build failures surfaced by CI on the Phase 9 branch:
1. NetworkOPsImp stores the ServiceRegistry as
std::reference_wrapper<ServiceRegistry> registry_, so calls must go
through registry_.get().<method>(). The MetricsRegistry hooks added
in setMode() and recvValidation() dereferenced the wrapper directly,
which compiles against a pre-existing accessor on the wrapper type
on some toolchains but fails on clang 16/17/20 and gcc 13/15 with
"no member named 'getMetricsRegistry' in
std::reference_wrapper<xrpl::ServiceRegistry>".
2. MetricsRegistry::app_ and MetricsRegistry::journal_ are only used
inside XRPL_ENABLE_TELEMETRY-guarded code paths (gauge callbacks
and JLOG). When telemetry is disabled, clang's
-Werror=-Wunused-private-field tripped. Move the two fields under
the same #ifdef and guard the constructor initialisers with
[[maybe_unused]] so the no-op build continues to compile cleanly.
Addresses code review findings on PR #6513:
1. registerAsyncGauges() was ~730 lines, violating the CLAUDE.md
rule "No function longer than 80 lines." Split into fifteen
per-domain helpers (cache, TxQ, object count, load factor,
NodeStore, server info, build info, complete ledgers, DB,
validator health, peer quality, ledger economy, state tracking,
storage detail, validation agreement) dispatched from a thin
shell. Each helper now stays at or below the 80-line limit.
2. PerfLogImp::rpcEnd() only updated the in-memory counter and
never advanced the OTel xrpld_rpc_method_finished_total,
xrpld_rpc_method_errored_total, or xrpld_rpc_method_duration_us
instruments. rpcStart() was already wired up, so the finished
and errored counters stayed at zero for every RPC call.
rpcEnd() now computes the duration once, records it under the
existing mutex, and forwards finish/error events to
MetricsRegistry::recordRpcFinished / recordRpcErrored outside
the counter mutex to avoid lock nesting with the OTel SDK.
3. Added class-level Doxygen for MetricsRegistry with an ASCII
collaborator diagram and explicit @note tags covering
thread-safety, lifetime, and extension guidance.
Tempo /api/traces/{id} returns OTLP-shaped JSON with a top-level
"batches" key, not "data". The cross-check in check_log_correlation
was querying jq '.data | length' which always returned null, causing
the Log-Tempo cross-check to fail even when the trace existed.
MockServiceRegistry in MetricsRegistry.cpp still used the old method
names (timeKeeper, cachedSLEs, validators, overlay, cluster, app, etc.)
while ServiceRegistry has been standardized on getXxx()/isXxx() forms.
Windows CI caught this as C3668 "did not override any base class methods"
errors and C2259 "cannot instantiate abstract class".
Rename all 13 mismatched overrides to match the current interface:
timeKeeper -> getTimeKeeper
cachedSLEs -> getCachedSLEs
validators -> getValidators
validatorSites -> getValidatorSites
validatorManifests -> getValidatorManifests
publisherManifests -> getPublisherManifests
overlay -> getOverlay
cluster -> getCluster
peerReservations -> getPeerReservations
pendingSaves -> getPendingSaves
openLedger (x2) -> getOpenLedger
getPathRequests -> getPathRequestManager (type rename too)
journal -> getJournal
logs -> getLogs
trapTxID -> getTrapTxID
app -> getApp
Also regenerate levelization ordering.txt to reflect the new
tests.libxrpl -> xrpl.core edge introduced by ServiceRegistry.h include.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Consensus.h (Phase 4 tracing) depends on DisputedTx::getYays()/getNays()
to build disputeResolve span events. Both accessors were removed by
earlier 'duplicate accessor' cleanup commits on this branch, leaving
Consensus.h referencing non-existent members. CI caught this on
macOS/clang-17/gcc-13/Windows builds.
Restore the accessors on the branch where they were dropped so downstream
phase branches inherit a compiling DisputedTx.h via merge.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Clang-tidy fixes:
- Concatenate nested namespaces (modernize-concat-nested-namespaces)
in OTelCollector.h, OTelCollector.cpp, ValidationTracker.h/.cpp
- Add missing direct includes (misc-include-cleaner) in
ValidationTracker.cpp, test, CollectorManager.cpp, OTelCollector.cpp
- Make lock_guard variables const (misc-const-correctness)
- Add braces around single-line if/else (readability-braces-around-statements)
- Use designated initializer for WindowEvent (modernize-use-designated-initializers)
- Initialize LedgerEvent::seq field (cppcoreguidelines-pro-type-member-init)
Linker fix:
- Add ValidationTracker.cpp as source to xrpl.test.telemetry target
(it lives in src/xrpld/ but the test links against libxrpl only)
Levelization fix:
- Remove stale dependency edges from ordering.txt that were introduced
by the erroneous develop-merge commit
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
These were already added in an earlier phase branch. The duplicate
with slightly different Doxygen wording was introduced by the
erroneous merge/revert cycle.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace GetSpan() with direct context value check in Logs::format()
to avoid heap allocation (new DefaultSpan) on the no-span path
- Restore Phase 7 documentation accidentally deleted during merge
- Fix undefined $JAEGER variable → use $TEMPO in integration test
- Remove useless LCOV_EXCL markers around #ifdef block
- Fix indentation inconsistencies in Log.cpp injection block
- Remove incorrect url field from loki.yaml derivedFields
- Update stale code sample in Phase8_taskList.md to match implementation
- Correct "<10ns" performance claims to accurate ~15-20ns (no-span)
and ~50ns (active-span) measurements across all docs
- Replace Jaeger references with Tempo in TESTING.md (port 16686→3200)
- Improve error handling in check_log_correlation(): track files_scanned,
detect missing log files, fix silent grep error masking
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Fix use-after-free: extract gauge callback to static function and call
RemoveCallback in ~OTelGaugeImpl() before unregistering from collector
- Use memory_order_acq_rel on callHooks() debounce CAS for proper
happens-before relationship between hook invocations
- Add explicit 2s timeout to ForceFlush() in destructor to prevent
blocking indefinitely when OTLP endpoint is unreachable at shutdown
- Add OTLP receiver to metrics pipeline so native OTel metrics from
xrpld are actually received by the collector
- Remove stale health check port from docker-compose (extension was
removed from collector config)
- Clarify fallback docs: StatsD path requires re-enabling receiver/port
- Fix comments: Counter uses uint64_t not int64_t, gauge clamps to
[0, INT64_MAX] not [0, UINT64_MAX]
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The blanket revert of f4555c80fe also un-reverted some files that had
been correctly matched to phase-6 (nodestore Backend API refactor,
Vault_test changes). Restore those to the base branch state so the
phase-7 PR only contains telemetry changes.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Build fixes in PeerImp.cpp:
- Rename duplicate `span` variable to `consSpan` in proposal and
validation handlers to avoid redefinition error
- Fix `->` on non-pointer SpanGuard (now correctly on shared_ptr)
- Fix move-only type copy in lambda capture
Clang-tidy fixes:
- Concatenate nested namespaces in LedgerSpanNames.h and PeerSpanNames.h
- Add missing SpanNames.h includes in BuildLedger.cpp, LedgerMaster.cpp,
PeerImp.cpp for direct seg:: symbol usage
- Add missing <chrono> and <cstdint> includes in BuildLedger.cpp
- Remove unused Feature.h include from BuildLedger.cpp
Rename check fix:
- Run docs.sh to rename rippled_ metric prefixes to xrpld_ in
09-data-collection-reference.md and telemetry-runbook.md
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- protocol/README.md: restore historical GitHub URL path (src/ripple/)
- Config.cpp: restore configLegacyName as "rippled.cfg" (legacy name
must remain as-is for backward compatibility)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Reverts 259 files that carried unrelated upstream changes through the
phase-6 merge: enum class removals (cppcoreguidelines-use-enum-class),
scoped_lock→lock_guard conversions (modernize-use-scoped-lock),
nodestore Backend API changes (void const* key), .clang-tidy config,
test infrastructure deletions, and miscellaneous develop changes.
These changes belong on develop, not in the telemetry PR chain.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Revert reusable-build-test-config.yml to develop (action SHA update
and "Show test failure summary" step removal don't belong here)
- Revert upload-conan-deps.yml to develop (action SHA update)
- Revert features.macro: BatchInnerSigs and Batch back to Supported::no
(these feature flag changes are unrelated to telemetry)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Separate local declarations from assignments to avoid hiding errors,
and use [[ instead of [ for non-POSIX comparisons.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Prettier formatting for markdown docs and OTelCollector header
- docs.sh rippled→xrpld renames in OTelCollector.cpp comments/strings
- Updated levelization ordering with new dependency edges
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add [[nodiscard]] to getConsensusTraceStrategy, getYays, getNays
- Add missing <string>, SpanGuard.h, SpanNames.h includes
- Fix widening cast placement (cast before arithmetic, not after)
- Replace nested ternary with lambda for const dir variable
- Add braces to if/else-if chains in Consensus.h
- Concatenate nested namespaces in ConsensusSpanNames.h
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Concatenate nested namespaces in SpanNames.h, RpcSpanNames.h, GrpcSpanNames.h
- Remove unused InfoSub.h and NetworkOPs.h includes from RPCHandler.cpp
- Add missing <string_view> includes in RPCHandler.cpp and GRPCServer.cpp
- Replace nested ternary with if/else-if in RPCHandler.cpp
- Add IWYU pragma keep for json_body.h in ServerHandler.cpp
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Clang-tidy misc-include-cleaner requires direct includes for all used
symbols. Application.cpp calls toBase58(TokenType::NodePublic, ...) at
line 1359 but did not directly include PublicKey.h.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Document the propagation infrastructure: send-side injection in
NetworkOPs/RCLConsensus, receive-side extraction in PeerImp via
PropagationHelpers.h and ConsensusReceiveTracing.h. Update
consensus receive span descriptions to reflect parent extraction.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add [[maybe_unused]] to RAII spans in TxQ.cpp
- Include Telemetry.h in RCLConsensus.cpp for complete type
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fix quorum attribute to use actual validator quorum instead of proposer
count, add missing ConsensusState::Expired handling in haveConsensus()
span, move ConsensusSpanNames.h to xrpld/consensus/ to resolve
levelization cycle, remove unused constants, enrich proposal receive
span with sequence, and correct stale documentation references.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add cons_span::event namespace with disputeResolve and txIncluded
constants; replace hardcoded strings in Consensus.h and RCLConsensus.cpp
- Move proposal.receive and validation.receive spans in PeerImp into
shared_ptr captured by job lambdas so they measure checkPropose and
checkValidation timing, not just message parsing
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update Phase4_taskList.md and 06-implementation-phases.md to reflect
completed implementation of all remaining Phase 4/4a tasks (4.2-4.6,
4a.5, 4a.6, 4a.8). Update exit criteria and summary tables.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implement remaining Phase 4/4a consensus tracing tasks:
- Add consensus.phase.open span (open → closeLedger lifecycle)
- Add consensus.proposal.receive span in PeerImp with trusted attr
- Add consensus.validation.receive span in PeerImp with trusted/seq attrs
- Add tx_count attr on accept.apply, disputes_count on update_positions
- Add tx.included events with txId in doAccept transaction loop
- Enhance dispute.resolve event with yays/nays fields
- Add avalanche_threshold attr on update_positions span
- Reparent accept/accept.apply as children of round span via childSpan()
Also adds compile-time constants in ConsensusSpanNames.h and updates
the span hierarchy diagram.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The 4-arg hashSpan overload was duplicated during a prior rebase
cascade — it appeared at both line 240 and line 305 in SpanGuard.cpp.
This would cause a linker error (multiple definition).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Record the close time voting threshold and consensus state on
consensus.update_positions and consensus.check spans:
- xrpl.consensus.close_time_threshold: the avCT_CONSENSUS_PCT (75%)
threshold required for close time agreement
- xrpl.consensus.have_close_time_consensus: whether validators
reached close time consensus in this iteration
These attributes enable dashboards to show how the close time
voting process converges (or stalls) across consensus iterations.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Remove the span-replacement logic in startRoundTracing() that was
discarding the hash-derived round span and replacing it with a linked
span (which gets a random trace_id). The deterministic trace_id from
the ledger hash is the key feature enabling cross-node correlation —
replacing it broke correlation on all rounds after the first.
Also: use thread_local mt19937 for hashSpan() span IDs (same fix as
phase-3 txSpan), add Doxygen to establish tracing method declarations
in Consensus.h, and update SpanGuard.h diagram with hashSpan/addEvent.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Instrument the consensus subsystem with OpenTelemetry spans covering
the full round lifecycle: round start, establish phase, proposal send,
ledger close, position updates, consensus check, accept, validation
send, and mode changes.
Key design choices adapted from the original Phase 4 implementation
to the new SpanGuard factory pattern introduced in Phase 3:
- Add SpanGuard::hashSpan() for category-gated hash-derived trace IDs
(consensus round spans share trace_id across validators via ledger hash)
- Add SpanGuard::addEvent() overload with key-value attribute pairs
(used for dispute.resolve events during position updates)
- Add ConsensusSpanNames.h with compile-time span name constants
following the colocated *SpanNames.h pattern from Phase 3
- Add consensusTraceStrategy config option ("deterministic"/"attribute")
for cross-node trace correlation strategy selection
- Use SpanGuard::linkedSpan() for follows-from relationships between
consecutive rounds and cross-thread validation spans
- Use SpanGuard::captureContext() for thread-safe context propagation
from consensus thread to jtACCEPT worker thread
Spans produced: consensus.round, consensus.proposal.send,
consensus.ledger_close, consensus.establish, consensus.update_positions,
consensus.check, consensus.accept, consensus.accept.apply,
consensus.validation.send, consensus.mode_change
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Wire trace context into P2P message flow so distributed traces
link across nodes. TX relay injects SpanGuard context via
PropagationHelpers.h; consensus propose/validate injects via
TraceContextPropagator.h. Receive-side extraction in PeerImp
creates child spans for proposals and validations.
- Add TraceBytes struct and SpanGuard::getTraceBytes() for
extracting raw trace context without OTel type dependencies
- Add PropagationHelpers.h: injectSpanContext(SpanGuard, proto)
- Add ConsensusReceiveTracing.h: proposalReceiveSpan(),
validationReceiveSpan() with parent context extraction
- NetworkOPs::apply(): inject tx.process context before relay
- RCLConsensus::propose()/validate(): inject active span context
- PeerImp: create receive spans for proposals and validations
with sender's trace context as parent
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move TxQSpanNames.h include to correct alphabetical position, update
levelization results for new xrpld.telemetry module dependencies,
and apply rename script to docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- tx.receive span in PeerImp: convert to shared_ptr, capture in
checkTransaction lambda so it measures actual processing, not just
message parsing
- tx.process span in NetworkOPs: convert to shared_ptr, store in
TransactionStatus so it lives until the batch job processes the entry;
sync path unchanged (span destructs on function return)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace SpanGuard::txSpan(prefix, name, hash) with the generic
SpanGuard::hashSpan(TraceCategory, name, hash) that accepts a
TraceCategory parameter instead of hardcoding Transactions. This
enables reuse for consensus round spans (Phase 4) and any future
subsystem needing deterministic cross-node trace correlation via
hash-derived trace IDs.
Both overloads are replaced:
- hashSpan(cat, name, hash, size) — standalone with random span_id
- hashSpan(cat, name, hash, size, parentSpanId, parentSize, flags)
— with remote parent from protobuf context propagation
Add full span name constants (tx_span::receive, tx_span::process)
to TxSpanNames.h following the ConsensusSpanNames.h pattern.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Mark local variables in extractFromProtobuf() and injectToProtobuf()
as const since they are not modified after initialization: traceId,
spanId, flags, spanCtx, and span.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace thread_local mt19937 with xrpl::default_prng() for span ID
generation — uses the project's existing thread-local xor-shift engine.
One call yields a uint64_t (8 bytes), filling the span ID in a single
memcpy without loops.
Fix compilation failure when XRPL_ENABLE_TELEMETRY is not defined:
move xrpl.pb.h include outside the #ifdef guard in TxTracing.h since
protocol::TMTransaction is used unconditionally in the function
signature.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace per-call std::random_device with thread_local std::mt19937 in
txSpan() for span ID generation. random_device is ~423x slower due to
/dev/urandom syscalls on each construction; mt19937 is seeded once per
thread and reused for all subsequent span IDs.
Update the SpanGuard class ASCII diagram to include txSpan factory
methods that were added in the hash-derived trace ID commit.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move TxSpanNames.h and TxQSpanNames.h from src/xrpld/telemetry/ to sit
next to the classes they instrument, matching the PathFindSpanNames.h
convention.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Derive trace_id from txHash[0:16] so all nodes handling the same
transaction produce spans under the same trace. Protobuf span_id
propagation provides parent-child relay ordering when available.
- Add SpanGuard::txSpan() factory methods (hash-derived trace ID)
- Add TxTracing.h helpers: txReceiveSpan(), txProcessSpan()
- Update PeerImp and NetworkOPs to use the new helpers
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add trace_id = txHash[0:16] strategy so all nodes handling the same
transaction independently produce spans under the same trace_id,
combined with protobuf span_id propagation for parent-child ordering.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move scattered string literals from PeerImp.cpp and NetworkOPs.cpp into
compile-time constants in src/xrpld/telemetry/TxSpanNames.h. Follows
the same StaticStr/join() pattern established in Phase 1c for RPC spans.
Constants cover: span prefixes (tx), operations (receive, process),
attribute keys (hash, local, path, suppressed, status, peerId,
peerVersion), and values (sync, async, knownBad).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace references to old XRPL_TRACE_TX/CONSENSUS macros with
SpanGuard::span(TraceCategory, ...) factory calls introduced in Phase 1c.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds xrpl.peer.version attribute to tx.receive spans for version-mismatch
correlation during network upgrades.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add "hicpp" to cspell dictionary for NOLINT annotations
- Concatenate nested namespaces in RpcSpanNames.h
- Fix include hygiene and nested ternary in RPCHandler.cpp
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Document the propagation infrastructure: send-side injection in
NetworkOPs/RCLConsensus, receive-side extraction in PeerImp via
PropagationHelpers.h and ConsensusReceiveTracing.h. Update
consensus receive span descriptions to reflect parent extraction.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fix quorum attribute to use actual validator quorum instead of proposer
count, add missing ConsensusState::Expired handling in haveConsensus()
span, move ConsensusSpanNames.h to xrpld/consensus/ to resolve
levelization cycle, remove unused constants, enrich proposal receive
span with sequence, and correct stale documentation references.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add cons_span::event namespace with disputeResolve and txIncluded
constants; replace hardcoded strings in Consensus.h and RCLConsensus.cpp
- Move proposal.receive and validation.receive spans in PeerImp into
shared_ptr captured by job lambdas so they measure checkPropose and
checkValidation timing, not just message parsing
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update Phase4_taskList.md and 06-implementation-phases.md to reflect
completed implementation of all remaining Phase 4/4a tasks (4.2-4.6,
4a.5, 4a.6, 4a.8). Update exit criteria and summary tables.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implement remaining Phase 4/4a consensus tracing tasks:
- Add consensus.phase.open span (open → closeLedger lifecycle)
- Add consensus.proposal.receive span in PeerImp with trusted attr
- Add consensus.validation.receive span in PeerImp with trusted/seq attrs
- Add tx_count attr on accept.apply, disputes_count on update_positions
- Add tx.included events with txId in doAccept transaction loop
- Enhance dispute.resolve event with yays/nays fields
- Add avalanche_threshold attr on update_positions span
- Reparent accept/accept.apply as children of round span via childSpan()
Also adds compile-time constants in ConsensusSpanNames.h and updates
the span hierarchy diagram.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The 4-arg hashSpan overload was duplicated during a prior rebase
cascade — it appeared at both line 240 and line 305 in SpanGuard.cpp.
This would cause a linker error (multiple definition).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Record the close time voting threshold and consensus state on
consensus.update_positions and consensus.check spans:
- xrpl.consensus.close_time_threshold: the avCT_CONSENSUS_PCT (75%)
threshold required for close time agreement
- xrpl.consensus.have_close_time_consensus: whether validators
reached close time consensus in this iteration
These attributes enable dashboards to show how the close time
voting process converges (or stalls) across consensus iterations.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Remove the span-replacement logic in startRoundTracing() that was
discarding the hash-derived round span and replacing it with a linked
span (which gets a random trace_id). The deterministic trace_id from
the ledger hash is the key feature enabling cross-node correlation —
replacing it broke correlation on all rounds after the first.
Also: use thread_local mt19937 for hashSpan() span IDs (same fix as
phase-3 txSpan), add Doxygen to establish tracing method declarations
in Consensus.h, and update SpanGuard.h diagram with hashSpan/addEvent.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Instrument the consensus subsystem with OpenTelemetry spans covering
the full round lifecycle: round start, establish phase, proposal send,
ledger close, position updates, consensus check, accept, validation
send, and mode changes.
Key design choices adapted from the original Phase 4 implementation
to the new SpanGuard factory pattern introduced in Phase 3:
- Add SpanGuard::hashSpan() for category-gated hash-derived trace IDs
(consensus round spans share trace_id across validators via ledger hash)
- Add SpanGuard::addEvent() overload with key-value attribute pairs
(used for dispute.resolve events during position updates)
- Add ConsensusSpanNames.h with compile-time span name constants
following the colocated *SpanNames.h pattern from Phase 3
- Add consensusTraceStrategy config option ("deterministic"/"attribute")
for cross-node trace correlation strategy selection
- Use SpanGuard::linkedSpan() for follows-from relationships between
consecutive rounds and cross-thread validation spans
- Use SpanGuard::captureContext() for thread-safe context propagation
from consensus thread to jtACCEPT worker thread
Spans produced: consensus.round, consensus.proposal.send,
consensus.ledger_close, consensus.establish, consensus.update_positions,
consensus.check, consensus.accept, consensus.accept.apply,
consensus.validation.send, consensus.mode_change
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Wire trace context into P2P message flow so distributed traces
link across nodes. TX relay injects SpanGuard context via
PropagationHelpers.h; consensus propose/validate injects via
TraceContextPropagator.h. Receive-side extraction in PeerImp
creates child spans for proposals and validations.
- Add TraceBytes struct and SpanGuard::getTraceBytes() for
extracting raw trace context without OTel type dependencies
- Add PropagationHelpers.h: injectSpanContext(SpanGuard, proto)
- Add ConsensusReceiveTracing.h: proposalReceiveSpan(),
validationReceiveSpan() with parent context extraction
- NetworkOPs::apply(): inject tx.process context before relay
- RCLConsensus::propose()/validate(): inject active span context
- PeerImp: create receive spans for proposals and validations
with sender's trace context as parent
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move TxQSpanNames.h include to correct alphabetical position, update
levelization results for new xrpld.telemetry module dependencies,
and apply rename script to docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- tx.receive span in PeerImp: convert to shared_ptr, capture in
checkTransaction lambda so it measures actual processing, not just
message parsing
- tx.process span in NetworkOPs: convert to shared_ptr, store in
TransactionStatus so it lives until the batch job processes the entry;
sync path unchanged (span destructs on function return)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace SpanGuard::txSpan(prefix, name, hash) with the generic
SpanGuard::hashSpan(TraceCategory, name, hash) that accepts a
TraceCategory parameter instead of hardcoding Transactions. This
enables reuse for consensus round spans (Phase 4) and any future
subsystem needing deterministic cross-node trace correlation via
hash-derived trace IDs.
Both overloads are replaced:
- hashSpan(cat, name, hash, size) — standalone with random span_id
- hashSpan(cat, name, hash, size, parentSpanId, parentSize, flags)
— with remote parent from protobuf context propagation
Add full span name constants (tx_span::receive, tx_span::process)
to TxSpanNames.h following the ConsensusSpanNames.h pattern.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Mark local variables in extractFromProtobuf() and injectToProtobuf()
as const since they are not modified after initialization: traceId,
spanId, flags, spanCtx, and span.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace thread_local mt19937 with xrpl::default_prng() for span ID
generation — uses the project's existing thread-local xor-shift engine.
One call yields a uint64_t (8 bytes), filling the span ID in a single
memcpy without loops.
Fix compilation failure when XRPL_ENABLE_TELEMETRY is not defined:
move xrpl.pb.h include outside the #ifdef guard in TxTracing.h since
protocol::TMTransaction is used unconditionally in the function
signature.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace per-call std::random_device with thread_local std::mt19937 in
txSpan() for span ID generation. random_device is ~423x slower due to
/dev/urandom syscalls on each construction; mt19937 is seeded once per
thread and reused for all subsequent span IDs.
Update the SpanGuard class ASCII diagram to include txSpan factory
methods that were added in the hash-derived trace ID commit.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move TxSpanNames.h and TxQSpanNames.h from src/xrpld/telemetry/ to sit
next to the classes they instrument, matching the PathFindSpanNames.h
convention.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Derive trace_id from txHash[0:16] so all nodes handling the same
transaction produce spans under the same trace. Protobuf span_id
propagation provides parent-child relay ordering when available.
- Add SpanGuard::txSpan() factory methods (hash-derived trace ID)
- Add TxTracing.h helpers: txReceiveSpan(), txProcessSpan()
- Update PeerImp and NetworkOPs to use the new helpers
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add trace_id = txHash[0:16] strategy so all nodes handling the same
transaction independently produce spans under the same trace_id,
combined with protobuf span_id propagation for parent-child ordering.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move scattered string literals from PeerImp.cpp and NetworkOPs.cpp into
compile-time constants in src/xrpld/telemetry/TxSpanNames.h. Follows
the same StaticStr/join() pattern established in Phase 1c for RPC spans.
Constants cover: span prefixes (tx), operations (receive, process),
attribute keys (hash, local, path, suppressed, status, peerId,
peerVersion), and values (sync, async, knownBad).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace references to old XRPL_TRACE_TX/CONSENSUS macros with
SpanGuard::span(TraceCategory, ...) factory calls introduced in Phase 1c.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds xrpl.peer.version attribute to tx.receive spans for version-mismatch
correlation during network upgrades.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add [[maybe_unused]] to RAII span variables in PathFind/RipplePathFind
handlers (Clang -Wunused-variable with -Werror)
- Restore over-renamed values: rippledb, rippled.cfg, historical GitHub URL
- Concatenate nested namespaces in SpanNames.h and PathFindSpanNames.h
(modernize-concat-nested-namespaces)
- Add missing includes and const qualifiers in test files
- Suppress intentional use-after-move in SpanGuardFactory move test
- Remove unused NetworkOPs.h include from PathRequest.cpp
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Document the propagation infrastructure: send-side injection in
NetworkOPs/RCLConsensus, receive-side extraction in PeerImp via
PropagationHelpers.h and ConsensusReceiveTracing.h. Update
consensus receive span descriptions to reflect parent extraction.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Wire trace context into P2P message flow so distributed traces
link across nodes. TX relay injects SpanGuard context via
PropagationHelpers.h; consensus propose/validate injects via
TraceContextPropagator.h. Receive-side extraction in PeerImp
creates child spans for proposals and validations.
- Add TraceBytes struct and SpanGuard::getTraceBytes() for
extracting raw trace context without OTel type dependencies
- Add PropagationHelpers.h: injectSpanContext(SpanGuard, proto)
- Add ConsensusReceiveTracing.h: proposalReceiveSpan(),
validationReceiveSpan() with parent context extraction
- NetworkOPs::apply(): inject tx.process context before relay
- RCLConsensus::propose()/validate(): inject active span context
- PeerImp: create receive spans for proposals and validations
with sender's trace context as parent
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The StatsD receiver config was lost during a branch rebase (--ours
conflict resolution dropped it). Re-add the statsd receiver to the
OTel Collector config and wire it into the metrics pipeline so
beast::insight UDP metrics flow to Prometheus.
Also fixes:
- Metric prefix mismatch: docs used xrpld_ but dashboards/tests use
rippled_ — align all documentation to match the runnable stack
- Remove phantom Peer_Disconnects_Charges from docs (plain atomic,
not a beast::insight gauge)
- Remove premature .codecov.yml exclusions for Phase 7 OTelCollector
files that don't exist on this branch
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add missing xrpl.consensus.quorum attribute to consensus.accept in runbook
- Fix dashboard legend formats: add exported_instance, use Title Case
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add 14 missing spans to runbook (6 TxQ + 8 consensus)
- Fix tx.receive attributes and config table in runbook
- Document dispute.resolve and tx.included span events
- Add spanmetrics dimensions for close_time_correct and tx.suppressed
- Fix Close Time Agreement and TX Receive vs Suppressed panel PromQL
- Wire $consensus_mode template variable to all consensus panels
- Add 10 Tempo search filters for operational attributes
- Apply rename script artifacts
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Document service_instance_id, use_tls, tls_ca_cert, batch_size,
batch_delay_ms, and max_queue_size in the [telemetry] section.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace remaining rippled/Ripple references with xrpld/XRPL in
data collection reference, implementation phases, and runbook docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Resolve merge conflicts taking phase 4 consensus span improvements,
fix bashate indentation in integration test script, and apply rename
script to Phase5 integration test docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move TxQSpanNames.h include to correct alphabetical position, update
levelization results for new xrpld.telemetry module dependencies,
and apply rename script to docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add unknownCommand and wsUpgrade span name constants to RpcSpanNames.h,
fix SpanGuardFactory tests to use the 3-argument SpanGuard::span() API,
update levelization results, and apply rename script to docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Use C++17 concatenated namespaces, add [[nodiscard]] to query methods,
add missing direct includes, and use pass-by-value + std::move in
NullTelemetry constructor.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fix quorum attribute to use actual validator quorum instead of proposer
count, add missing ConsensusState::Expired handling in haveConsensus()
span, move ConsensusSpanNames.h to xrpld/consensus/ to resolve
levelization cycle, remove unused constants, enrich proposal receive
span with sequence, and correct stale documentation references.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- tx.receive span in PeerImp: convert to shared_ptr, capture in
checkTransaction lambda so it measures actual processing, not just
message parsing
- tx.process span in NetworkOPs: convert to shared_ptr, store in
TransactionStatus so it lives until the batch job processes the entry;
sync path unchanged (span destructs on function return)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add cons_span::event namespace with disputeResolve and txIncluded
constants; replace hardcoded strings in Consensus.h and RCLConsensus.cpp
- Move proposal.receive and validation.receive spans in PeerImp into
shared_ptr captured by job lambdas so they measure checkPropose and
checkValidation timing, not just message parsing
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update Phase4_taskList.md and 06-implementation-phases.md to reflect
completed implementation of all remaining Phase 4/4a tasks (4.2-4.6,
4a.5, 4a.6, 4a.8). Update exit criteria and summary tables.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implement remaining Phase 4/4a consensus tracing tasks:
- Add consensus.phase.open span (open → closeLedger lifecycle)
- Add consensus.proposal.receive span in PeerImp with trusted attr
- Add consensus.validation.receive span in PeerImp with trusted/seq attrs
- Add tx_count attr on accept.apply, disputes_count on update_positions
- Add tx.included events with txId in doAccept transaction loop
- Enhance dispute.resolve event with yays/nays fields
- Add avalanche_threshold attr on update_positions span
- Reparent accept/accept.apply as children of round span via childSpan()
Also adds compile-time constants in ConsensusSpanNames.h and updates
the span hierarchy diagram.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sort PathFindSpanNames.h after AssetCache.h alphabetically in
PathRequestManager.cpp and Pathfinder.cpp to satisfy the project's
include-order convention enforced by pre-commit hooks.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move SpanGuard.h (associated header) to first include position,
separated by blank line from other project includes, per the
project's include-order convention enforced by pre-commit hooks.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move Telemetry.h (associated header) to first include position in
Telemetry.cpp per the project's include-order convention. Trim
trailing whitespace from POC_taskList.md markdown table columns.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add 5 new panels to the consensus-health Grafana dashboard using Tempo
TraceQL queries against consensus.accept.apply span attributes:
- Close Time: Raw Proposals (Per Node) — each node's unrounded
wall-clock close_time_self, reveals clock drift across validators
- Close Time: Effective / Quantized — the consensus-agreed close_time
after rounding to resolution bins, written to ledger header
- Close Time Vote Bins & Resolution — number of distinct vote bins
(close_time_vote_bins) and bin size (close_resolution_ms) on dual axes
- Close Time Resolution Direction — whether resolution increased
(coarser), decreased (finer), or stayed unchanged
- Close Time Bin Distribution — bar chart showing how raw proposals
distribute across quantized bins per round
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Integrate the existing StatsD metrics pipeline (beast::insight) into
the OpenTelemetry observability stack and add new trace spans for
ledger build/store/validate and peer proposal/validation receive.
Phase 5b — Ledger, peer, and transaction spans:
- Add ledger.build span with close time attributes in BuildLedger.cpp
- Add tx.apply span with tx_count/tx_failed in BuildLedger.cpp
- Add ledger.store and ledger.validate spans in LedgerMaster.cpp
- Add peer.proposal.receive span with trusted attribute in PeerImp.cpp
- Add peer.validation.receive span with ledger_hash, full, trusted
attributes in PeerImp.cpp
- Add ledger-operations and peer-network Grafana dashboards
Phase 6 — StatsD metrics integration:
- Add StatsD UDP receiver (port 8125) to OTel Collector
- Add 5 StatsD Grafana dashboards: node health, network traffic,
overlay traffic detail, ledger data sync, RPC pathfinding
- Add 09-data-collection-reference.md cataloging all metrics/spans
- Update existing dashboards with new span panels
- Expand telemetry runbook and integration test script
- Add codecov exclusions for telemetry modules
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Run .github/scripts/rename/docs.sh to replace rippled → xrpld
references in TESTING.md, xrpld-telemetry.cfg, and
telemetry-runbook.md, fixing the check-rename CI failure.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Mark Tasks 5.3 (alert definitions) and 5.6 (training materials) as
"Deferred — post-MVP" in the implementation phases document to
accurately reflect current delivery scope. Add status column to the
Phase 5 task table.
Also fix stale reference to XRPL_TRACE_* macros in Phase 4a section —
the implementation uses SpanGuard factory methods.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The 4-arg hashSpan overload was duplicated during a prior rebase
cascade — it appeared at both line 240 and line 305 in SpanGuard.cpp.
This would cause a linker error (multiple definition).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Record the close time voting threshold and consensus state on
consensus.update_positions and consensus.check spans:
- xrpl.consensus.close_time_threshold: the avCT_CONSENSUS_PCT (75%)
threshold required for close time agreement
- xrpl.consensus.have_close_time_consensus: whether validators
reached close time consensus in this iteration
These attributes enable dashboards to show how the close time
voting process converges (or stalls) across consensus iterations.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Remove the span-replacement logic in startRoundTracing() that was
discarding the hash-derived round span and replacing it with a linked
span (which gets a random trace_id). The deterministic trace_id from
the ledger hash is the key feature enabling cross-node correlation —
replacing it broke correlation on all rounds after the first.
Also: use thread_local mt19937 for hashSpan() span IDs (same fix as
phase-3 txSpan), add Doxygen to establish tracing method declarations
in Consensus.h, and update SpanGuard.h diagram with hashSpan/addEvent.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Instrument the consensus subsystem with OpenTelemetry spans covering
the full round lifecycle: round start, establish phase, proposal send,
ledger close, position updates, consensus check, accept, validation
send, and mode changes.
Key design choices adapted from the original Phase 4 implementation
to the new SpanGuard factory pattern introduced in Phase 3:
- Add SpanGuard::hashSpan() for category-gated hash-derived trace IDs
(consensus round spans share trace_id across validators via ledger hash)
- Add SpanGuard::addEvent() overload with key-value attribute pairs
(used for dispute.resolve events during position updates)
- Add ConsensusSpanNames.h with compile-time span name constants
following the colocated *SpanNames.h pattern from Phase 3
- Add consensusTraceStrategy config option ("deterministic"/"attribute")
for cross-node trace correlation strategy selection
- Use SpanGuard::linkedSpan() for follows-from relationships between
consecutive rounds and cross-thread validation spans
- Use SpanGuard::captureContext() for thread-safe context propagation
from consensus thread to jtACCEPT worker thread
Spans produced: consensus.round, consensus.proposal.send,
consensus.ledger_close, consensus.establish, consensus.update_positions,
consensus.check, consensus.accept, consensus.accept.apply,
consensus.validation.send, consensus.mode_change
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace SpanGuard::txSpan(prefix, name, hash) with the generic
SpanGuard::hashSpan(TraceCategory, name, hash) that accepts a
TraceCategory parameter instead of hardcoding Transactions. This
enables reuse for consensus round spans (Phase 4) and any future
subsystem needing deterministic cross-node trace correlation via
hash-derived trace IDs.
Both overloads are replaced:
- hashSpan(cat, name, hash, size) — standalone with random span_id
- hashSpan(cat, name, hash, size, parentSpanId, parentSize, flags)
— with remote parent from protobuf context propagation
Add full span name constants (tx_span::receive, tx_span::process)
to TxSpanNames.h following the ConsensusSpanNames.h pattern.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Mark local variables in extractFromProtobuf() and injectToProtobuf()
as const since they are not modified after initialization: traceId,
spanId, flags, spanCtx, and span.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace thread_local mt19937 with xrpl::default_prng() for span ID
generation — uses the project's existing thread-local xor-shift engine.
One call yields a uint64_t (8 bytes), filling the span ID in a single
memcpy without loops.
Fix compilation failure when XRPL_ENABLE_TELEMETRY is not defined:
move xrpl.pb.h include outside the #ifdef guard in TxTracing.h since
protocol::TMTransaction is used unconditionally in the function
signature.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace per-call std::random_device with thread_local std::mt19937 in
txSpan() for span ID generation. random_device is ~423x slower due to
/dev/urandom syscalls on each construction; mt19937 is seeded once per
thread and reused for all subsequent span IDs.
Update the SpanGuard class ASCII diagram to include txSpan factory
methods that were added in the hash-derived trace ID commit.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move TxSpanNames.h and TxQSpanNames.h from src/xrpld/telemetry/ to sit
next to the classes they instrument, matching the PathFindSpanNames.h
convention.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Derive trace_id from txHash[0:16] so all nodes handling the same
transaction produce spans under the same trace. Protobuf span_id
propagation provides parent-child relay ordering when available.
- Add SpanGuard::txSpan() factory methods (hash-derived trace ID)
- Add TxTracing.h helpers: txReceiveSpan(), txProcessSpan()
- Update PeerImp and NetworkOPs to use the new helpers
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add trace_id = txHash[0:16] strategy so all nodes handling the same
transaction independently produce spans under the same trace_id,
combined with protobuf span_id propagation for parent-child ordering.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move scattered string literals from PeerImp.cpp and NetworkOPs.cpp into
compile-time constants in src/xrpld/telemetry/TxSpanNames.h. Follows
the same StaticStr/join() pattern established in Phase 1c for RPC spans.
Constants cover: span prefixes (tx), operations (receive, process),
attribute keys (hash, local, path, suppressed, status, peerId,
peerVersion), and values (sync, async, knownBad).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace references to old XRPL_TRACE_TX/CONSENSUS macros with
SpanGuard::span(TraceCategory, ...) factory calls introduced in Phase 1c.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds xrpl.peer.version attribute to tx.receive spans for version-mismatch
correlation during network upgrades.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Instrument the path finding subsystem with full span coverage:
- pathfind.request: wraps doPathFind() and doRipplePathFind() RPC handlers
- pathfind.compute: wraps PathRequest::doUpdate() with fast/normal attr
- pathfind.update_all: wraps PathRequestManager::updateAll() on ledger
close with ledger_index attr
- pathfind.discover: wraps Pathfinder::findPaths() graph exploration
with search_level attr
- pathfind.rank: wraps Pathfinder::computePathRanks() liquidity
validation with num_paths attr
New file: PathFindSpanNames.h with compile-time constants following
the StaticStr/join() pattern from Phase 1c.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Move node health attribute strings to compile-time constants in
SpanNames.h (attr::nodeAmendmentBlocked, attr::nodeServerState)
- Add Tempo search filters for node health attributes
- Remove unnecessary .c_str() on strOperatingMode() return
- Add samplingRatio clamping test (values > 1.0 and < 0.0)
- Fix Task 2.3 status: delivered in Phase 1c, not Phase 2
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- serviceName default is "xrpld" not "rippled"
- Remove references to nonexistent exporterType field
- Pass networkId (4th param) to setup_Telemetry()
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Mark deferred tasks (2.1→Phase 3, 2.5→low priority) with rationale.
Mark superseded tasks (2.2→Phase 1c SpanGuard factory). Add Task 2.7
for Grafana search filters. Update summary table with status column.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Grafana Tempo datasource: add rpc-command, rpc-status, rpc-role
search filters for the Explore UI
- Unit tests: TelemetryConfig (config parsing defaults and sections),
SpanGuardFactory (null guard safety, move semantics, discard, all
factory methods)
- Test CMake registration with optional OTel linking
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add amendment_blocked and server_state span attributes to every
rpc.command.* span so operators can correlate RPC behavior with node state.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Introduces task list documents for Phases 2 through 5, with Tempo
references (replacing Jaeger) and Task 2.8 dashboard parity spec.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Centralise scattered string literals into compile-time constants using
StaticStr<N> and join() for dot-separated composition. Shared primitives
live in SpanNames.h; RPC-specific names in RpcSpanNames.h. Future modules
(consensus, peer, ledger) add their own *SpanNames.h without bloating
the central header.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add [[maybe_unused]] to the RAII span in processSession() — the
variable is not read but its lifetime scopes the active OTel context
for child spans created in processRequest()
- Regenerate levelization: remove premature xrpld.telemetry entries
that reference a module not yet present on this branch
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Use grpc_span::val::resourceExhausted constant instead of raw
"resource_exhausted" string in GRPCServer.cpp
- Fix unbounded span name cardinality in RPCHandler.cpp error path:
use fixed rpc_span::val::unknownCommand as span name instead of
user-supplied cmdName (attacker-controlled input). The actual
command is still captured in the xrpl.rpc.command attribute.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Without this, cloned CallData instances (created for the next incoming
gRPC request) would have an empty name_, making subsequent span attrs
blank.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Follow project convention (PerfLog.h, SpanGuard.h) for documentation
diagrams. Show HTTP single, HTTP batch, and WebSocket span nesting.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Document the span hierarchy, covered paths, and known instrumentation
gaps directly in the header that developers reference when adding spans.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Centralise scattered string literals into compile-time constants using
StaticStr<N> and join() for dot-separated composition. Shared primitives
live in SpanNames.h; RPC-specific names in RpcSpanNames.h. Future modules
(consensus, peer, ledger) add their own *SpanNames.h without bloating
the central header.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace rpcSpan(fullName) calls with span(TraceCategory::Rpc, prefix,
name). Add 'using namespace telemetry' to both RPC files so call sites
read cleanly without repeated namespace qualifiers.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Delete TracingInstrumentation.h and replace all XRPL_TRACE_* macro
invocations with direct SpanGuard::rpcSpan() calls. SpanGuard's pimpl
design and global Telemetry accessor eliminate the need for macro
wrappers and explicit Telemetry instance passing at call sites.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
TracingInstrumentation.h:
- Unify all span-creation macros to use std::optional<SpanGuard>
(fixes type mismatch between XRPL_TRACE_SPAN and SET_ATTR)
- Wrap XRPL_TRACE_SET_ATTR/EXCEPTION in do-while(0) (dangling-else)
- Move macros outside namespace blocks (macros are global)
- Cache telemetry reference to avoid double-evaluation
- Remove leaked _xrpl_span_ intermediate variable
- Add @note tags for thread safety, scope, and usage constraints
- Add 3 usage examples per CLAUDE.md requirements
ServerHandler.cpp:
- Remove misleading rpc.request span from onRequest() (span ended
before coroutine runs, producing orphan spans)
- Add rpc.http_request span to HTTP processSession() (runs inside
the coroutine, correct parent for rpc.process/rpc.command spans)
- Add XRPL_TRACE_EXCEPTION and error status in both catch blocks
(WS processSession and processRequest)
SpanGuard.h:
- Add null guards to all mutating methods (setOk, setStatus,
setAttribute, addEvent, recordException) for safety after discard()
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Concatenate nested namespaces (modernize-concat-nested-namespaces)
- Add [[nodiscard]] to factory and accessor methods
- NOLINT no-op stub instance methods that must stay non-static for API
parity with the real implementation
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove the single-arg span(name) factory that creates unconditional
spans without category gating. All call sites use the 3-arg
span(TraceCategory, prefix, name) variant which checks whether the
category is enabled in config before creating a span. The 1-arg form
was dead code with no callers.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace references to non-existent TracingInstrumentation.h with
SpanGuard.cpp pimpl implementation that actually exists on this branch.
Update conditional compilation section to describe the pimpl approach.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace rpcSpan(), txSpan(), consensusSpan(), peerSpan(), ledgerSpan()
with a single span(TraceCategory, prefix, name) factory method. Adding
a new traceable subsystem now requires only a new enum value and one
switch case — no new methods or header changes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- SpanContext::isValid(): add inline no-op when XRPL_ENABLE_TELEMETRY
is not defined, preventing a linker error if called in that path
- linkedSpan(): set kIsRootSpanKey on the StartSpanOptions parent
context so linked spans start a genuinely independent sub-tree
instead of silently becoming children of the current active span
- Telemetry::instance_: use std::atomic with acquire/release ordering
to avoid a data race between start()/stop() and factory methods
called from worker threads
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Redesign SpanGuard with pimpl idiom to hide all OpenTelemetry types
from public headers. Add global Telemetry accessor so SpanGuard factory
methods work without explicit Telemetry references. Add child/linked
span creation and cross-thread context propagation. Update plan docs
to reflect macro removal in favor of SpanGuard factory pattern.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add DiscardFlag.h and FilteringSpanProcessor references to the file
tree, key files table, and implementation summary in OpenTelemetryPlan.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add span discard mechanism that drops unwanted spans before they enter
the batch export queue, saving both network bandwidth and storage.
FilteringSpanProcessor is a custom SpanProcessor decorator that wraps
BatchSpanProcessor. SpanGuard::discard() sets a thread-local flag
(tl_discardCurrentSpan) before calling Span::End(). The OTel SDK calls
OnEnd() synchronously on the same thread, where the flag is checked and
cleared to drop the span.
New file: DiscardFlag.h — zero-dependency header for the thread-local
flag, avoiding transitive include bloat from Telemetry.h.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove duplicate otlp/tempo exporter block, duplicate tempo service
definition, and jaeger dependency from docker-compose example.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Tempo is now the sole trace backend. Remove Jaeger all-in-one service
from docker-compose, otlp/jaeger exporter from OTel Collector config,
and Jaeger Grafana datasource provisioning file.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Run .github/scripts/rename/docs.sh to replace rippled → xrpld
references in all plan documentation files, fixing the check-rename
CI failure.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add Sub-task 7.10a: Per-Validator Validation Count (Flag Ledger Window)
to the Phase 7 task list. This metric tracks how many of the last 256
ledgers each UNL validator has validated — the key participation metric
for UNL health monitoring.
Implementation plan:
- Observable gauge rippled_validator_participation with validator label
- Data from RCLValidations::getTrustedForLedger() over 256-ledger window
- Emitted at flag ledger boundaries (~15 min interval)
- Grafana table panel with threshold coloring (green/yellow/red)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Populate baselines/baseline-timings.json from the green CI run
(24906110133, commit f11ebc1253). 25/31 metrics have non-null values;
6 span.rpc.* are null due to sparse data in the 3m window.
Remove the rpc_methods section from regression-metrics.json and its
thresholds. rippled_rpc_method_duration_us_bucket is never populated
because PerfLogImp::rpcEnd never calls MetricsRegistry::recordRpcFinished
— only recordRpcStarted is wired up (Phase 9 instrumentation gap).
The span-based rpc.request/rpc.process metrics via spanmetrics already
cover RPC latency.
Two CI failures traced to root cause:
1. rippled_jobq_job_count: 0 series — StatsDGaugeImpl declared
m_dirty{false} despite the constructor comment saying "start dirty".
Gauges whose value starts and stays at 0 never emitted, so Prometheus
never scraped them. Fix: m_dirty{true} on the member initializer.
2. TX error rate 82.8% — the submitter tracked account sequences
locally, but in a multi-node consensus network other nodes' txns
advance sequences independently. After a few ledger closes the
locally-tracked sequence fell behind the ledger, producing
tefPAST_SEQ for every subsequent submission. Fix: refresh account
sequences from account_info every 10 s during the submission loop.
- capture_timings.py: fail when captured/total ratio < 50%
(--min-capture-ratio). Prevents silent pass on unreachable Prometheus.
- run-full-validation.sh: set REGRESSION_EXIT=2 on capture failure so
the final exit code reflects it. Update exit code docs in header.
- compare_to_baseline.py: extract _skip_delta helper to bring
compute_delta under 80 lines. Fix 0.0-as-falsy bug in abs_bound
resolution (use explicit None check instead of `or`). Remove dead
variable override_prefix_key.
- prom_queries.py: extract _build_simple_entries and _build_job_entries
to bring build_query_plan under 80 lines. Fix module docstring return
type example. Use aiohttp.ClientTimeout instead of bare int.
- telemetry-validation.yml: add set -euo pipefail to regression summary
step; guard jq calls with -e flag and fallback; fail on missing
baseline file; emit ::warning annotation when timings.json missing.
- baselines/README.md: document the placeholder field.
Captures per-span / per-RPC / per-job timings from Prometheus after the
workload run and diffs them against a committed baseline. Regression
requires breaching both a percentage and an absolute bound, tolerating
small-value noise. When the baseline is a placeholder, the comparator
emits the captured JSON in the exact schema for one-time paste into
baselines/baseline-timings.json, and the CI Step Summary surfaces that
block for the reviewer.
Scope: gate only — automated baseline persistence, benchmark.sh
PromQL migration, and the historical trend dashboard remain follow-ups.
Plan documents referenced Application.h and app_ for getTelemetry()
but the codebase now uses ServiceRegistry as the interface. Updated:
- 05-configuration-reference.md: getTelemetry() on ServiceRegistry,
deferred serviceInstanceId pattern in ApplicationImp
- POC_taskList.md Task 4: target ServiceRegistry.h not Application.h,
correct config file path and constructor pattern
- 04-code-samples.md: fix overlay() -> getOverlay(), rewrite JobQueue
sample to reflect actual architecture (no app_ member)
- 03-implementation-strategy.md: fix file impact table path
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Strip effort/risk columns from task tables and remove the §6.9 Effort
Summary section with its pie chart and resource requirements table.
Renumber §6.10 Quick Wins → §6.9.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Split document index into Plan Documents and Task Lists sections.
These files were introduced in this branch but missing from the index.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The Jaeger-removal rebase used --ours conflict resolution which
dropped content added by intermediate phases (6-8): StatsD receiver,
filelog receiver, Loki service/exporter, health_check extension,
and OTLP metrics pipeline. Restore from pre-rebase origin minus
Jaeger references.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Fix remaining Jaeger references that accumulated across intermediate
branches in the stacked PR chain. These were in files modified by
multiple phases where the per-branch fixes didn't cover all additions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Migrate validate_telemetry.py to Tempo TraceQL search API, remove
Jaeger service from workload docker-compose, update readiness checks.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Pre-include boost/asio/detail/socket_types.hpp on Windows before OTel
SDK headers to ensure WinSock2.h is included before WinSock.h.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add ValidationTracker.cpp to xrpl.test.telemetry target sources
(implementation lives in src/xrpld/ but has no OTel SDK dependency)
- Change BEAST_DEFINE_TESTSUITE namespace from ripple to xrpl
- Replace recursive *.md glob with non-recursive GLOB in XrplDocs.cmake
to avoid picking up .claude/instructions.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Update MockServiceRegistry to match current ServiceRegistry interface
(17 method renames: get* prefix, PathRequests→PathRequestManager)
- Make throwUnimplemented() static to satisfy clang-tidy
- Regenerate levelization ordering.txt and loops.txt
- Remove 'rippled' prefix from 3 StatsD dashboard titles
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix validations_checked_total recording site (NetworkOPs, not LedgerMaster)
- Add Slide 11 to presentation: External Dashboard Parity overview with
Mermaid diagrams for new metric categories, ValidationTracker sequence,
and new dashboard summary
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Remove duplicate 'system-node-health' UID from expected_metrics.json
(already covered by 'rippled-system-node-health')
- Add parity span attributes to expected_spans.json: node health on
rpc.command.*, validation hash/full on consensus.validation.send,
quorum/proposers on consensus.accept, validation hash/full on
peer.validation.receive
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add agreement_pct_7d, agreements_7d, missed_7d labels to the
rippled_validation_agreement observable gauge, matching the external
xrpl-validator-dashboard's 7-day tracking.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add ValidationTracker member to MetricsRegistry with a public accessor,
register a rippled_validation_agreement observable gauge that calls
reconcile() and reports 1h/24h agreement percentages and counts, and
hook recordOurValidation/recordNetworkValidation into RCLConsensus
validate() and LedgerMaster setValidLedger() respectively.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Wire incrementValidationsChecked() into NetworkOPs::recvValidation() so
each received network validation increments the counter.
Note: incrementJqTransOverflow() hook is deferred — JobQueue has no
explicit overflow event path; the counter is reserved for future use.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add rippled_validation_agreements_total and rippled_validation_missed_total
counter declarations and creation (wiring to ValidationTracker pending rebase)
- Fix peer-quality dashboard: query rippled_server_info{metric="peer_disconnects_resources"}
instead of non-existent rippled_Overlay_Peer_Disconnects_Charges
- Remove dead getCountsJson() call in storageDetail callback
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Rename fee labels to match spec: base_fee_drops -> base_fee_xrp,
reserve_base_drops -> reserve_base_xrp, reserve_inc_drops -> reserve_inc_xrp
- Add peers_insane_count (stub with TODO for PeerImp::tracking_ exposure)
- Add transaction_rate to ledger economy gauge
- Replace node_store_writes/node_written_bytes with nudb_bytes per spec
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add validator health, peer quality, ledger economy, state tracking, and
storage detail observable gauges plus 5 synchronous counters with recording
hooks for ledger close, validation send, state change, and overflow events.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add window7d_ deque, agreementPct7d(), agreements7d(), missed7d() to
match the external xrpl-validator-dashboard's 7-day agreement tracking.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Cover normal agreement, missed validation, late repair, empty window,
grace period boundary, max pending trimming, mixed results, duplicate
recording, and only-we-validated scenarios.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Use >= instead of > for grace period comparison to reconcile at exactly
8 seconds rather than skipping the boundary
- Two-pass hard trim: first remove entries past late-repair window, then
any reconciled entry, to avoid sabotaging late repairs
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Standalone class that tracks whether this validator's validations agree
with network consensus, maintaining rolling 1h/24h windows and lifetime
totals with a late-repair mechanism for out-of-order arrivals.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds ValidationTracker (agreement computation with 8s grace period),
validator health, peer quality, ledger economy, state tracking,
storage detail gauges, 7 synchronous counters, and agreement gauge.
29 new metrics covering validation agreement, peer quality, UNL health,
ledger economy, state tracking, and upgrade awareness.
Part of the external dashboard parity initiative across phases 2-11.
See docs/superpowers/specs/2026-03-30-external-dashboard-parity-design.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Bridge the existing beast::insight gauge for resource-limit peer
disconnects (peerDisconnectsCharges_) into the StatsD metric inventory.
Part of the external dashboard parity initiative.
See docs/superpowers/specs/2026-03-30-external-dashboard-parity-design.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The xrpld/app/misc/CanonicalTXSet.h header doesn't exist — it was
incorrectly added during a rebase conflict resolution. The correct
include xrpl/ledger/CanonicalTXSet.h is already present.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add ledger hash and full-validation flag to peer.validation.receive
spans for trace-level agreement analysis across validators.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move xrpld data paths from ./data/ to docker/telemetry/data/ so runtime
files stay within the docker telemetry directory. Add .gitignore to
exclude the data directory from version control.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move validation_quorum and proposers_validated attributes from
consensus.accept.apply to consensus.accept span to match the design
spec. Both values are available in onAccept() scope.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add validation ledger hash and full-validation flag to
consensus.validation.send spans, plus quorum and proposer count to
consensus.accept spans for trace-level agreement analysis.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds ledger_hash, validation.full to validation send/receive spans,
and validation_quorum, proposers_validated to consensus.accept spans.
Foundation for Phase 7 ValidationTracker agreement computation.
Part of the external dashboard parity initiative across phases 2-11.
See docs/superpowers/specs/2026-03-30-external-dashboard-parity-design.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Tag transaction receive spans with the relaying peer's rippled version
to enable version-mismatch correlation during network upgrades.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds xrpl.peer.version attribute to tx.receive spans for version-mismatch
correlation during network upgrades.
Part of the external dashboard parity initiative across phases 2-11.
See docs/superpowers/specs/2026-03-30-external-dashboard-parity-design.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add amendment_blocked and server_state span attributes to every
rpc.command.* span so operators can correlate RPC behavior with node state.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds node health context (amendment_blocked, server_state) to rpc.command.*
spans, inspired by the community xrpl-validator-dashboard.
Part of the external dashboard parity initiative across phases 2-11.
See docs/superpowers/specs/2026-03-30-external-dashboard-parity-design.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Tempo is now the sole trace backend. Remove Jaeger all-in-one service
from docker-compose, otlp/jaeger exporter from OTel Collector config,
and Jaeger Grafana datasource provisioning file.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Strip effort/risk columns from task tables and remove the §6.9 Effort
Summary section with its pie chart and resource requirements table.
Renumber §6.10 Quick Wins → §6.9.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Split document index into Plan Documents and Task Lists sections.
These files were introduced in this branch but missing from the index.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
| A | No stray dotted span-attribute key (only the derived resource keys may be dotted). |
| G | Attribute keys are `lower_snake_case` (`^[a-z][a-z0-9_]*$` per dot-segment) — no camelCase, UPPERCASE, or spaces. |
| F | No string literals as attribute keys or span-name arguments in `setAttribute`/`addEvent`/`span`/`rootSpan`/`childSpan` (`rootSpan` shares `span`'s `(cat, prefix, name)` signature). Attribute _values_ are exempt (runtime data); `*SpanNames.h` definitions and test files are exempt. |
| B | Every collector `spanmetrics.dimensions` name exists in the L1 key set. |
| C | Every Tempo span-filter tag exists in the L1 key set. |
| D | Every dashboard label resolves to an L1 span attribute, a native-metric label (L6, emitted by MetricsRegistry), or a Prometheus/Grafana builtin. TraceQL scope prefixes (`span.`/`resource.`/…) are stripped before the L1 lookup. |
| E | No dotted `xrpl.<domain>.<field>` attribute key in the runbook (only the L1 resource attrs `xrpl.network.*` may be dotted). Span names, filenames, OTel-standard keys, and metric labels are not flagged. |
| I | No string literals as **metric** instrument names or label keys — the mirror of Rule F. Applies to the name passed to an `XRPL_METRIC_*` macro or a `meter->Create*` factory and to the label _keys_ in its label set. Label _values_, descriptions, `*MetricNames.h`, `MetricMacros.h` and test files are exempt. Scoped by metric **family** (first underscore segment): declaring a constant opts that family in, so the metric surface can be converted subsystem by subsystem. Unconverted families warn as Rule L. |
| J | Metric instrument names follow the suffix conventions: `lower_snake_case`, no `xrpld_`/`xrpl_` prefix (the exporter adds it), a counter ends `_total`, a histogram ends `_us`/`_ms`/`_seconds`, a gauge does not end `_total`. The instrument **kind** is read from the emit site, never guessed from words in the name — so a multi-series gauge carrying units in its label values (e.g. `nodestore_state` observing `write_mean_us`) is not a violation. A name created through two different factories is itself reported as a kind conflict, since no suffix can be correct for both. |
| K | Every metric named in `docker/telemetry/workload/expected_metrics.json` resolves to a declared constant, so a rename in code cannot leave the workload validator asserting a name nothing emits. PromQL selectors (`m{label="v"}`) and exporter-appended histogram suffixes (`_bucket`/`_count`/`_sum`) are normalized away first; groups fed by another emit path (`statsd_gauges`, `statsd_counters`, `spanmetrics`) are out of scope by design. |
Rule F runs **unconditionally** (it is a purely syntactic check on the
call-sites and needs no `*SpanNames.h`), so a code path that calls
`SpanGuard::span`/`setAttribute` directly without ever defining a header is
| H | A namespace-qualified constant (e.g. `foo::bar::myKey`) used at a telemetry call-site is not defined in any `*SpanNames.h`. The constant should live in the proper header; defining it in-place bypasses rules A/G/F. Warns rather than fails — the argument may be a legitimately dynamic value, and the header may live on a later branch. Bare locals and `std::` names are not warned. |
| L | A literal metric name in a family that has no `*MetricNames.h` constants yet. Rule I's ratchet defers these instead of failing the build on the whole pre-existing metric surface at once; the warning keeps the outstanding conversion work visible rather than silently accepted. |
## Presence-gated
Every rule runs **only when the source files it needs are present** in the tree
and is otherwise skipped (printed as `SKIP: <rule> — <reason>`), never failed.
This keeps the check correct no matter how telemetry work is split across PRs —
a stacked chain, one large PR, or independent per-stage PRs where (for example)
the collector config lands before the dashboards. The collector/Tempo/dashboard/
runbook layers are introduced in later phases; on a branch without them, only
@@ -3,6 +3,12 @@ name: Build and test configuration
on:
workflow_call:
inputs:
benchmark:
description:"Whether to smoke-run the benchmarks after testing."
required:false
type:boolean
default:false
build_only:
description:'Whether to only build or to build and test the code ("true", "false").'
required:true
@@ -63,6 +69,12 @@ on:
type:string
default:""
toolchain:
description:'Where the toolchain comes from ("nix" to build the flake CI environment on the runner, empty for the system one). macOS only: Linux always builds in a Nix image, and Nix has no Windows support.'
required:false
type:string
default:""
secrets:
CODECOV_TOKEN:
description:"The Codecov token to use for uploading coverage reports."
@@ -100,23 +112,32 @@ jobs:
# header files are copied into separate directories by CMake, which will
@@ -42,6 +42,7 @@ This section contains changes targeting a future version.
### Bugfixes
-`get_aggregate_price`: Duplicate entries in the `oracles` request array are now ignored. [#6586](https://github.com/XRPLF/rippled/pull/6586)
- Peer Crawler: The `port` field in `overlay.active[]` now consistently returns an integer instead of a string for outbound peers. [#6318](https://github.com/XRPLF/rippled/pull/6318)
-`ping`: The `ip` field is no longer returned as an empty string for proxied connections without a forwarded-for header. It is now omitted, consistent with the behavior for identified connections. [#6730](https://github.com/XRPLF/rippled/pull/6730)
- gRPC `GetLedgerDiff`: Fixed error message that incorrectly said "base ledger not validated" when the desired ledger was not validated. [#6730](https://github.com/XRPLF/rippled/pull/6730)
@@ -53,6 +54,8 @@ This section contains changes targeting a future version.
-`submit`: The `fail_hard` field now returns an error if the value is not a boolean. [#6529](https://github.com/XRPLF/rippled/pull/6529)
-`subscribe`: The `taker` field in the `books` array now returns `actMalformed` instead of `badIssuer` if the value is not a valid account. [#6529](https://github.com/XRPLF/rippled/pull/6529)
- Fixed a bug in `Forwarded` HTTP header parsing where the extracted IP address could be incorrect when no comma or semicolon delimiter follows the address. This could cause the server to misidentify a client's IP address when operating behind a reverse proxy. [#6529](https://github.com/XRPLF/rippled/pull/6529)
-`gateway_balances`: The `account` and `ident` fields now return an `invalidParams` error if the value is not a string, instead of an `internal` error. [#7655](https://github.com/XRPLF/rippled/pull/7655)
-`account_lines`: The `peer` field now returns an error if the value is not a string. [#7728](https://github.com/XRPLF/rippled/pull/7728)
If you're seeing an error like `ERROR: Package 'snappy/1.1.10' not resolved: Unable to find 'snappy/1.1.10#968fef506ff261592ec30c574d4a7809%1756234314.246' in remotes.`,
please [add `xrplf` remote](#add-xrplf-remote) or re-run `conan export` for [patched recipes](./docs/build/advanced_conan.md#patched-recipes).
please [set Conan up](#set-up-conan) so the `xrplf` remote is configured, or re-run `conan export` for [patched recipes](./docs/build/advanced_conan.md#patched-recipes).
### `protobuf/port_def.inc` file not found
@@ -400,7 +412,6 @@ For example, if you want to build Debug:
1. For conan install, pass `--settings build_type=Debug`
@@ -225,8 +225,9 @@ environment, so you don't need to install most of the individual tools
yourself. The version of each hook sourced from an external repository
(`clang-format`, `gersemi`, etc.) is pinned in that file, so running the hooks
locally uses exactly the same versions as CI. A few `local` hooks — most notably
`clang-tidy` — run tools from your own environment; see
[Installing clang-tidy](#installing-clang-tidy) for how to get those.
`clang-tidy` and `cargo fmt` — run tools from your own environment; see
[Installing clang-tidy](#installing-clang-tidy) and
[Rust](./docs/build/environment.md#rust) for how to get those.
To get started, install `pre-commit` and enable the git hook scripts:
@@ -255,6 +256,7 @@ The hooks configured in this repository include, among others:
-`clang-tidy` — C++ static analysis (see [Clang-tidy](#clang-tidy)); opt in with `TIDY=1`
-`fix-include-style`, `fix-pragma-once`, `check-doxygen-style` — C++ hygiene
-`gersemi` — CMake formatting
-`cargo fmt` — Rust formatting for the crates in `crates/`
-`prettier`, `black`, `shfmt` — formatting for JavaScript/JSON/Markdown, Python, and shell
-`cspell` — spell checking
@@ -319,7 +321,11 @@ See the [environment setup guide](./docs/build/environment.md#clang-tidy) for ho
### Running clang-tidy locally
Before running clang-tidy, you must build the project to generate required files (particularly protobuf headers). Refer to [`BUILD.md`](./BUILD.md) for build instructions.
Before running clang-tidy, you must generate the files it depends on (protobuf headers, and, when the project is configured with `-Drust=ON`, the cxxbridge headers from the Rust crates). Configure the project as described in [`BUILD.md`](./BUILD.md), then build the `tidy_prerequisites` target, which generates all of them:
`-format` reformats the fixed code with [`.clang-format`](./.clang-format); without it the fixes are inserted in LLVM style and the `clang-format` hook rewrites them afterwards.
## Telemetry span attribute naming
OpenTelemetry span attribute keys follow these rules so they stay consistent
across the code, the OTel collector, Tempo, Grafana dashboards, and docs. The
constants in the `*SpanNames.h` headers are the single source of truth; every
other layer must match them. A CI check enforces this end to end.
1. Per-span unique attribute: bare field name — allowed when the field is
recorded by a single span/workflow, so the span name already supplies the
domain (e.g. `command`, `local`, `version` on `rpc.command` / `tx.process`).
2. Shared attribute (same concept on more than one span): ONE key, reused
verbatim on every span that records it — the span name tells the occurrences
apart, so no per-emitter prefix is added. Pick the name by the field's
meaning: a property of a domain object keeps that object's bare field name
(`ledger_hash`, `ledger_seq`, `tx_hash`, `peer_id`, `full_validation`); a
field already qualified by a sub-kind keeps that qualifier on every emitter
(`proposal_trusted` on both `consensus.proposal.receive` and
`peer.proposal.receive`; `validation_trusted` likewise). Define it once in
the base `SpanNames.h``namespace attr` block and re-export (`using`) it from
each domain header, so all emitters share the exact string.
3. Collision qualifier: `<domain>_<field>` — only when a bare name would collide
with a DIFFERENT concept in the shared spanmetrics label space, or with the
Distributed tracing is a method for tracking data objects as they flow through distributed systems. In a network like XRP Ledger, a single transaction touches multiple independent nodes—each with no shared memory or logging. Distributed tracing connects these dots.
**Without tracing:** You see isolated logs on each node with no way to correlate them.
**With tracing:** You see the complete journey of a transaction or an event across all nodes it touched.
D --> E["ledger.update<br/>span_id: 005<br/>20ms"]
end
style A fill:#0d47a1,stroke:#082f6a,color:#ffffff
style B fill:#1b5e20,stroke:#0d3d14,color:#ffffff
style C fill:#1b5e20,stroke:#0d3d14,color:#ffffff
style D fill:#1b5e20,stroke:#0d3d14,color:#ffffff
style E fill:#bf360c,stroke:#8c2809,color:#ffffff
```
**Reading the diagram:**
- **tx.submit (blue, root)**: The top-level span representing the entire transaction submission; all other spans are its descendants.
- **tx.validate, tx.relay, tx.apply (green)**: Direct children of tx.submit, representing the three main stages -- validation, relay to peers, and application to the ledger.
- **ledger.update (red)**: A grandchild span nested under tx.apply, representing the actual ledger state mutation triggered by applying the transaction.
- **Arrows (parent to child)**: Each arrow indicates a parent-child span relationship where the parent's completion depends on the child finishing.
The same trace visualized as a **timeline (Gantt chart)**:
Spans don't always form simple parent-child trees. Distributed tracing defines several relationship types to capture different causal patterns:
### 1. Parent-Child (ChildOf)
The default relationship. The parent span **depends on** or **contains** the child span. The child runs within the scope of the parent.
```
tx.submit (parent)
├── tx.validate (child) ← parent waits for this
├── tx.relay (child) ← parent waits for this
└── tx.apply (child) ← parent waits for this
```
**When to use:** Synchronous calls, nested operations, any case where the parent's completion depends on the child.
### 2. Follows-From
A causal relationship where the first span **triggers** the second, but does **not wait** for it. The originator fires and moves on.
```
Time →
tx.receive [=======]
↓ triggers (follows-from)
tx.relay [===========] ← runs independently
```
**When to use:** Asynchronous jobs, queued work, fire-and-forget patterns. For example, a node receives a transaction and queues it for relay — the relay span _follows from_ the receive span but the receiver doesn't wait for relaying to complete.
> **OpenTracing** defined `FollowsFrom` as a first-class reference type alongside `ChildOf`.
> **OpenTelemetry** represents this using **Span Links** with descriptive attributes instead (see below).
### 3. Span Links (Cross-Trace and Non-Hierarchical)
Links connect spans that are **causally related but not in a parent-child hierarchy**. Unlike parent-child, links can cross trace boundaries.
| **Follows-From** | Usually | Causal but no dependency | Link + attributes |
| **Span Link** | Either | Correlation, no dependency | Link + attributes |
---
## Trace ID Generation
A `trace_id` is a 128-bit (16-byte) identifier that groups all spans belonging to one logical operation. How it's generated determines how easily you can find and correlate traces later.
### General Approaches
#### 1. Random (W3C Default)
Generate a random 128-bit ID when a trace starts. Standard approach for most services.
| Prefix search: "find all traces for TX abc" | Must propagate to maintain full trace_id |
| Unique per processing instance | More complex generation logic |
| Retries get distinct trace_ids | Partial correlation only (prefix match) |
### XRPL Workflow Analysis
XRPL has a unique advantage: its core workflows produce **globally unique 256-bit hashes** that are known on every node. This makes deterministic trace_id generation practical in ways most systems can't achieve.
#### Natural Identifiers by Workflow
| Workflow | Natural Identifier | Size | Known at Start? | Same on All Nodes? |
The string prefix (`"tx"`, `"cons"`, `"fetch"`) prevents collisions between workflows that might share underlying hashes.
**Why this works for XRPL:**
1.**Propagation-resilient** — Even if a P2P message drops trace context, every node independently computes the same trace_id from the same tx_hash or ledger_hash. Spans still correlate.
2.**Zero-cost search** — "Show me the trace for transaction ABC" becomes a direct lookup: compute `SHA-256("tx" || ABC)[0:16]` and query. No secondary index needed.
3.**Cross-workflow linking via Span Links** — A consensus trace links to individual transaction traces. A validation span links to the consensus trace. This connects the full picture without forcing everything into one giant trace.
### Cross-Workflow Correlation
Each workflow gets its own trace. Span Links tie them together:
- **Transaction Trace (blue)**: An independent trace whose `trace_id` is deterministically derived from the transaction hash. Contains receive, validate, and relay spans.
- **Consensus Trace (green)**: An independent trace whose `trace_id` is derived from the previous ledger hash and sequence number. Covers the open, propose, and accept phases.
- **Validation (red)**: Validation spans live within the consensus trace (not a separate trace). They are created after the accept phase completes.
- **Catch-Up Trace (purple)**: An independent trace for ledger acquisition, derived from the target ledger hash. Used when a node is behind and fetching missing ledgers.
- **Dotted arrows (span links)**: Cross-trace correlations. Consensus links to transaction traces it included; catch-up links to the consensus trace that produced the target ledger.
- **Solid arrow (C3 to V1)**: A parent-child relationship -- validation spans are direct children of the consensus accept span within the same trace.
**How a query flows:**
```
"Why was TX abc slow?"
1. Compute trace_id = SHA-256("tx" || abc)[0:16]
2. Find transaction trace → see it was included in consensus round N
3. Follow span link → consensus trace for round N
4. See which phase was slow (propose? accept?)
5. If a node was catching up, follow link → catch-up trace
| **Retries get same trace_id** | Add `attempt` attribute to root span; spans have unique span_ids and timestamps |
| **256→128 bit truncation** | Birthday-bound collision at ~2⁶⁴ operations — negligible for XRPL's throughput |
| **Non-standard generation** | OTel spec allows any 16-byte non-zero value; tooling works on the hex string |
| **Hash computation cost** | SHA-256 is ~0.3μs per call; XRPL already computes these hashes for other purposes |
| **Late-binding identifiers** | Ledger hash isn't known until after consensus — validation spans use ledger_seq as fallback, then link to the consensus trace |
---
## Distributed Traces Across Nodes
In distributed systems like xrpld, traces span **multiple independent nodes**. The trace context must be propagated in network messages:
```mermaid
sequenceDiagram
participant Client
participant NodeA as Node A
participant NodeB as Node B
participant NodeC as Node C
Client->>NodeA: Submit TX<br/>(no trace context)
Note over NodeA: Creates new trace<br/>trace_id: abc123<br/>span: tx.receive
Note over NodeC: Creates child span<br/>span: tx.relay<br/>parent_span_id: 001
Note over NodeA,NodeC: All spans share trace_id: abc123<br/>enabling correlation across nodes
```
**Reading the diagram:**
- **Client**: The external entity that submits a transaction. It does not carry trace context -- the trace originates at the first node.
- **Node A**: The entry point that creates a new trace (trace_id: abc123) and the root span `tx.receive`. It relays the transaction to peers with trace context attached.
- **Node B and Node C**: Peer nodes that receive the relayed transaction along with the propagated trace context. Each creates a child span under Node A's span, preserving the same `trace_id`.
- **Arrows with trace context**: The relay messages carry `trace_id` and `parent_span_id`, allowing each downstream node to link its spans back to the originating span on Node A.
---
## Context Propagation
For traces to work across nodes, **trace context must be propagated** in messages.
xrpld P2P messages such as `TMTransaction` carry the trace context in two added byte fields alongside the existing payload: `trace_parent` holds the W3C traceparent (`trace_id`, `span_id`, and `trace_flags`), and `trace_state` holds the optional W3C tracestate. Together they propagate the trace across the P2P boundary so a receiving node can attach its spans to the sender's span.
---
## Sampling
Not every trace needs to be recorded. **Sampling** reduces overhead:
### Head Sampling (at trace start)
```
Request arrives → Random N% chance → Record or skip entire trace
```
- ✅ Low overhead
- ❌ May miss interesting traces
> **xrpld note**: xrpld intentionally fixes head sampling at 100% (sample
> everything) and does not expose a configurable ratio. A per-node ratio
> would let different nodes make divergent keep/drop decisions for the same
> distributed trace, producing broken/partial traces. xrpld uses a
> `ParentBased` sampler so spans with a remote parent honor the upstream
> decision. Volume reduction is delegated to collector-side tail sampling.
- **Core Services (blue)**: The entry points into xrpld -- RPC Server handles client requests, Overlay manages peer-to-peer networking, Consensus drives agreement, and ValidatorList manages trusted validators.
- **JobQueue (center)**: The asynchronous thread pool that decouples Core Services from the Processing and Application layers. All work flows through it.
- **Application Services (purple)**: Higher-level features -- PathFinding computes payment routes, TxQ manages fee-based queuing, and LoadManager tracks server load.
- **Existing Observability (orange)**: The current monitoring stack (PerfLog, Insight, Journal logging) that OpenTelemetry will complement, not replace.
- **Arrows (Services to JobQueue to layers)**: Work originates at Core Services, is enqueued onto the JobQueue, and dispatched to Processing or Application layers for execution.
| **SHAMap** | `src/xrpld/shamap/` | State hash tree | Sync speed, missing nodes |
---
## 1.3 Transaction Flow Diagram
Transaction flow spans multiple nodes in the network. Each node creates linked spans to form a distributed trace:
```mermaid
sequenceDiagram
participant Client
participant PeerA as Peer A (Receive)
participant PeerB as Peer B (Relay)
participant PeerC as Peer C (Validate)
Client->>PeerA: 1. Submit TX
rect rgb(230, 245, 255)
Note over PeerA: tx.receive SPAN START
PeerA->>PeerA: HashRouter Deduplication
PeerA->>PeerA: tx.validate (child span)
end
PeerA->>PeerB: 2. Relay TX (with trace ctx)
rect rgb(230, 245, 255)
Note over PeerB: tx.receive (linked span)
end
PeerB->>PeerC: 3. Relay TX
rect rgb(230, 245, 255)
Note over PeerC: tx.receive (linked span)
PeerC->>PeerC: tx.process
end
Note over Client,PeerC: DISTRIBUTED TRACE (same trace_id: abc123)
```
**Reading the diagram:**
- **Client**: The external entity that submits a transaction to Peer A. It has no trace context -- the trace starts at the first node.
- **Peer A (Receive)**: The entry node that creates the root span `tx.receive`, runs HashRouter deduplication to avoid processing duplicates, and creates a child `tx.validate` span.
- **Peer A to Peer B arrow**: The relay message carries trace context (trace_id + parent span_id), enabling Peer B to create a linked span under the same trace.
- **Peer B (Relay)**: Receives the transaction and trace context, creates a `tx.receive` span linked to Peer A's trace, then relays onward.
- **Peer C (Validate)**: Final hop in this example. Creates a linked `tx.receive` span and runs `tx.process` to fully process the transaction.
- **Blue rectangles**: Highlight the span boundaries on each node, showing where instrumentation creates and closes spans.
### Trace Structure
```
trace_id: abc123
├── span: tx.receive (Peer A)
│ ├── span: tx.validate
│ └── span: tx.relay
├── span: tx.receive (Peer B) [parent: Peer A]
│ └── span: tx.relay
└── span: tx.receive (Peer C) [parent: Peer B]
└── span: tx.process
```
---
## 1.4 Consensus Round Flow
Consensus rounds are multi-phase operations that benefit significantly from tracing:
- **consensus.round (orange, root span)**: The top-level span encompassing the entire consensus round, with attributes like ledger sequence, mode, and proposer count.
- **consensus.phase.open (blue)**: The first phase where the node waits (~3s) to collect incoming transactions before proposing.
- **consensus.phase.establish (green)**: The negotiation phase where validators exchange proposals, resolve disputes, and converge on a transaction set. Child spans track each proposal received/sent and each dispute resolved.
- **consensus.phase.accept (pink)**: The final phase where the agreed transaction set is applied, a new ledger is built, and the ledger is validated. Child spans cover `ledger.build` and `ledger.validate`.
- **Arrows (open to establish to accept)**: The sequential flow through the three consensus phases. Each phase must complete before the next begins.
---
## 1.5 RPC Request Flow
> **WS** = WebSocket
RPC requests support W3C Trace Context headers for distributed tracing across services:
```mermaid
flowchart TB
subgraph request["rpc.request (root span)"]
http["HTTP Request — POST /<br/>traceparent:<br/>00-abc123...-def456...-01"]
- **rpc.request (green, root span)**: The outermost span representing the full RPC request lifecycle, from HTTP receipt to response. Carries the W3C `traceparent` header for distributed tracing.
- **HTTP Request node**: Shows the incoming POST request with its `traceparent` header and extracted attributes (method, peer IP, command name).
- **jobqueue.enqueue (blue)**: The span covering the asynchronous handoff from the RPC thread to the JobQueue worker thread. The trace context is preserved across this async boundary.
- **rpc.command.submit (orange)**: The span for the actual command execution, with child spans for deserialization, local validation, and network submission.
- **Response node**: The final output with HTTP status and total duration, marking the end of the root span.
- **Arrows (top to bottom)**: The sequential processing pipeline -- receive request, extract attributes, enqueue job, execute command, return response.
---
## 1.6 Key Trace Points
> **TxQ** = Transaction Queue
The following table identifies priority instrumentation points across the codebase:
| **Memory Spike** | Large batch of spans correlating with memory increase | Tune batch_size or sampling |
| **Network Partition** | Traces missing cross-node links for specific peer | Check peer connectivity |
| **Path Computation Slow** | pathfind.compute span shows high latency; cache miss rate in attributes | Warm the RippleLineCache, check order book depth |
- **xrpld Nodes (blue)**: The source of telemetry data. Each xrpld node exports spans via OTLP/HTTP on port 4318 (the only exporter shipped in Phase 1b).
- **OpenTelemetry Collector (red)**: The central aggregation point that receives spans from all nodes. Can run as a sidecar (per-node) or standalone (shared). Handles batching, filtering, and routing.
- **Observability Backends (green)**: The storage and visualization destinations. Tempo is the recommended backend for both development and production, and Elastic APM is an alternative. The Collector routes to one or more backends.
- **Arrows (nodes to collector to backends)**: The data pipeline -- spans flow from nodes to the Collector over HTTP, then the Collector fans out to the configured backends.
### 2.2.1 OTLP/HTTP (Shipped in Phase 1b)
OTLP/HTTP is the only exporter wired up in Phase 1b. It is configured via
`OtlpHttpExporterOptions` with the collector traces endpoint
(`http://localhost:4318/v1/traces` by default) and a JSON content type
(binary protobuf is also available).
### 2.2.2 OTLP/gRPC (Future Work — Planned Upgrade)
OTLP/gRPC is planned as a future upgrade from the HTTP exporter. The gRPC
transport offers lower per-span overhead and tighter back-pressure semantics
than HTTP/JSON, making it attractive for production deployments once the HTTP
path is validated in earlier phases.
Required to land this upgrade:
1. Add `opentelemetry-cpp::otlp_grpc_exporter` to the Conan recipe (the
dependency already exists but is not linked in Phase 1b builds).
2. Extend `TelemetryConfig.cpp` to parse an `exporter` key (`otlp_http`
default, `otlp_grpc` opt-in) and a gRPC endpoint override.
3. In `Telemetry::start()` branch on the parsed exporter type and construct
either `OtlpHttpExporterFactory::Create(httpOpts)` or
| `tx.process` | Full transaction processing | Live | — |
| `tx.relay` | Transaction relay to peers | **Not built** | **Superseded by metrics.** Relay volume is carried by the overlay traffic counters (`total_bytes_in/out`, `total_messages_in/out`, per-`TrafficCount` category). Relay is also per-peer fan-out, so one span per relay multiplies by peer count for data the counters already aggregate. |
| `tx.apply` | Apply transaction to ledger | Live | — |
| `rpc.command.*` | Specific RPC command (dynamic) | Live | — |
| `peer.connect` | Peer connection establishment | **Not built** | **Gap, scoped as its own change** — see [09 §6.4](./09-data-collection-reference.md#64-peer-span-coverage-gap-not-implemented). Adding these changes the 41-family span count and the 40 catalogued in `expected_spans.json`. |
| `peer.disconnect` | Peer disconnection | **Not built** | **Gap.** Partially observable: the aggregate count via the `Overlay.Peer_Disconnects` insight gauge and resource-charge drops via `server_info{metric="peer_disconnects_resources"}`, but not per-reason. Disconnect reasons are only recoverable from `debug.log` (the `log-derived-insights` dashboard). |
| `peer.message.send` | Send protocol message | **Not built** | **Gap.** Of the 13 protocol message families only `mtGET_OBJECTS` has native instrumentation (`getobject_*`); byte/message volume is aggregated by `TrafficCount` category, not traced per message. |
| `peer.message.receive` | Receive protocol message | **Not built** | **Gap.** Same as `peer.message.send`. |
| `ledger.acquire` | Ledger acquisition from network | Live | — |
| `ledger.build` | Build new ledger | Live | — |
| `ledger.validate` | Ledger validation | Live | — |
| `ledger.replay` | Ledger replay executed | **Not built** | **Gap, no substitute.**`LedgerReplayer.cpp` and `LedgerReplayTask.cpp` contain zero `SpanGuard` uses and no metric covers the replay path. A real hole, not a trade-off. |
| `ledger.delta` | Delta-based ledger acquired | **Not built** | **Gap, no substitute.**`LedgerDeltaAcquire.cpp` contains zero `SpanGuard` uses. The `acquire_*` stats cover whole-ledger acquisition, not the delta path. |
| `fee.escalate` | Fee escalation triggered | **Not built** | **Superseded by metrics + existing spans.** Escalation state is `txq_metrics{metric=…}` and `load_factor_metrics{metric=…}`; the queueing path that triggers it is already traced by the six `txq.*` spans. An event span would restate a gauge. |
| `validator.list.fetch` | UNL list fetched | **Not built** | **Superseded by metrics.**`validator_health{metric="unl_expiry_days"}`, `{metric="unl_blocked"}` and `{metric="validation_quorum"}` carry the outcome. A fetch span would fire on a slow timer and tell an operator nothing the gauges do not. |
| `validator.manifest` | Manifest update processed | **Not built** | **Superseded by logs.** Per-master-key manifest dispositions are on the `log-derived-insights` dashboard (`ManifestCache` partition, requires `log_level ManifestCache debug`). |
| `amendment.vote` | Amendment voting executed | **Not built** | **Deferred to Phase 11.**`validator_health{metric="amendment_blocked"}` covers the blocked state in the meantime. |
| `shamap.sync` | State tree synchronization | **Not built** | **Superseded by metrics.** Covered by the nine `acquire_*` stats, `nodestore_state{metric=…}` and the five `getobject_*` families. Per-node-fetch spans would be prohibitive volume. |
| `job.enqueue` | Job added to queue | **Not built** | **Superseded by metrics.**`job_queued_total` and `job_queued_us{job_type}` plus the 105 per-job-type `jobq_*` gauges. A span per enqueue is one span per unit of daemon work, for latency the histogram already records exactly. |
| `job.execute` | Job execution | **Not built** | **Superseded by metrics.**`job_started_total`, `job_finished_total`, `job_running_us{job_type}`. Same volume argument as `job.enqueue`. |
| `service.name` | `"xrpld"` | Standard `SERVICE_NAME` | Set |
| `service.version` | `build_info::getVersionString()` | Standard `SERVICE_VERSION` | Set |
| `service.instance.id` | node public key (base58), or `[telemetry] service_instance_id` | Standard `SERVICE_INSTANCE_ID` | Set — but the node-key fallback reaches traces only; see [05 §5.1.1](./05-configuration-reference.md) |
| `xrpl.network.id` | network id (e.g. 0 for mainnet) | Network identifier | Set |
| `xrpl.node.type` | `"validator"` \| `"stock"` \| `"reporting"` | Node role | **Not implemented** — no constant, no set-site. Node role is therefore not queryable from a trace. (Dashboards do offer an `$xrpl_node_role` filter, but it matches a Prometheus label stamped by the external perf-iac deployment — `check_otel_naming.py:872` — not by anything in this repo) |
| `xrpl.node.cluster` | cluster name | Cluster name, if clustered | **Not implemented** — no constant, no set-site |
The collector adds two more resource attributes of its own (`deployment.environment`
and, when the node did not stamp it, `xrpl.network.type`) via the
`resource/tier` processor, and deletes the SDK-injected `telemetry.sdk.*` trio
via `resource/stripsdk`. See [05 §5.5.1](./05-configuration-reference.md).
### 2.4.2 Span Attributes by Category
> Span attribute keys use the underscore form from §2.3.3 (shared/qualified
> keys are `<domain>_<field>`; per-span unique keys are bare). The dotted form
> is reserved for the resource attributes in §2.4.1 above. This catalog lists
> the planned attribute set by category; the exact emitted key **and its type**
> for each implemented span is defined by the `*SpanNames.h` constants and their
> set-sites, which win where the two differ. The types in the tables below are
> the ones originally planned and are **not** all what shipped — `peer_id` is
> the notable case (planned as a base58 string, shipped as an int64). §2.4.3
| **Peer** | `peer_id` (**int64**, the process-local `Peer::id_` slot number — not a key of any kind; also set on `tx.receive`), `proposal_trusted`, `validation_trusted`, `ledger_hash`, `full_validation`. (`peer_version` is **not** a peer-span key: the constant lives in `TxSpanNames.h:79` and its only set-site is `PeerImp.cpp:1342` on the `tx.receive` span — see the Transaction row) | `peer_address`, `peer_latency_ms`, `peer_cluster`, `message_type`, `message_size_bytes`, `message_compressed` — the peer-span coverage gap (§2.3.2) | Network topology analysis |
| **Ledger** | `ledger_seq`, `tx_count`, `tx_failed`, `validations`, `acquire_reason`, `timeouts`, `peer_count`, `outcome`, `close_time`, `close_time_correct`, `close_resolution_ms` | `ledger_index` (→ `ledger_seq`), `ledger_tx_count` (→ `tx_count`). `ledger_hash` is a live key, but **no `ledger.*` span sets it** — only `consensus.validation.send` (`RCLConsensus.cpp:977`; that span is the one returned by `createValidationSpan()`, which names `cs::validationSend` at `RCLConsensus.cpp:1365,1373`) and `peer.validation.receive` (`PeerImp.cpp:2573`) do. The `LedgerSpanNames.h:41``using` alias has zero uses. `consensus.ledger_close` sets **no** hash: its four attributes are `ledger_seq`, `consensus_mode`, `tx_count_open` and `close_time_resolution_ms` (`RCLConsensus.cpp:354-361`) | Ledger progression tracking |
| **Job** | — (no job spans exist) | `job_type`, `job_queue_ms`, `job_worker`. JobQueue is observed via **metrics**, not spans — but by **two disjoint families**, and only one of them has a `job_type` label. See the note below the table | JobQueue performance |
| **Fee** | — (no `fee.escalate` span, §2.3.2) | `fee_load_factor`, `fee_escalation_level`. Fee escalation is dashboarded from metrics (`fee-market`), not spans | Fee escalation monitoring |
| **Validator** | — (no `validator.*` span, §2.3.2) | `validator_list_size`, `validator_list_age_sec`. UNL health is dashboarded from metrics (`validator-health`) | UNL health monitoring |
> | `beast::insight` `jobq` group: `jobq_<jobtype>_waiting` / `_running` / `_deferred` / `_q` | In the **metric name itself** — there is **no** `job_type` label at all | `JobTypeData.h:29-32` (naming contract), `:35-38` (suffixes), `Application.cpp:392` (group) |
>
> **The trap:** `sum by (job_type)(jobq_…)` collapses every job type into a
> single series with an empty `job_type`, because an absent PromQL label is
> equivalent to `""` — the query returns a plausible-looking number rather than
> an error. Aggregate the `jobq_*` family with a name matcher
> (`{__name__=~"jobq_.*_waiting"}`) and reserve `by (job_type)` for the
> `job_*_total` / `job_*_us` family.
### 2.4.4 Privacy & Sensitive Data Policy
> **PII** = Personally Identifiable Information
OpenTelemetry instrumentation is designed to collect **operational metadata only**, never sensitive content.
#### Data NOT Collected
The following data is explicitly **excluded** from telemetry collection:
| **IP Addresses** | **Never in spans** — no span sets an address attribute (`peer_address` has zero set-sites); peer spans identify peers by `peer_id`, an int64 process-local slot number. **But the log pipeline is a different story** — see the note below this table |
> **Peer IPs DO leave the node — via the log pipeline, not via spans.** The
> "IP Addresses" row above is scoped to spans, and only to spans. This same
> document describes a log pipeline (§2.6.5) that carries peer addresses:
>
> 1. `PeerImp`'s constructor logs the peer's `remoteAddress_` — an `IP:port` —
> at `info` severity (`PeerImp.h:837-842`), and other overlay call sites log
> addresses too. These land in the ordinary `debug.log` stream.
> 2. The collector's `filelog` receiver tails exactly that file
| **Account Hashing** | Account addresses are hashed both SDK-side (`pathfind_source_account`, `pathfind_dest_account` — always hashed before emission) and again at the collector level, so raw addresses never reach storage |
| **Unconditional Redaction** | Account redaction is **not** configurable and cannot be turned off: `redactAccount()` (`Redaction.cpp:14-29`) hashes every **non-empty** address handed to it, with no flag and no bypass (an empty input returns empty — `Redaction.cpp:18-19` — so there is no raw value to leak either way). That is a stronger guarantee than a config switch: there is no insecure-by-default state to misconfigure |
| **Collector Tail Sampling** | **Optional, and OFF in the base stack.** xrpld head sampling is fixed at 1.0 (`Telemetry.h:234``static constexpr double samplingRatio = 1.0;`), so 100% of traces leave the node. `docker/telemetry/otel-collector-config.yaml` has **no**`tail_sampling` processor either, so the local stack stores 100%. The only shipped policy is in the Grafana Cloud overlay (`otel-collector-config.grafanacloud.yaml:60-67`, wired at `:261`): one `probabilistic` policy at **0.5%**, on the trace-storage branch only so spanmetrics still see every span. Treat sampling as a cost control you opt into — not as a privacy control |
| **Local Control** | Node operators have full control over what gets exported |
| **No Raw Payloads** | Transaction content is never recorded, only metadata (hash, type, result) |
| **Collector-Level Filtering** | Additional redaction/hashing can be configured at OTel Collector |
#### Account Address Hashing
Account addresses are **always** hashed before they reach the telemetry
backend — there is no opt-out flag and therefore no insecure-by-default
failure mode. Protection is applied in two independent layers:
1.**SDK-side** (this node): the path-finding RPC handlers call
`redactAccount()` (`xrpl::telemetry`, `Redaction.h`) before setting the
`pathfind_source_account` / `pathfind_dest_account` span attributes. For a
non-empty address the helper emits the first 16 characters of
`sha512Half(address)` as lowercase hex — deterministic (spans for one
account still correlate) but non-reversible. An empty address returns empty
rather than the hash of the empty string (`Redaction.cpp:18-19`).
2.**Collector-side** (defense-in-depth): an `attributes/hash` processor in
the OpenTelemetry Collector re-hashes those same attributes, so any node
that emitted a raw value is still redacted before storage.
#### Collector-Level Data Protection
The shipped base config does exactly one thing here, and it is the
defense-in-depth layer described above: an `attributes/hash` processor
(`otel-collector-config.yaml:105-110`) hashing `pathfind_source_account` and
`pathfind_dest_account`.
**No `peer_address` or `params` scrubbing rule is needed on the trace pipeline,
and none is shipped.** Earlier drafts prescribed `delete` actions for both.
Neither attribute is ever emitted: `peer_address` has zero set-sites in the code
(peer spans carry `peer_id`, an int64 process-local slot number — not an IP and
not a key), and no span sets a `params` attribute — RPC spans carry `command`,
(`trace_transactions`, `trace_consensus`, `trace_rpc`, `trace_peer` — the last
often disabled due to high volume) select which spans are emitted. Account
address hashing is not configurable: addresses are hashed unconditionally by
the SDK helper described above, with collector-level hashing as a second
layer.
> **Key Principle**: Telemetry collects **operational metadata** (timing, counts, hashes) — never **sensitive content** (keys, balances, amounts, raw payloads).
> **See also**: [Securing the OTel Pipeline](./secure-OTel.md) covers transport-level protection for telemetry leaving the node — mTLS to the collector and validation of incoming peer trace context. Privacy controls in this section keep sensitive data out of spans; the security doc keeps the spans themselves out of untrusted hands.
---
## 2.5 Context Propagation Design
> **WS** = WebSocket
### 2.5.0 Deterministic Trace ID Strategy
Both transaction and consensus tracing use **deterministic trace IDs** derived from
a globally known hash, so all nodes handling the same workflow independently produce
spans under the same `trace_id`. This is combined with protobuf `span_id` propagation
for parent-child relay ordering when available.
#### Transactions — `trace_id = txHash[0:16]`
Every node that handles a transaction knows its `txID` (the `uint256` transaction
hash). The first 16 bytes of this hash are used as the OTel `trace_id`:
- **HTTP/WebSocket - RPC (blue)**: For client-facing RPC requests, trace context is propagated using the W3C `traceparent` header. This is the standard approach and works with any OTel-compatible client.
- **Protocol Buffers - P2P (green)**: For peer-to-peer messages between xrpld nodes, trace context is embedded as a protobuf `TraceContext` message carrying trace_id, span_id, flags, and optional trace_state.
- **JobQueue / Coroutines - Internal Async (red)**: For asynchronous work within a single node, the ambient OTel context follows the coroutine rather than being carried on the work item. `include/xrpl/core/Job.h` has **no** telemetry include and no `traceContext_` member — an earlier draft of this diagram showed one, and that was never built. Instead `xrpl::telemetry::CoroAwareContextStorage` (`include/xrpl/telemetry/CoroAwareContextStorage.h:84`) overrides the SDK's `RuntimeContextStorage` with a per-coroutine context stack, and is installed as the global storage in `Telemetry::start()` (`Telemetry.cpp:416-419`) before the tracer provider and before the first span. That fixes the wrong-thread scope pop across coroutine yield/resume and keeps log↔trace correlation intact. The storage is never reset — tearing it down while spans may still exist is undefined behaviour in the SDK — so it lives for the process lifetime.
- **xrpld Process (dark gray)**: The single xrpld node running all three observability frameworks side by side. Each framework operates independently with no interference.
- **PerfLog to perf.log**: PerfLog writes JSON-formatted event logs to a local file. Grafana can ingest these via Loki or a file-based datasource.
- **Beast Insight to StatsD Server**: Insight sends aggregated metrics (counters, gauges) over UDP to a StatsD server. Grafana reads from StatsD-compatible backends like Graphite or Prometheus (via StatsD exporter).
- **OpenTelemetry to OTLP Collector**: OTel exports spans over OTLP/HTTP to a Collector, which then forwards to a trace backend (Tempo). (OTLP/gRPC is future work — §2.2.2.)
- **Grafana (red, unified UI)**: All three data streams converge in Grafana, enabling operators to correlate logs, metrics, and traces in a single dashboard.
**Phase 7 outcome (as shipped)**: Beast Insight gained an `OTelCollector`
`Collector` implementation that rides the global MeterProvider and exports via
OTLP/HTTP to the same collector as traces. It is selected with
`[insight] server=otel`.
The three back ends are **co-equal branches of one `if/else` chain** in
`makeCollectorManager()` (`CollectorManager.cpp:37-75`), not a migration path:
| `otel` | `OTelCollector` | OTLP/HTTP to the OTel Collector — the recommended setting |
| `statsd` | `StatsDCollector` | Unchanged from before Phase 7. **Not deprecated**: no warning is logged, no removal is scheduled, and the code path is not marked legacy |
| absent / anything else | `NullCollector` | **The default.** A node with no `[insight]` section emits no metrics at all |
Two corrections to earlier drafts, both of which matter operationally: StatsD
is not a "deprecated fallback", and `otel` is not the default — you must set it
explicitly. See [06-implementation-phases.md §6.8](./06-implementation-phases.md),
[Phase7_taskList.md](./Phase7_taskList.md), and
[05 §5.8.6](./05-configuration-reference.md) for which `[insight]` keys are live
under `server=otel` (most are inert).
### 2.6.5 Correlation with Logs
**Shipped in Phase 8 — and not the way this section originally planned it.**
The design here was a `setTraceId` hook on PerfLog, fed from the
`rpc.command.<method>` span in `RPCHandler.cpp`. That hook was never built:
`setTraceId` has zero occurrences in **source** — the only hits in the tree are
in these plan documents, describing the design that was dropped — and PerfLog's
JSON output carries no trace ID.
What shipped instead is broader and needs no per-call-site wiring: the **journal
sink** stamps the IDs onto _every_ log line written while a span is active.
| 100% | ~500 | ~250 KB/s | **The only reachable node→collector figure.** Plan capacity against this row |
| 0.5% | ~2.5 | ~1.25 KB/s | Collector→backend only, and only with the Grafana Cloud overlay's `tail_sampling` (`otel-collector-config.grafanacloud.yaml:60-67`) |
| 10% / 1% / error-only | — | — | **Not implemented.** No shipped config produces these ratios; treat them as illustrative of what a tail-sampling policy could do |
The rows below 100% therefore reduce _storage_ cost at the backend, never the
node's egress. Note also that the shipped 0.5% policy is applied to the
trace-storage branch only, so the spanmetrics-derived RED metrics still see
Instrumentation is gated on two levels. A compile-time feature flag reduces the trace macros to no-ops when telemetry is built out, so disabled builds carry zero cost. At runtime, per-component guards (e.g. `shouldTracePeer()`) skip span creation for components whose tracing is turned off, incurring no overhead beyond a single boolean check.
> The compile-time gate is the macro `XRPL_ENABLE_TELEMETRY`, but that macro is
> **not** the switch you flip. It is a compile definition added by
> `CMakeLists.txt` (`add_compile_definitions(XRPL_ENABLE_TELEMETRY)`) when the CMake option `telemetry` is ON.
> That option is declared ON today (`option(telemetry "Enable OpenTelemetry tracing" ON)`)
> only so that CI compiles the instrumented build while the telemetry branches are
> in review; **OFF is the intended default once merged**, flipped in a separate
> change. Select the value explicitly instead of relying on the default:
> `-Dtelemetry=ON|OFF` (CMake) or `-o telemetry=True|False` (Conan). See
> [05 §5.4.2](./05-configuration-reference.md).
---
## 3.8 Links to Detailed Documentation
- **[Configuration Reference](./05-configuration-reference.md)**: Configuration options and collector setup
- **[Implementation Phases](./06-implementation-phases.md)**: Detailed timeline and milestones
---
## 3.9 Code Intrusiveness Assessment
> **TxQ** = Transaction Queue
This section provides a detailed assessment of how intrusive the OpenTelemetry integration is to the existing xrpld codebase.
| **Build System** | Low | A single CMake option (`telemetry`) selects the whole feature in or out, and builds work either way (`-Dtelemetry=ON` / `-Dtelemetry=OFF`). It is declared ON today only so CI compiles the instrumented paths; **OFF is the intended default once merged**, so the shipped build is opt-in |
| **Dependencies** | Medium | `opentelemetry-cpp/1.28.0` is a **conditional** requirement, never a hard one: `conanfile.py:152-153` adds it only `if self.options.telemetry`, and `:238-239` adds the matching `libxrpl` component requirement the same way. The option's declared default is `True` today (`conanfile.py:59`), so a default `conan install` does resolve it; with `-o telemetry=False` it never enters the graph and the null implementation supplies the factory |
| **Config File** | ✅ Full | New `[telemetry]` section is optional |
| **Protocol** | ✅ Full | Optional protobuf fields with high field numbers |
| **Build** | ✅ Full | `-Dtelemetry=OFF` (or `-o telemetry=False`) produces a binary with all tracing compiled out, whatever the option's declared default happens to be. **Not**`-DXRPL_ENABLE_TELEMETRY=OFF`, which does not disable anything — it is not a CMake option, only a compile definition that `CMakeLists.txt:152` adds inside the `if(telemetry)` block. CMake does flag it (`Manually-specified variables were not used by the project`) at the end of configuration, so it is not literally silent — but the warning is easy to scroll past and the resulting binary still has telemetry compiled in. See [05 §5.4.2](./05-configuration-reference.md) |
| **Runtime** | ✅ Full | `enabled=0` produces zero overhead |
| **API** | ✅ Full | No changes to public RPC or P2P APIs |
### 3.9.6 Rollback Strategy
If issues are discovered after deployment:
1.**Immediate**: Set `enabled=0` in `[telemetry]` and restart (zero code change).
Also set `[insight] server=` to something other than `otel` if metrics must
stop too — `enabled=0` governs tracing, and the metrics pipeline is selected
2.**Quick**: Rebuild with `-Dtelemetry=OFF` (CMake) or `-o telemetry=False`
(Conan). Pass the flag explicitly — an omitted flag resolves to the option's
declared default, which is ON today and OFF once the feature is merged; a
build that already has telemetry off needs no rebuild at all.
**Do not use `-DXRPL_ENABLE_TELEMETRY=OFF`** — it is not a CMake option, so
it is ignored (CMake reports it under `Manually-specified variables were not
used by the project`) and the rebuilt binary still has telemetry compiled in.
This step also drops the `opentelemetry-cpp` dependency, so expect a full
rebuild rather than an incremental one.
3.**Complete**: Revert telemetry commits (clean separation makes this easy)
### 3.9.7 Code Change Examples
**Minimal RPC Instrumentation (Low Intrusiveness):** Instrumenting an RPC handler adds roughly 3-4 lines: one macro to start the span and one or two `setAttribute` calls (command name, status). The span ends automatically via RAII, so the existing control flow — process the request, send the result — is untouched.
**Consensus Instrumentation (Medium Intrusiveness):** Consensus is slightly more intrusive because child spans in later phase transitions need the round's context. Beyond the span-start and attribute macros, this requires **four** new member variables on the adaptor rather than the single `currentRoundContext_` this section originally sketched (`RCLConsensus.h:103,113,123,143`):
-`std::optional<telemetry::SpanGuard> roundSpan_` (`:103`) — the round span
itself. It is **created and ended in one place**, `startRoundTracing()`: the
previous round's guard is released at `RCLConsensus.cpp:1288-1289`
(`if (roundSpan_) roundSpan_.reset();`) and the new one is emplaced a few
lines later — at `:1306` or `:1310` on the `"attribute"` strategy, at `:1319`
on the default `"deterministic"` one. `preStartRound()` does not create it; it
calls `startRoundTracing()` at `:1229`. There is no `reset()` method — the
span simply lives until the next round begins. A `SpanGuard` owns no
thread-local scope, so emplacing and resetting on different job workers is
safe.
-`telemetry::SpanContext roundSpanContext_` (`:113`) — a lightweight value-type
snapshot, captured at the end of `startRoundTracing()` (`:1350`). Child spans
link through this, not through an ambient parent, so code running on another
worker (e.g. `createValidationSpan()` on `jtACCEPT`) never touches
`roundSpan_` cross-thread.
-`telemetry::SpanContext prevRoundSpanContext_` (`:123`) — the prior round's
context, saved at `:1282`**before** the new span overwrites
`roundSpanContext_`, so the new round span can carry a follows-from link and
consecutive rounds stay navigable.
-`telemetry::SpanContext acceptSpanContext_` (`:143`) — the current round's
accept-span context, set at `:544` and cleared at `:1286` on each new round.
`createValidationSpan()` prefers it as the parent and falls back to
`roundSpanContext_` (`:1363-1373`), so a stale value must not survive into the
next round.
The split is the point: the guard is owned by one thread, the contexts are
copied freely. The existing round logic itself remains unchanged.
The authoritative `[telemetry]` example lives in `cfg/xrpld-example.cfg`. Telemetry is disabled by default (`enabled=0`); enabling it turns on distributed tracing for transaction flow, consensus, and RPC calls, with traces exported to an OpenTelemetry Collector over OTLP. Head sampling is intentionally fixed at 1.0 (sample everything) and is not configurable — per-node head-sampling would produce broken/partial distributed traces, so volume reduction is delegated to the collector's tail sampling (see Section 7.4.2). Transaction trace IDs are always deterministic (`trace_id = txHash[0:16]`); there is no strategy switch for the transaction path. The full option reference follows.
> **`service_instance_id` is effectively required for `beast::insight`
> metrics — and only for those.** Three producers resolve the instance id
> independently, and exactly one of them lacks a node-key fallback:
>
> | Producer | Resource built by | Unset `service_instance_id` yields |
| `consensus_trace_strategy` | string | `"deterministic"` | Consensus trace ID strategy: `"deterministic"` (trace_id = prevLedgerHash[0:16]) or `"attribute"` (random). Parsed at `TelemetryConfig.cpp:155-156`, consumed at `RCLConsensus.cpp:1291,1296`. **Not validated** — see the note below |
| `service_name` | string | `"xrpld"` | Service name (`service.name`) for traces and metrics |
| `service_instance_id` | string | node public key (base58) | Instance identifier (`service.instance.id`). Traces, span metrics and native `XRPL_METRIC_*` metrics all fall back to the node key; **`beast::insight` metrics do not** — see the note in §5.1.1 |
**`consensus_trace_strategy` is not validated.** `TelemetryConfig.cpp:155-156`
copies the raw string into `Setup::consensusTraceStrategy` without checking it
against an allowed set, and the only comparison in the code is
`strategy == "attribute"` (`RCLConsensus.cpp:1296`). Any unrecognised value —
including a typo — silently takes the deterministic branch with no log warning.
The two accepted values are documented at `include/xrpl/telemetry/Telemetry.h:287-292`.
**Not a config key — deterministic transaction trace IDs are unconditional.**
Earlier drafts of this document listed a `tx_trace_strategy` option
(`"deterministic"` \| `"attribute"`). No such key exists: `TelemetryConfig.cpp`
parses no transaction-strategy key, and the transaction trace ID is always
derived from the transaction hash. Only the **consensus** path has a
switchable strategy.
**Planned (not yet implemented)**: the following options appear in the design
documents but are not parsed by `TelemetryConfig.cpp`. They will be added as
> **`exporter` is not read, so do not set it.** Both shipped sample configs
> (`docker/telemetry/xrpld-telemetry.cfg`,
> `docker/telemetry/xrpld-telemetry-mainnet.cfg`) used to carry
> `exporter=otlp_http`; the line had no effect and has since been replaced with
> a comment saying so. OTLP/HTTP is the only transport that exists (§2.2.1), and
> `endpoint` / `metrics_endpoint` are the only transport knobs, until the §2.2.2
> gRPC work lands.
---
## 5.2 Configuration Parser
> **TxQ** = Transaction Queue
The parser `makeTelemetrySetup()` in `src/libxrpl/telemetry/TelemetryConfig.cpp` reads the `[telemetry]``Section` and populates a `Telemetry::Setup` struct, applying the defaults listed in Section 5.1.2 via `section.valueOr(...)`. It takes `serviceInstanceId` from the `nodePublicKey` argument when the key is absent, applies one unconditional `endpoint` default (`dflt::endpoint`, `TelemetryConfig.cpp:61`, used at `:108`) — the parser has no notion of exporter type — and leaves the sampling ratio at its fixed 1.0 default (a `static constexpr` member, so there is nothing to parse; `TelemetryConfig.cpp:139`, `Telemetry.h:234`). It also rejects two contradictory mTLS configurations outright (`tls_client_cert` without `tls_client_key`, and either without `use_tls=1`) rather than failing open at handshake time.
`metrics_endpoint` is deliberately **not** handled here: it is read separately in `ApplicationImp::startTelemetry()` (`Application.cpp:1670`) and passed to `MetricsRegistry::start()`. Note the consequence — the two metric exporters resolve their URL differently:
| Metric source | Exporter built by | URL comes from |
Setting a non-default `endpoint` therefore moves the insight metrics with it, but leaves the native metrics on localhost unless `metrics_endpoint` is set too.
---
## 5.3 Application Integration
### 5.3.1 ApplicationImp Changes
> **Deferred identity**: The node public key (`nodeIdentity_`) is not
> available during `ApplicationImp`'s member initializer list — it is
> resolved later in `setup()`. The `Telemetry` object is therefore
> constructed with an empty `serviceInstanceId` and patched via
> `setServiceInstanceId()` once `setup()` has called `getNodeIdentity()`.
> the one `beast::insight` metrics use — is already frozen by then (§5.1.1), so
> those metrics keep whatever `service_instance_id` the config supplied (`""`
> if it supplied none). Native `XRPL_METRIC_*` metrics do not go through this
> patch at all: `startTelemetry()` re-reads the config key and applies its own
> node-key fallback when building `MetricsRegistry`'s separate resource
> (`Application.cpp:1674-1679`).
`ApplicationImp` (in `src/xrpld/app/main/Application.cpp`) owns a `std::unique_ptr<telemetry::Telemetry> telemetry_`. It is built in the member initializer list via `makeTelemetry(makeTelemetrySetup(...))` with an empty `serviceInstanceId`, then patched in `setup()` by calling `setServiceInstanceId()` with the Base58 node public key (unless the user supplied a custom `service_instance_id`). `start()` and `run()` forward to `telemetry_->start()` / `telemetry_->stop()`, and `getTelemetry()` returns the owned instance.
### 5.3.2 ServiceRegistry Interface Addition
`include/xrpl/core/ServiceRegistry.h` gains a pure-virtual `telemetry::Telemetry& getTelemetry()` (with a forward declaration of `telemetry::Telemetry`), giving every component a uniform accessor for the tracing subsystem.
> **Note:** `Application` extends `ServiceRegistry`, so `getTelemetry()` is
> available on both. Components that hold a `ServiceRegistry&` (e.g.
> `NetworkOPsImp`) call `registry_.get().getTelemetry()`. Components that
> still hold an `Application&` (e.g. `ServerHandler`, `PeerImp`,
> **Production hardening**: The configurations in this section are starting points. For production deployments where xrpld ships telemetry across a network to a centrally-hosted collector, see [Securing the OTel Pipeline](./secure-OTel.md) for the required mTLS receiver config, NetworkPolicy, and peer trace-context validation.
The authoritative collector config lives in the repo at `docker/telemetry/otel-collector-config.yaml` (with Tempo backend config in `docker/telemetry/tempo.yaml`). The sections below summarize the development and production shapes of that pipeline.
### 5.5.1 Development / Base Configuration
`docker/telemetry/otel-collector-config.yaml` is the base config used by the
local stack and by CI. It carries **three** pipelines, not one:
`metrics/cloud`, `logs`. `tail_sampling` is applied in **`traces/store`**
(`:259-261`) — the branch feeding Tempo and Grafana Cloud — not in a pipeline
named `traces`, which does not exist in the overlay. The `traces/metrics`
branch feeds `spanmetrics` unsampled, so the derived RED metrics stay exact
while stored traces are ~1/200 of ingested ones.
> **Known issue — the cloud path does not hash pathfinding accounts.** The base
> config runs `attributes/hash` on its `traces` pipeline
> (`otel-collector-config.yaml:105-110`), hashing `pathfind_source_account` and
> `pathfind_dest_account` as defense in depth behind the node-side hashing. The
> overlay declares no such processor and lists none on any of its five
> pipelines, so on the Grafana Cloud config those two attributes reach **both**
> Grafana Cloud and the local Tempo with whatever value the node sent. Any node
> that emits raw addresses loses its second line of defense. Adding
> `attributes/hash` to `traces/store` and `traces/metrics` would close the gap.
Hardening a collector for a real deployment (TLS/mTLS on the receiver,
NetworkPolicy, peer trace-context validation) is covered in
[Securing the OTel Pipeline](./secure-OTel.md) — not by any config file in
`docker/telemetry/`.
---
## 5.6 Docker Compose Development Environment
> **OTLP** = OpenTelemetry Protocol
The authoritative development stack lives in the repo at `docker/telemetry/docker-compose.yml`. It brings up **six** services on a shared `xrpld-telemetry` bridge network. All images are pinned to exact tags.
| `renderer` | `grafana/grafana-image-renderer:v5.12.0` | `8081` | Panel→PNG rendering for image export and alert screenshots |
Two corrections to earlier drafts:
- **`prometheus` is not optional.** `grafana` lists it in `depends_on` (along
with `tempo`, `loki` and `renderer`), and 7 of the 15 dashboards query
`span_calls_total` from it. Removing it blanks most panels.
- **Port `13133` is not published.** The collector's `health_check` extension
listens on `13133` inside the container, but the base compose file publishes
only `4317`, `4318` and `8889`. Health checks from the host must either add a
port mapping or run `docker compose exec`.
The collector also bind-mounts the xrpld log root read-only
(`${XRPLD_LOG_DIR:-./data/logs}` → `/var/log/xrpld`) for the `filelog`
receiver, and the `grafana` service reads Slack/email alert secrets from an
optional gitignored `.env.alerting`.
---
## 5.7 Configuration Architecture
> **OTLP** = OpenTelemetry Protocol
```mermaid
flowchart TB
subgraph config["Configuration Sources"]
cfgFile["xrpld.cfg<br/>[telemetry] section"]
cmake["CMake option: telemetry<br/>ON today for CI, OFF once merged<br/>when ON, defines XRPL_ENABLE_TELEMETRY"]
end
subgraph init["Initialization"]
parse["makeTelemetrySetup()"]
factory["makeTelemetry()"]
end
subgraph runtime["Runtime Components"]
tracer["TracerProvider"]
exporter["OTLP Exporter"]
processor["BatchProcessor"]
end
subgraph collector["Collector Pipeline"]
recv["Receivers"]
proc["Processors"]
exp["Exporters"]
end
cfgFile --> parse
cmake -->|"compile flag"| parse
parse --> factory
factory --> tracer
tracer --> processor
processor --> exporter
exporter -->|"OTLP"| recv
recv --> proc
proc --> exp
style config fill:#e3f2fd,stroke:#1976d2
style runtime fill:#e8f5e9,stroke:#388e3c
style collector fill:#fff3e0,stroke:#ff9800
```
**Reading the diagram:**
- **Configuration Sources**: `xrpld.cfg` provides runtime settings (endpoint, per-component trace toggles) while the CMake `telemetry` option controls whether telemetry is compiled in at all. That option is declared ON today only so CI compiles the instrumented paths; OFF is the intended default once merged, so treat the build gate as something to pass explicitly, and the runtime gate is opt-in either way (`enabled=0` by default). Head sampling is fixed at 1.0 and is not a config option; volume reduction happens via tail sampling in the collector.
- **Initialization**: `makeTelemetrySetup()` parses config values, then `makeTelemetry()` constructs the provider, processor, and exporter objects.
- **Runtime Components**: The `TracerProvider` creates spans, the `BatchProcessor` buffers them, and the `OTLP Exporter` serializes and sends them over the wire.
- **OTLP arrow to Collector**: Trace data leaves the xrpld process via OTLP/HTTP and enters the external Collector pipeline. (OTLP/gRPC is future work — see design decisions §2.2.2.)
- **Collector Pipeline**: `Receivers` ingest OTLP data, `Processors` apply sampling/filtering/enrichment, and `Exporters` forward traces to storage backends (Tempo, etc.).
---
## 5.8 Grafana Integration
> **APM** = Application Performance Monitoring
Step-by-step instructions for integrating xrpld traces with Grafana.
### 5.8.1 Data Source Configuration
Three datasources are provisioned from `docker/telemetry/grafana/provisioning/datasources/`. There is **no** Elastic APM datasource — `elastic-apm.yaml` was described in an earlier draft but never existed. Elastic remains a _possible_ backend (§7.2); nothing in this repo provisions it.
| `address` | `:39` | StatsD only — the UDP endpoint. |
| `prefix` | `:41`, `:53` | **Inert.** Stored on the OTel collector but `formatName()` prepends nothing (`OTelCollector.cpp:855-866`); only StatsD applies it. |
| `endpoint` | `:50` | **Inert.** Logged for diagnostics (`OTelCollector.cpp:730`), then unused. |
- **Budget Constraints? (Yes)**: Leads to open-source options. If you already run Grafana or Elastic, pick the matching backend; otherwise default to Grafana Tempo.
- **Budget Constraints? (No) → Prefer SaaS?**: If you want a managed service, choose between Datadog (enterprise support) and Honeycomb (developer-focused). If not, fall back to open-source.
- **Terminal nodes (Tempo / Elastic / Honeycomb / Datadog)**: Each represents a concrete backend choice, all of which feed into the same final step.
- **Configure Collector**: Regardless of backend, you always finish by configuring the OTel Collector to export to your chosen destination.
- **Validator / Stock Nodes**: All xrpld nodes emit trace data via OTLP. Validators and stock nodes are grouped separately because they may reside in different network zones.
- **Collector Cluster (DC1, DC2)**: Regional collectors receive OTLP from nodes in their datacenter, apply processing (sampling, enrichment), and fan out to multiple backends. Enrichment includes deployment-tier tagging: each collector stamps `deployment.environment` and (as a fallback) `xrpl.network.type` so one Grafana stack can filter data from many collectors by tier.
- **Storage Backends**: Tempo and Elastic provide queryable trace storage; S3/GCS Archive provides long-term cold storage for compliance or post-incident analysis.
- **Grafana Dashboards**: The single visualization layer that queries both Tempo and Elastic, giving operators a unified view of all traces.
> **Note**: Production deployments should use multiple collector instances behind a load balancer for high availability. The diagram shows a simplified single-collector topology for clarity.
- **xrpld Node (three signals, one transport)**: spans and metrics both leave over OTLP/HTTP on port 4318. Logs do not leave the node at all — the node just writes `debug.log`, and the journal sink prefixes `trace_id=`/`span_id=` whenever a span is active (`Log.cpp:304-338`).
- **OTel Collector (single agent)**: an `otlp` receiver takes spans and metrics; a `filelog` receiver tails `/var/log/xrpld/*/debug.log` and regex-parses the trace/span IDs out of each line. A `spanmetrics` connector derives RED metrics from the trace stream and feeds them into the metrics pipeline. Three pipelines, three exporters — see [05 §5.5.1](./05-configuration-reference.md).
- **PerfLog is not in this picture.** It still writes `perf.log`, but nothing collects it and it carries no trace ID; the `setTraceId` hook once planned for it was never built ([02 §2.6.5](./02-design-decisions.md)).
- **StatsD is not in this picture either.** It remains a supported `[insight] server=` choice, but selecting it takes metrics _out_ of this pipeline and requires a StatsD receiver you would have to add yourself — the compose file's StatsD port mapping is commented out.
- **Grafana**: correlation is bidirectional and configured in the datasources, not in a bespoke panel — Tempo's `tracesToLogs` (`filterByTraceID: true`) jumps trace → logs, and `loki.yaml`'s derived fields jump log → trace.
| **Trace** | `tx_hash` | — | Live as a span attribute for search; **not** used as a cross-signal join key (`tags: []`) |
| **Trace** | `ledger_seq` | — | Live as a span attribute; not a join key |
| **Journal log** | `trace_id`, `span_id` | Traces | **Live.** Emitted by `Log.cpp:304-338` into `debug.log`, parsed by the collector's `filelog` receiver, jumped via `loki.yaml` derived fields |
| **PerfLog** | `trace_id` | Traces | **Not implemented.** PerfLog output has no trace ID; the planned `setTraceId` hook was never built. Use the journal log instead |
| **Insight** | `exemplar.trace_id` | Traces | **Not implemented.** No exemplar configuration exists anywhere in the code or collector config — no `exemplar_filter` on the SDK side, no `exemplarTraceIdDestinations` on the Prometheus datasource. Metric spike → trace jumps must be done by time range today |
| [Phase11_taskList.md](./Phase11_taskList.md) | Third-party data collection pipelines (future) |
> **Only Phase 11 is still "future".** Phase 9 ships on
> `pratik/otel-phase9-metric-gap-fill` (18 task entries, 9.1–9.17 plus 9.7a) and
> Phase 10 on `pratik/otel-phase10-workload-validation` (7 tasks). Their task
> lists are present on every branch from those points forward, so a reader on a
> later branch sees plans that are already implemented, not proposals. Phase 11
> (13 tasks) has no implementation branch.
> **Note**: Phases 1 and 6 do not have separate task list files. Phase 1 tasks are documented in [06-implementation-phases.md §6.2](./06-implementation-phases.md). Phase 6 tasks are documented in [06-implementation-phases.md §6.7](./06-implementation-phases.md).
---
## 8.6 Phase 9–11 Cross-Reference Guide
This guide maps Phase 9–11 content to its location across the documentation.
### Phase 9: Internal Metric Instrumentation Gap Fill
# [OpenTelemetry](00-tracing-fundamentals.md) Distributed Tracing Implementation Plan for xrpld
## Executive Summary
> **OTLP** = OpenTelemetry Protocol
This document provides a comprehensive implementation plan for integrating OpenTelemetry distributed tracing into the xrpld XRP Ledger node software. The plan addresses the unique challenges of a decentralized peer-to-peer system where trace context must propagate across network boundaries between independent nodes.
### Key Benefits
- **End-to-end transaction visibility**: Track transactions from submission through consensus to ledger inclusion
- **Consensus round analysis**: Understand timing and behavior of consensus phases across validators
- **RPC performance insights**: Identify slow handlers and optimize response times
- **Network topology understanding**: Visualize message propagation patterns between peers
- **Incident debugging**: Correlate events across distributed nodes during issues
| **7** | [Observability Backends](./07-observability-backends.md) | Backend selection guide and production architecture |
| **8** | [Appendix](./08-appendix.md) | Glossary, references, version history |
| **9** | [Data Collection Reference](./09-data-collection-reference.md) | Complete inventory of spans, attributes, metrics, and dashboards |
| **Sec** | [Securing the OTel Pipeline](./secure-OTel.md) | Threat model and hardening (mTLS, peer trace-context validation) |
> Note there is no document 4: `04-code-samples.md` was removed during the
> rollout, and the numbering was left as-is rather than renumbering every
> cross-reference in the chain.
---
## 0. Tracing Fundamentals
This document introduces distributed tracing concepts for readers unfamiliar with the domain. It covers what traces and spans are, how parent-child and follows-from relationships model causality, how context propagates across service boundaries, and how sampling controls data volume. It also maps these concepts to xrpld-specific scenarios like transaction relay and consensus.
The xrpld node consists of several key components that require instrumentation for comprehensive distributed tracing. The main areas include the RPC server (HTTP/WebSocket), Overlay P2P network, Consensus mechanism (RCLConsensus), JobQueue for async task execution, PathFinding, Transaction Queue (TxQ), fee escalation (LoadManager), ledger acquisition, validator management, and existing observability infrastructure (PerfLog, Insight/StatsD, Journal logging).
Key trace points span across transaction submission via RPC, peer-to-peer message propagation, consensus round execution, ledger building, path computation, transaction queue behavior, fee escalation, and validator health. The implementation prioritizes high-value, low-risk components first: RPC handlers provide immediate value with minimal risk, while consensus tracing requires careful implementation to avoid timing impacts.
➡️ **[Read full Architecture Analysis](./01-architecture-analysis.md)**
The OpenTelemetry C++ SDK is selected for its CNCF backing, active development, and native performance characteristics. Traces are exported via OTLP/HTTP to an OpenTelemetry Collector, which provides flexible routing and sampling. OTLP/gRPC is planned future work (see design decisions §2.2.2).
Span naming follows a hierarchical `<component>.<operation>` convention (e.g., `rpc.command.server_info`, `tx.process`, `consensus.round`). Context propagation uses W3C Trace Context headers for HTTP and embedded Protocol Buffer fields for P2P messages. The implementation coexists with existing PerfLog and Insight observability systems through correlation IDs.
**Data Collection & Privacy**: Telemetry collects only operational metadata (timing, counts, hashes) — never sensitive content (private keys, balances, amounts, raw payloads). Account addresses are hashed **unconditionally** by the SDK helper and hashed again at the collector; there is no redaction config key and therefore no insecure-by-default state. Trace volume is _not_ reduced on the node (head sampling is fixed at 100%); reduction, where wanted, is a collector-side tail-sampling decision. Node operators control which subsystems are traced via the `[telemetry]` per-component toggles.
➡️ **[Read full Design Decisions](./02-design-decisions.md)**
---
## 3. Implementation Strategy
The telemetry code is organized under `include/xrpl/telemetry/` for headers, `src/libxrpl/telemetry/` for implementation, and `src/xrpld/telemetry/` for the native-metrics module added in Phases 7 and 9. Key principles include RAII-based span management via `SpanGuard` (with `discard()` for dropping unwanted spans), a `FilteringSpanProcessor` that intercepts `OnEnd()` to prevent discarded spans from entering the export pipeline, conditional compilation behind the `XRPL_ENABLE_TELEMETRY` compile definition (set by the CMake `telemetry` option, which defaults to **ON** — build it out with `-Dtelemetry=OFF`), and minimal runtime overhead through batch processing.
Performance optimization strategies include head sampling fixed at 100% (intentionally not configurable, so trace keep/drop decisions stay coherent across nodes), optional tail-based sampling at the collector to reduce stored volume (not enabled in the base stack — the only shipped policy is a 0.5% probabilistic one in the Grafana Cloud overlay), batch export to reduce network overhead, and conditional instrumentation that compiles to no-ops when disabled.
➡️ **[Read full Implementation Strategy](./03-implementation-strategy.md)**
Configuration is handled through the `[telemetry]` section in `xrpld.cfg` with options for enabling/disabling, TLS/mTLS, batch tuning, and component-level filtering. Exporter selection is _not_ configurable — OTLP/HTTP is the only transport. Head sampling is fixed at 1.0 (not operator-configurable); volume reduction is done by tail sampling in the collector. CMake integration uses the `telemetry` option (default **ON**) for compile-time control.
Endpoints are spread across **three** keys in two sections, not one "traces and metrics" pair:
`[telemetry]` itself has exactly **one**`endpoint` key, and it is traces-only.
The repo ships one collector config (`docker/telemetry/otel-collector-config.yaml`, three pipelines: traces, metrics, logs) plus a Grafana Cloud overlay that adds 0.5% tail sampling. A six-service Docker Compose stack — collector, Tempo, Loki, Prometheus, Grafana, renderer — gives a complete local environment.
➡️ **[View full Configuration Reference](./05-configuration-reference.md)**
---
## 6. Implementation Phases
The plan was originally scoped at **13 weeks across 8 phases** — the table below
is that original scope. As delivered it grew to **11 phases through week 20**;
Phases 9-11 were added after the original plan was written. See
[06-implementation-phases.md §6.12.6](./06-implementation-phases.md) for the
authoritative per-phase status, and treat the eight rows below as the
originally-planned subset rather than the current timeline:
Grafana Tempo is recommended for all environments due to its cost-effectiveness and Grafana integration, and it is the only backend this repo provisions. Elastic APM remains a reasonable choice for organizations with existing Elastic infrastructure, but nothing here configures it.
The recommended production architecture uses a gateway collector pattern with regional collectors performing tail-based sampling, routing traces to multiple backends (Tempo for primary storage, Elastic for log correlation, S3/GCS for long-term archive). Note that several subsections of doc 7 predate the shipped dashboards and alert rules and are marked superseded in place, pointing at [09-data-collection-reference.md](./09-data-collection-reference.md) and `docs/telemetry-runbook.md`.
The appendix contains a glossary of OpenTelemetry and xrpld-specific terms, references to external documentation and specifications, version history for this implementation plan, and a complete document index.
➡️ **[View Appendix](./08-appendix.md)**
---
## 9. Data Collection Reference
A single-source-of-truth reference documenting every piece of telemetry data collected by xrpld: the OpenTelemetry span inventory with per-span attributes, the `beast::insight` and native `XRPL_METRIC_*` instruments (gauges, counters, histograms, overlay traffic), the SpanMetrics-derived Prometheus metrics, and the **15** Grafana dashboards. Includes Tempo search guides and Prometheus query examples. Consult that document rather than this index for any count — it tracks the code, this summary does not.
➡️ **[View Data Collection Reference](./09-data-collection-reference.md)**
---
## Securing the OTel Pipeline
Threat model and hardening guidance for production deployments where xrpld nodes ship telemetry to a centrally-hosted collector across an untrusted network. Covers the two attack surfaces (collector ingress and peer trace-context spoofing) and the chosen defenses: mTLS as primary collector auth, NetworkPolicy as defense-in-depth, and source-side validation plus per-peer rate limiting for the `protocol::TraceContext` field on peer messages.
➡️ **[View Securing the OTel Pipeline](./secure-OTel.md)**
---
_This document provides a comprehensive implementation plan for integrating OpenTelemetry distributed tracing into the xrpld XRP Ledger node software. For detailed information on any section, follow the links to the corresponding sub-documents._
> **Goal**: Build tools that generate realistic XRPL traffic to validate the full Phases 1-9 telemetry stack end-to-end — all spans, attributes, metrics, dashboards, and log-trace correlation — under controlled load.
# Phase 11: Third-Party Data Collection Pipelines — Task List
> **Status**: Not started — 0 of 13 tasks complete (`grep -c '^## Task 11\.'` = 13:
> Tasks 11.1 through 11.13). Verified against the tree:
> no `.go` files exist anywhere, `docker/telemetry/otel-rippled-receiver/` does
> not exist, `docker/telemetry/prometheus/` does not exist (so no
> `prometheus/rippled-alerts.yml`), and no `network-topology` / `dex-amm`
> dashboards are present under `docker/telemetry/grafana/dashboards/`. **No Phase 11 work has
> been done, so no task box below may be ticked.**
>
> One **prerequisite** box is ticked, and only one: Task 11.12's
> "`state_tracking` gauge implemented (Task 7.12)". That is an upstream
> dependency satisfied by Phase 7/9 code, not Phase 11 work — see the citation
> there.
>
> **Goal**: Build a custom OTel Collector receiver that periodically polls xrpld's admin RPCs and exports structured metrics for external consumers — making all XRPL health, validator, peer, fee, and DEX data available as Prometheus/OTLP metrics without xrpld code changes.
| [09-data-collection-reference.md](./09-data-collection-reference.md) | Defines full metric inventory including third-party metrics |
| [Phase10_taskList.md](./Phase10_taskList.md) | Prerequisite — validation harness for testing |
### Third-Party Consumer Gap Analysis
This phase addresses the cross-cutting gap identified during research: **xrpld has no native Prometheus/OTLP metrics export for data accessible only via RPC**. Every consumer (exchanges, payment processors, analytics providers, validators, researchers, compliance firms, custodians) must build custom JSON-RPC polling and conversion. This receiver centralizes that work.
| Consumer Category | Data Unlocked by This Phase |
- `xrpl_object_count{type="<name>"}` for each counted object type (Transaction, Ledger, NodeObject, STTx, STLedgerEntry, InboundLedger, Pathfinder, etc.)
**Cardinality note**: Per-peer metrics use truncated keys. For large peer sets (50+), the aggregate distribution gauges are preferred over per-peer labels.
---
## Task 11.5: Validator & Amendment Collector
**Objective**: Poll `validators` and `feature` to export validator health and amendment voting status.
**What to do**:
- Implement `validatorCollector`:
**From `validators` RPC:**
- `xrpl_trusted_validators_count`
- `xrpl_validator_signing` (0 or 1 — whether local validator is signing)
**From `feature` RPC:**
- `xrpl_amendment_enabled_count` — total enabled amendments
- `xrpl_amendment_majority_count` — amendments with majority but not yet enabled
- Add alerting playbook (what to do for each Tier 1/Tier 2 alert)
---
## Task 11.12: Alert Rules for External Dashboard Parity Metrics
> **Source**: [External Dashboard Parity](./06-implementation-phases.md#appendix-external-dashboard-parity) — 18 alert rules ported from the community [xrpl-validator-dashboard](https://github.com/realgrapedrop/xrpl-validator-dashboard).
**Objective**: Document the external dashboard's "fast path" pattern as a future optimization for real-time panels.
**Pattern**: A lightweight Prometheus scrape endpoint (separate from OTLP pipeline) that polls critical metrics every 2-5s, bypassing the 10s OTLP metric reader interval and Prometheus scrape interval.
**Use case**: Real-time state panels (server state, ledger age, peer count) where 10-15s latency is too slow for operational dashboards.
**Decision**: Document as a future option, not implement now. The current 10s interval is acceptable for v1. The external dashboard achieves 2-5s freshness by polling RPC directly, which is what the Phase 11 receiver already does. Adding a separate scrape endpoint to xrpld would only be needed if sub-second metric freshness is required from the internal metrics pipeline.
**What to document**:
- Architecture comparison: OTLP pipeline (10-15s) vs. direct scrape (2-5s) vs. push gateway
- When to consider: operator feedback indicating 10s is insufficient for alerting SLOs
- How to implement if needed: add `/metrics` HTTP endpoint to xrpld with Prometheus client library
- Trade-offs: additional port, additional dependency, duplication with OTLP metrics
> **Goal**: Complete RPC tracing coverage with unit tests, Grafana search filters, PathFind instrumentation, and config hardening. Build on the Phase 1c SpanGuard factory foundation to achieve production-quality RPC observability.
>
> **Scope**: Unit tests for core telemetry, Grafana Tempo search filters, PathFind RPC tracing, config validation (`std::clamp`).
**Reason**: W3C context propagation (`traceparent`/`tracestate` headers) requires a consumer — in Phase 2, RPC spans are entirely local to the node. Phase 3 introduces cross-node transaction tracing via protobuf context propagation, which is the first use case for extracted trace context. Implementing it here without a consumer would be dead code.
**Status**: COMPLETE (superseded by Phase 1c design)
**Original plan**: Add `XRPL_TRACE_PEER` and `XRPL_TRACE_LEDGER` macros.
**Actual implementation**: Phase 1c replaced all tracing macros with the `SpanGuard::span(TraceCategory, prefix, name)` factory pattern. The `TraceCategory` enum (`Rpc`, `Transactions`, `Consensus`, `Peer`, `Ledger`) serves the same conditional-creation purpose without macros. No separate task needed — the factory already supports all categories.
---
## Task 2.3: Add shouldTraceLedger() to Telemetry Interface
**Objective**: The `Setup` struct has a `traceLedger` field but there's no corresponding virtual method. Add it for interface completeness.
-`src/tests/libxrpl/CMakeLists.txt` — test target with optional OTel linking
---
## Task 2.5: Enhance RPC Span Attributes
**Status**: DEFERRED (low priority)
**Reason**: The high-value attributes (`command`, `version`, `role`, `status`) are already set by Phase 1c. The remaining HTTP transport-level attributes (`http.method`, `net.peer.ip`, `http.status_code`) provide limited additional insight since:
-`http.method` is always POST for JSON-RPC
-`net.peer.ip` is debug-level info available in logs
-`duration_ms` is redundant with span duration (OTel captures start/end time natively)
These can be added later if dashboard queries specifically need them. The node health attributes (Task 2.8) provide far more operational value and were prioritized instead.
---
## Task 2.6: Build Verification and Performance Baseline
**Objective**: Verify the build succeeds with and without telemetry, and establish a performance baseline.
**What to do**:
1. Build with `telemetry=ON` and verify no compilation errors
2. Build with `telemetry=OFF` and verify no regressions
Node health (`amendment_blocked`, `server_state`) is not part of the telemetry surface. Operators consume the same data via the existing `server_info` / `server_state` RPC commands, so duplicating it on traces adds storage and cardinality cost without new value. The OTel C++ SDK 1.18.0 also does not support runtime updates to the resource, ruling out resource-level emission of these dynamic-by-nature flags.
---
## Task 2.9: PathFind RPC Instrumentation
**Status**: COMPLETE
**Objective**: Trace the path_find and ripple_path_find RPC handlers to capture request latency and computation cost.
**Spans added**:
-`pathfind.request` — wraps `doPathFind()` and `doRipplePathFind()` RPC handlers
-`pathfind.update_all` — wraps `PathRequestManager::updateAll()` on ledger close (`pathfind_ledger_index`, `pathfind_num_requests` attrs; emitted only when active subscriptions exist)
-`pathfind.discover` — wraps the entire per-source-asset loop in `PathRequest::findPaths()` (`pathfind_search_level`, `pathfind_num_paths` attrs). One span per RPC call instead of N (one per source asset). Trade-off: per-asset breakdown is lost; storage and cardinality bounded.
**Attribute namespacing**: All pathfind attributes use the `pathfind_*` underscore form per the Phase 1c naming-spec rule 5.
## Task 2.10: RPC and PathFind Span Attribute Gap Fill
**Status**: COMPLETE
**Objective**: Wire up workflow-identifying attributes that enable filtering and grouping traces by request characteristics without drilling into child spans.
> **Goal**: Trace the full transaction lifecycle from RPC submission through peer relay, including cross-node context propagation via Protocol Buffer extensions. This is the WALK phase that demonstrates true distributed tracing.
## Task 3.5: Instrument HashRouter for Dedup Visibility
**Objective**: Make transaction deduplication visible in traces by recording HashRouter decisions as span attributes/events.
**What to do**:
- Edit `src/xrpld/overlay/detail/PeerImp.cpp` (in handleTransaction):
- After calling `HashRouter::shouldProcess()` or `addSuppressionPeer()`:
- Record `suppressed` attribute (true/false)
- Record `tx_flags` showing current HashRouter state (SAVED, TRUSTED, etc.)
- Add `tx.first_seen` or `tx.duplicate` event
- This is NOT a modification to HashRouter itself — just recording its decisions as span attributes in the existing PeerImp instrumentation from Task 3.3.
**Key modified files**:
- `src/xrpld/overlay/detail/PeerImp.cpp` (same changes as 3.3, logically grouped)
---
## Task 3.6: Context Propagation in Transaction Relay
**Objective**: Verify all Phase 3 changes compile and work correctly.
**What to do**:
1. Build with `telemetry=ON` — verify no compilation errors
2. Build with `telemetry=OFF` — verify no regressions
3. Run existing unit tests
4. Verify protobuf regeneration produces correct C++ code
5. Document any issues encountered
**Verification Checklist**:
- [ ] Protobuf changes generate valid C++
- [ ] Build succeeds with telemetry ON
- [ ] Build succeeds with telemetry OFF
- [ ] Existing tests pass
- [ ] No undefined symbols from new telemetry calls
---
## Task 3.8: Transaction Span Peer Version Attribute
> **Source**: [External Dashboard Parity](./06-implementation-phases.md#appendix-external-dashboard-parity) — adds peer version context inspired by the community [xrpl-validator-dashboard](https://github.com/realgrapedrop/xrpl-validator-dashboard).
>
> **Upstream**: Phase 2 (RPC span infrastructure must exist).
> **Downstream**: Phase 10 (validation checks for this attribute).
**Objective**: Add the relaying peer's xrpld version to `tx.receive` spans so operators can correlate transaction issues with peer version mismatches during network upgrades.
**What to do**:
- Edit `src/xrpld/overlay/detail/PeerImp.cpp`:
- In the `tx.receive` span block (after existing `peer_id` setAttribute call):
- Add `peer_version` (string) — from `this->getVersion()`
- Only set if `getVersion()` returns a non-empty string (avoid empty-string attributes)
**Rationale**: Transaction relay is where version mismatches cause subtle serialization or validation bugs. Tracing "this tx came from a v2.3.0 peer" helps diagnose compatibility issues. The community dashboard tracks peer versions externally; this brings version awareness into the trace itself.
**Key modified files**:
- `src/xrpld/overlay/detail/PeerImp.cpp`
**Exit Criteria**:
- [ ] `tx.receive` spans carry `peer_version` attribute with a non-empty version string
- [ ] Attribute is omitted (not set to empty string) when `getVersion()` returns empty
- [ ] Attribute visible in Tempo trace detail view
**Objective**: Add workflow-identifying attributes to transaction spans so operators can filter by transaction type and see outcomes without off-chain correlation.
| 3.11 | TX/TxQ span attribute gap fill | 0 | 5 | 3.3, 3.10 |
**Parallel work**: Tasks 3.1 and 3.4 can start in parallel. Task 3.2 depends on 3.1. Tasks 3.3 and 3.5 depend on 3.2. Task 3.6 depends on 3.3 and 3.5. Task 3.8 depends on 3.3 (span must exist). Task 3.9 depends on 3.2 and 3.3. Task 3.10 depends on 3.4 (tx.process span must exist).
> **Goal**: Production readiness — Grafana dashboards, spanmetrics pipeline, operator runbook, alert definitions, and final integration testing. This phase ensures the telemetry system is useful and maintainable in production.
- **OTelGaugeImpl**: Uses `opentelemetry::metrics::ObservableGauge<uint64_t>` with an async callback. `set(value)` stores value atomically; callback reads it during collection.
- **OTelMeterImpl**: Wraps `opentelemetry::metrics::Counter<uint64_t>`. `increment(amount)` calls `counter->Add(amount)`. Semantically identical to Counter but unsigned.
- **OTelEventImpl**: Wraps `opentelemetry::metrics::Histogram<double>`. `notify()` calls `histogram->Record(value.count())`. Declares its unit from `beast::insight::Unit`, which is what selects its bucket ladder: the histogram views in `Telemetry.cpp` match on unit, so a `ms` instrument gets the millisecond ladder and a `By` instrument the byte ladder. Bucket edges live in `include/xrpl/telemetry/HistogramBuckets.h` — do not restate them here. The millisecond ladder must contain every representable edge of the collector's spanmetrics ladder and may extend above it (jobs outlive spans); `.github/scripts/telemetry/check_bucket_parity.py` enforces that. An earlier version of this line specified `[1, 5, 10, 25, 50, 100, 250, 500, 1000, 5000] ms` as "matching SpanMetrics" — true when written, then silently false once the collector ladder was extended on its own, which capped every quantile above 5s at a flat 5000.
- **OTelHookImpl**: Stores handler function. Called during periodic metric collection (same 1s pattern via PeriodicMetricReader).
- **OTelCollectorImp**: Main class.
- Creates `MeterProvider` with `PeriodicMetricReader` (1s export interval)
- Creates `OtlpHttpMetricExporter` pointing to `[telemetry]` endpoint
- The prefix + group + name concatenation must produce the same string as `StatsDCollector`'s format
- Use underscores as separators (matching StatsD convention)
- Verify in integration test that key Prometheus queries still return data:
- `xrpld_LedgerMaster_Validated_Ledger_Age`
- `xrpld_Peer_Finder_Active_Inbound_Peers`
- `xrpld_rpc_requests`
**Key consideration**: OTel Prometheus exporter may normalize metric names differently than StatsD receiver. Test this early (Task 7.2) and adjust naming strategy if needed. The OTel SDK's Prometheus exporter adds `_total` suffix to counters and converts dots to underscores — match existing conventions.
---
## Task 7.6: Update Grafana Dashboards
**Objective**: Update the 3 StatsD dashboards if any metric names change due to OTLP export format differences.
**What to do**:
- If Task 7.5 confirms metric names are preserved exactly, no dashboard changes needed.
- If OTLP export produces different names (e.g., `_total` suffix on counters), update:
> **Source**: [External Dashboard Parity](./06-implementation-phases.md#appendix-external-dashboard-parity) — the most valuable metric from the community [xrpl-validator-dashboard](https://github.com/realgrapedrop/xrpl-validator-dashboard).
**Objective**: Implement a stateful class that tracks whether our validator's validations agree with network consensus, maintaining rolling 1h and 24h windows with an 8-second grace period and 5-minute late repair window.
**Parallel work**: Tasks 7.4 and 7.5 can run in parallel after 7.2/7.3 complete. Task 7.6 depends on 7.5's findings. Tasks 7.7 and 7.8 can run in parallel after 7.6. Tasks 7.10-7.14 can all run in parallel after 7.2. Task 7.15 depends on 7.2. Task 7.16 depends on 7.9. Task 7.9 depends on 7.2 and Phase 4 Task 4.8.
# Phase 8: Log-Trace Correlation and Centralized Log Ingestion — Task List
> **Goal**: Inject trace context (trace_id, span_id) into xrpld's Journal log output for log-trace correlation, and add OTel Collector filelog receiver to ingest logs into Grafana Loki for unified observability.
>
> **Scope**: Two independent sub-phases — 8a (code change: trace_id in logs) and 8b (infra only: filelog receiver to Loki). No changes to the `beast::Journal` public API.
**Objective**: Add OTel trace context to every log line that is emitted within an active, sampled span. The sampled flag matters because a span dropped by the `ParentBasedSampler` still carries its parent's ids, so emitting them would advertise a trace that was never exported.
**What to do**:
- Edit `src/libxrpl/basics/Log.cpp`:
- In `Logs::format()` (around line 346), after severity is appended, check for active OTel span. The implementation checks the context value directly to avoid the heap allocation that `GetSpan()` performs on the no-span path:
```cpp
#ifdef XRPL_ENABLE_TELEMETRY
{
auto context = opentelemetry::context::RuntimeContext::GetCurrent();
auto spanValue = context.GetValue(opentelemetry::trace::kSpanKey);
- Add `#include` for OTel context headers, guarded by `#ifdef XRPL_ENABLE_TELEMETRY`
- Edit `include/xrpl/basics/Log.h`:
- No changes needed — format() signature unchanged
**Key modified files**:
- `src/libxrpl/basics/Log.cpp`
**Performance note**: The implementation checks the thread-local context value directly (avoiding the heap allocation that `GetSpan()` performs on the no-span path). On threads without an active span (~99% of log lines), the cost is a thread-local read + variant type check (~15-20ns). On the active-span path, an additional shared_ptr copy + `GetContext()` + `IsValid()`/`IsSampled()` adds ~50ns total. Overhead is negligible at typical logging rates.
---
## Task 8.2: Add Loki to Docker Compose Stack
**Objective**: Add Grafana Loki as a log storage backend in the development observability stack.
**What to do**:
- Edit `docker/telemetry/docker-compose.yml`:
- Add Loki service:
```yaml
loki:
image: grafana/loki:3.7.6
ports:
- "3100:3100"
command: -config.file=/etc/loki/local-config.yaml
```
- Add Loki as a Grafana data source in provisioning
<!-- The all-caps macro name XRPL_METRIC_HISTOGRAM_RECORD trips cspell's
compound-word splitter, which emits the subword "ISTOGRAM"; ignore it here. -->
# Phase 9: Internal Metric Instrumentation Gap Fill — Task List
> **Status**: Complete for Tasks 9.1-9.13. Tasks 9.14-9.17 remain open by design
> (see each task for the blocker).
>
> **Goal**: Instrument xrpld to emit ~50+ metrics that exist in `get_counts`/`server_info`/TxQ/PerfLog but currently lack time-series export via the OTel or beast::insight pipelines.
>
> **Scope**: Hybrid approach — extend `beast::insight` for metrics near existing registrations, use OTel Metrics SDK `ObservableGauge` callbacks for new categories (TxQ, PerfLog, CountedObjects).
- Create a centralized `MetricsRegistry` class that holds all OTel async gauge registrations, polled at 10-second intervals by the `PeriodicMetricReader`.
**Objective**: Fill the remaining metric gaps between the external `push_metrics.py` script (in `ripplex-ansible`) and the internal OTel `MetricsRegistry` observable gauges. After this task, all metrics collected by `push_metrics.py` that CAN be collected internally are covered.
**What was done**:
- Extended existing `cacheHitRateGauge_` callback with `AL_size` (AcceptedLedger cache size)
- Extended existing `nodeStoreGauge_` callback with 4 new metrics from `getCountsJson()`:
- Label sanitisation and mean scaling — `MetricsRegistrySanitiseHandler` (5
cases, incl. `output_domain_is_exactly_44_values`) and
`MetricsRegistryScaledMean` (6 cases)
- ❌ Not covered: asserting metric values match `get_counts` JSON output —
that needs a live `Application`, so it is left to the Phase 10 harness
**Key files**:
-`src/tests/libxrpl/telemetry/MetricsRegistry.cpp` (new). The originally
planned `src/test/telemetry/MetricsRegistry_test.cpp` was **never created** —
Phase 9 tests are GTest under `src/tests/libxrpl/`, per project convention.
-`src/tests/libxrpl/telemetry/MetricMacros.cpp`, `GetMeter.cpp` (new — cover
the `XRPL_METRIC_*` macros and meter lookup)
---
## Task 9.11: Validator Health Dashboard (External Dashboard Parity)
> **Source**: [External Dashboard Parity](./06-implementation-phases.md#appendix-external-dashboard-parity) — dashboards for Phase 7 metrics inspired by the community [xrpl-validator-dashboard](https://github.com/realgrapedrop/xrpl-validator-dashboard).
>
> **Upstream**: Phase 7 Tasks 7.9-7.16 (metrics must be emitting).
|`xrpld-jobqueue`|JobQueueTxOverflow|`sum by (service_instance_id) (increase(jq_trans_overflow_total[15m]))`|`gt [0]`|2m|warning|
|`xrpld-jobqueue`|JobQueueLatencyHigh|`histogram_quantile(0.99, sum by (le, service_instance_id) (rate(job_queued_us_bucket[5m])))`|`gt [1000000]`(µs =1s)|5m|warning|
|`xrpld-jobqueue`|NodeStoreIOLatencyHigh|`histogram_quantile(0.95, sum by (le, service_instance_id) (rate(ios_latency_milliseconds_bucket[10m])))`|`gt [1000]`(ms)|10m|warning|
|`xrpld-overlay`|PeerResourceDisconnects|`sum by (service_instance_id) (increase(server_info{metric="peer_disconnects_resources"}[30m]))`|`gt [5]`|5m|warning|
# Securing OpenTelemetry Against Trace Context Spoofing
> **Part of**: [OpenTelemetry Implementation Plan](./OpenTelemetryPlan.md) — see also [Design Decisions § Privacy](./02-design-decisions.md#244-privacy--sensitive-data-policy) (what we don't collect) and [Configuration Reference § 5.5](./05-configuration-reference.md#55-opentelemetry-collector-configuration) (collector base config).
Trace context spoofing (or poisoning) occurs when untrusted actors inject tampered or stale trace IDs into your system. If these requests are processed, the spans are appended to historical trace buckets, stretching trace durations, ruining p99 latency metrics, and breaking Grafana dashboards.
This guide outlines two categories of defense: mitigating tampered contexts and locking down the OpenTelemetry (OTel) Collector to trusted clients only.
---
## Part 1: Mitigating Tampered Trace Contexts
### 1. Perimeter Defense: Strip Headers at the API Gateway
The most effective way to prevent spoofing from external sources is to treat your API Gateway (Envoy, NGINX, AWS ALB) as a hard boundary. Strip incoming W3C tracing headers (`traceparent`, `tracestate`) from public traffic so the gateway is forced to generate a fresh, legitimate `trace_id`.
**NGINX Example (Stripping Headers):**
```nginx
server{
listen80;
location/{
# Clear out untrusted incoming trace headers
proxy_set_headertraceparent"";
proxy_set_headertracestate"";
proxy_passhttp://backend_service;
}
}
```
### **2. Timestamp-Anchored Trace IDs and OTTL Filtering**
If you use a custom trace ID generator that embeds a timestamp in the first few bytes (like AWS X-Ray or UUIDv7), you can use the OTel Collector's OpenTelemetry Transform Language (OTTL) to detect anomalies.
Ensure your Collector is not exposed to the public internet. If running in Kubernetes, use a NetworkPolicy to restrict ingress traffic to specific namespaces.
**Kubernetes NetworkPolicy Example:**
```yaml
apiVersion:networking.k8s.io/v1
kind:NetworkPolicy
metadata:
name:allow-internal-otel
namespace:observability
spec:
podSelector:
matchLabels:
app:opentelemetry-collector
policyTypes:
- Ingress
ingress:
- from:
- namespaceSelector:
matchLabels:
environment:production
ports:
- protocol:TCP
port:4317# gRPC
- protocol:TCP
port:4318# HTTP
```
### **Approach B: Transport Layer Security (Mutual TLS / mTLS)**
Require clients to present a valid cryptographic certificate to connect to the Collector.
**Collector Configuration (mTLS):**
```yaml
receivers:
otlp:
protocols:
grpc:
endpoint:0.0.0.0:4317
tls:
# Setting client_ca_file makes the collector require and verify a
# client cert, rejecting connections without a trusted one.
client_ca_file:/certs/client_ca.pem# CA that signs trusted client certs
Available routes to build on top of: https://github.com/XRPLF/rippled/pull/6425#discussion_r3234751995
---
# Analysis: Applying the Guide to xrpld
The guide above is written for HTTP-fronted web services. xrpld is a P2P node daemon, so the threat model and the applicable defenses differ. This section captures how each approach maps to xrpld and the chosen direction.
## Threat Model
xrpld has **two distinct attack surfaces**, not one. The original guide conflates them under "trace context spoofing"; for xrpld they need separate defenses.
| **Collector ingress** (xrpld → collector) | Anyone who can reach `4317`/`4318` on the collector host | Forged OTLP traffic, telemetry exfiltration, DoS on collector | mTLS + network policy |
| **Peer trace context** (peer → xrpld) | Malicious peer in the XRPL overlay | Crafted `protocol::TraceContext` field inside peer protobuf messages (TMTransaction, consensus, etc.) — used to forge `trace_id`/`span_id`, pollute p99, attach spans to historical traces | Validate + rate-limit at the receive boundary |
**Deployment context:** Across-network. xrpld nodes (potentially run by external operators or in different DCs) ship telemetry to a centrally-hosted collector across an untrusted network. The collector is NOT on the same host or private VPC as every node.
## Part 1 Applicability — Peer Trace-Context Validation
The guide's NGINX header stripping and OTTL stale-span filtering target HTTP gateways and post-hoc cleanup. Neither fits xrpld directly:
- **NGINX header stripping** — N/A. There is no HTTP gateway between peers and xrpld; trace context arrives inside protobuf peer messages (`protocol::TraceContext`), not as W3C `traceparent` headers. See [src/xrpld/telemetry/PropagationHelpers.h](../src/xrpld/telemetry/PropagationHelpers.h).
- **OTTL stale-span filtering** — Weak fit. Post-hoc cleanup at the collector loses peer identity (you can't tell _which_ peer poisoned the trace). Validation at the receive site is stronger.
**xrpld-specific Part 1 mitigations:**
1.**Validate extracted context at the boundary** in [src/xrpld/telemetry/ConsensusReceiveTracing.h](../src/xrpld/telemetry/ConsensusReceiveTracing.h) and any other peer-message receive site. Reject if `trace_id` is all-zero, wrong length, or fails W3C format checks. Treat invalid context as "no propagated context" — start a fresh span — rather than dropping the message.
2.**Per-peer sample rate limiting** so a hostile peer cannot flood the collector with spans bearing a fabricated `trace_id`. Use probabilistic sampling on the receive path keyed by peer identity.
## Part 2 — Comparison of Collector Hardening Approaches
Evaluated for the across-network deployment shape:
| Approach | Across-network fit | Cost | Verdict |
| **A. NetworkPolicy / firewall** | Necessary baseline (don't expose `4317`/`4318` to the internet), but insufficient on its own when traffic genuinely crosses networks — you cannot NetworkPolicy the public internet. | Cheap. | **Defense-in-depth, not primary.** |
| **B. mTLS** | Strongest fit. Every xrpld node holds a client cert; the collector verifies it via `client_ca_file` in the receiver's `tls` block. Encrypts in transit (raw OTLP over the internet leaks transaction patterns and validator identity). Compromised node = revoke one cert, no shared secret to rotate everywhere. | Cert issuance + rotation pipeline. | **Primary.** |
| **C. Basic Auth** | Worst shape for this topology. Single shared password across all xrpld nodes — one leaked node config compromises the whole fleet. Doesn't encrypt; you'd need TLS underneath anyway, at which point you're 80% of the way to mTLS. | Cheap to set up, expensive to operate (rotation across N operators). | **Skip.** |
## Decision
**Primary defense:** mTLS (Approach B) on the collector's OTLP receivers. The collector requires and verifies each client certificate when `client_ca_file` is set in the receiver's `tls` block (there is no `auth_type` field — setting `client_ca_file` is what enforces client-cert verification).
**Defense-in-depth:** NetworkPolicy / firewall rules (Approach A) so `4317`/`4318` are never reachable from outside the expected operator subnets even if mTLS were misconfigured.
**Skipped:** Basic Auth (Approach C) — wrong shape for an across-network, multi-operator topology.
**Plus xrpld-specific Part 1 work:** trace-context validation and per-peer rate limiting at peer-message receive sites.
| Cert source for mTLS | **Reuse XRPL node identity key** | One identity per node, no separate PKI to operate. Fits XRPL's existing trust model; requires small CA tooling step to derive/sign the OTel client cert from the node key. |
| Part 1 scope | **Include in this spec** | Collector hardening and peer trace-context validation share one threat model. Coherent design doc; can still be split into multiple PRs at implementation. |
| Dev impact | **Production-only** | Local `docker/telemetry/docker-compose.yml` keeps `insecure: true` and no auth for fast iteration. Only production deployment manifests gain mTLS. Accepted risk: minor dev/prod drift, mitigated by integration tests against a TLS-enabled collector in CI. |
## Out of Scope
- NGINX/Envoy header stripping (no HTTP gateway in front of xrpld-to-collector traffic).
- OTTL stale-span filtering at the collector (weaker than source validation; loses peer identity).
- Local development docker-compose hardening.
- Telemetry backend (Tempo) hardening — separate concern, downstream of the collector.
## Next Step
Write this up as a design doc with full sections covering:
1. Threat model & architecture (this section, expanded)
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
