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.
12 KiB
Phase 2: RPC Tracing Completion Task List
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).Branch:
pratik/otel-phase2-rpc-tracing(frompratik/otel-phase1c-rpc-integration)
Related Plan Documents
| Document | Relevance |
|---|---|
| 03-implementation-strategy.md | Code structure and instrumentation patterns (replaces the deleted 04-code-samples.md §4.4.2 / §4.5.3, removed by d6450631bf) |
| 02-design-decisions.md | W3C Trace Context (§2.5), span attributes (§2.4.2) |
| 06-implementation-phases.md | Phase 2 tasks (§6.3), definition of done (§6.11.2) |
Task 2.1: W3C Trace Context HTTP Header Extraction
Status: DEFERRED → Phase 3
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.
Implemented in: pratik/otel-phase3-tx-tracing — TraceContextPropagator.h/.cpp
Task 2.2: Per-Category Span Creation
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.
What to do:
-
Edit
include/xrpl/telemetry/Telemetry.h:- Add
virtual bool shouldTraceLedger() const = 0;
- Add
-
Update all implementations:
src/libxrpl/telemetry/Telemetry.cpp(TelemetryImpl, NullTelemetryOtel)src/libxrpl/telemetry/NullTelemetry.cpp(NullTelemetry)
Key modified files:
include/xrpl/telemetry/Telemetry.hsrc/libxrpl/telemetry/Telemetry.cppsrc/libxrpl/telemetry/NullTelemetry.cpp
Task 2.4: Unit Tests for Core Telemetry Infrastructure
Status: COMPLETE
Objective: Add unit tests for the core telemetry abstractions to validate correctness and catch regressions.
Implemented:
-
src/tests/libxrpl/telemetry/TelemetryConfig.cpp:- Test Setup defaults (all fields have correct initial values)
- Test
makeTelemetrySetupconfig parser (empty section, full section, edge cases) - Test
samplingRatioclamping (values outside 0.0-1.0)
-
src/tests/libxrpl/telemetry/SpanGuardFactory.cpp:- Test null guard methods are safe (setAttribute, setOk, setError, addEvent on null)
- Test category span returns null when telemetry disabled
- Test child/linked span null when no parent context
- Test move construction transfers ownership
- Test recordException safe on null guard
- Test discard() safe on null guard
-
src/tests/libxrpl/telemetry/main.cpp— GTest runner -
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.methodis always POST for JSON-RPCnet.peer.ipis debug-level info available in logsduration_msis 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:
- Build with
telemetry=ONand verify no compilation errors - Build with
telemetry=OFFand verify no regressions - Run existing unit tests to verify no breakage
- Document any build issues in lessons.md
Verification Checklist:
conan install . --build=missing -o telemetry=Truesucceedscmake --preset default -Dtelemetry=ONconfigures correctly- Build succeeds with telemetry ON
- Build succeeds with telemetry OFF
- Existing tests pass with telemetry ON
- Existing tests pass with telemetry OFF
Task 2.8: RPC Span Attribute Enrichment — Node Health Context
Status: DROPPED.
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— wrapsdoPathFind()anddoRipplePathFind()RPC handlerspathfind.compute— wrapsPathRequest::doUpdate()(pathfind_fastattr)pathfind.update_all— wrapsPathRequestManager::updateAll()on ledger close (pathfind_ledger_index,pathfind_num_requestsattrs; emitted only when active subscriptions exist)pathfind.discover— wraps the entire per-source-asset loop inPathRequest::findPaths()(pathfind_search_level,pathfind_num_pathsattrs). 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.
New file: src/xrpld/rpc/detail/PathFindSpanNames.h
Modified files:
src/xrpld/rpc/handlers/orderbook/PathFind.cppsrc/xrpld/rpc/handlers/orderbook/RipplePathFind.cppsrc/xrpld/rpc/detail/PathRequest.cppsrc/xrpld/rpc/detail/PathRequestManager.cppsrc/xrpld/rpc/detail/Pathfinder.cpp
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.
Attributes added:
| Span | Attribute | Type | Source |
|---|---|---|---|
rpc.http_request |
request_payload_size |
int64 | request.body().size() |
rpc.process |
is_batch |
bool | method == "batch" check |
rpc.process |
batch_size |
int64 | params.size() (only when batch) |
rpc.ws_message |
command |
string | jv[command] or jv[method] |
rpc.command.* |
load_type |
string | context.loadType.label() |
pathfind.compute |
pathfind_dest_currency |
string | to_string(saDstAmount_.asset()) |
pathfind.discover |
pathfind_num_source_assets |
int64 | sourceAssets.size() |
Note: pathfind_dest_amount was removed — the destination amount is a financial value excluded by the privacy policy (design §2.4.4).
New attr keys: RpcSpanNames.h (isBatch, batchSize, loadType), PathFindSpanNames.h (destCurrency, numSourceAssets).
Modified files:
src/xrpld/rpc/detail/RpcSpanNames.hsrc/xrpld/rpc/detail/PathFindSpanNames.hsrc/xrpld/rpc/detail/ServerHandler.cppsrc/xrpld/rpc/detail/RPCHandler.cppsrc/xrpld/rpc/detail/PathRequest.cpp
Summary
| Task | Description | Status | Notes |
|---|---|---|---|
| 2.1 | W3C Trace Context header extraction | Deferred → Phase 3 | No consumer in Phase 2; needs cross-node tracing |
| 2.2 | Per-category span creation | Complete (Phase 1c) | Superseded by TraceCategory enum + SpanGuard |
| 2.3 | Add shouldTraceLedger() interface method | Complete (Phase 1c) | Delivered in Phase 1c base branch |
| 2.4 | Unit tests for core telemetry | Complete | TelemetryConfig + SpanGuardFactory tests |
| 2.5 | Enhanced RPC span attributes (HTTP-level) | Deferred | Low value; span duration covers timing natively |
| 2.6 | Build verification and performance baseline | Complete | Verified in CI on Phase 1c |
| 2.7 | Grafana Tempo search filters | Complete | rpc-command, rpc-status, rpc-role filters |
| 2.8 | RPC span attribute enrichment (node health) | Dropped | Available via server_info/server_state RPC |
| 2.9 | PathFind RPC instrumentation | Complete | request, compute, update_all, discover |
| 2.10 | RPC/PathFind span attribute gap fill | Complete | Batch detection, payload size, load cost, pathfind params |
Delivered in this branch: Tasks 2.4, 2.7, 2.9, 2.10. Deferred with rationale: Tasks 2.1 (→Phase 3), 2.5 (low priority). Dropped: Task 2.8 (node health not duplicated on traces). Superseded: Task 2.2 (Phase 1c SpanGuard factory covers this).
Known Issues / Future Work
Thread safety of TelemetryImpl::stop() vs startSpan()
TelemetryImpl::stop() resets sdkProvider_ (a std::shared_ptr) without
synchronization. getTracer() reads the same member from RPC handler threads.
This is a data race if any thread calls startSpan() concurrently with stop().
Current mitigation: Application::stop() shuts down serverHandler_,
overlay_, and jobQueue_ before calling telemetry_->stop(), so no callers
remain. See comments in Telemetry.cpp:stop() and Application.cpp.
TODO: Add an std::atomic<bool> stopped_ flag checked in getTracer() to
make this robust against future shutdown order changes.
Macro incompatibility: XRPL_TRACE_SPAN vs XRPL_TRACE_SET_ATTR
XRPL_TRACE_SPAN and XRPL_TRACE_SPAN_KIND declare _xrpl_guard_ as a bare
SpanGuard, but XRPL_TRACE_SET_ATTR and XRPL_TRACE_EXCEPTION call
_xrpl_guard_.has_value() which requires std::optional<SpanGuard>. Using
XRPL_TRACE_SPAN followed by XRPL_TRACE_SET_ATTR in the same scope would
fail to compile.
Current mitigation: No call site currently uses XRPL_TRACE_SPAN — all
production code uses the conditional macros (XRPL_TRACE_RPC, XRPL_TRACE_TX,
etc.) which correctly wrap the guard in std::optional.
TODO: Either make XRPL_TRACE_SPAN/XRPL_TRACE_SPAN_KIND also wrap in
std::optional, or document that XRPL_TRACE_SET_ATTR is only compatible with
the conditional macros.