rippled/docker/telemetry/workload/validate_telemetry.py

#!/usr/bin/env python3
"""Telemetry Validation Suite for xrpld.

Validates that the full telemetry stack is emitting expected data after
a workload run. Queries Tempo (spans), Prometheus (metrics), Loki (logs),
and Grafana (dashboards) APIs to produce a pass/fail report.

Validation categories:
  1. Span validation     — Every required span type in expected_spans.json, each
                           carrying its required attributes
  2. Metric validation   — SpanMetrics, StatsD, and MetricsRegistry OTLP metrics
                           are non-zero, and each group's required_labels reach
                           Prometheus with non-empty values
  3. Sync diagnostics    — Fresh-node sync signals (bootstrap + acquire
                           pipeline) declared in the "sync_diagnostics" group
  4. Log-trace correlation — Loki logs contain trace_id/span_id fields
  5. Dashboard validation — Every dashboard uid in expected_metrics.json
                           provisions and loads (panel count only, not panel data)
  6. External parity     — Span attrs, metric existence, and value sanity for
                           external dashboard parity (validator-health,
                           peer-quality, node-health)

Usage:
    python3 validate_telemetry.py --report /tmp/validation-report.json

    # Custom API endpoints:
    python3 validate_telemetry.py \\
        --tempo http://localhost:3200 \\
        --prometheus http://localhost:9090 \\
        --loki http://localhost:3100 \\
        --grafana http://localhost:3000
"""

import argparse
import asyncio
import fnmatch
import json
import logging
import re
import sys
import time
from dataclasses import dataclass, field
from pathlib import Path
from typing import Any

import aiohttp

# Loki's default query window is the last hour. A validation run finishes in
# minutes, but bounding the range explicitly keeps the query reproducible when
# someone re-runs it later to investigate a result.
LOG_QUERY_WINDOW_SECONDS = 4 * 60 * 60

# Loki's OTLP ingestion promotes service.name to the stream label
# `service_name`; a `job` attribute arrives as structured metadata, which a
# stream selector cannot match (see otel-collector-config.yaml). Declared once
# so both log checks and the diagnostic below report on the same query -- a
# diagnostic that queried something else would describe a different failure
# than the one being investigated.
LOG_STREAM_SELECTOR = '{service_name="xrpld"}'
LOG_TRACE_LINE_FILTER = '|= "trace_id="'
LOG_CORRELATION_QUERY = f"{LOG_STREAM_SELECTOR} {LOG_TRACE_LINE_FILTER}"

logger = logging.getLogger("validate_telemetry")

# ---------------------------------------------------------------------------
# Configuration defaults
# ---------------------------------------------------------------------------

DEFAULT_TEMPO = "http://localhost:3200"
DEFAULT_PROMETHEUS = "http://localhost:9090"
DEFAULT_LOKI = "http://localhost:3100"
DEFAULT_GRAFANA = "http://localhost:3000"

SCRIPT_DIR = Path(__file__).parent
EXPECTED_SPANS_FILE = SCRIPT_DIR / "expected_spans.json"
EXPECTED_METRICS_FILE = SCRIPT_DIR / "expected_metrics.json"

# Some beast::insight gauges/counters (ledger-age, peer-finder, overlay
# traffic) only populate after the node validates ledgers and sustains peer
# traffic, then travel a 1s periodic OTLP export + a 15s Prometheus scrape
# before they are queryable. On a slow CI runner the fixed post-workload wait
# can end before that pipeline settles, so a single query races and reports 0
# series. Poll each missing metric for up to this long (covering two scrape
# cycles) before failing, so the check is robust to runner speed.
METRIC_POLL_TIMEOUT_SEC = 45.0
METRIC_POLL_INTERVAL_SEC = 5.0

# Group key in expected_metrics.json holding the fresh-node sync-diagnostics
# metrics (bootstrap + acquire pipeline). Owned by
# assert_sync_diagnostics_metrics() so those signals get their own report
# category and a single, explicit failure per missing metric.
SYNC_DIAGNOSTICS_GROUP = "sync_diagnostics"

# All metrics are polled concurrently against ONE shared deadline, so the
# metric phase costs a single poll window instead of one per metric. This caps
# how many /api/v1/series requests are in flight at a time, so the fan-out does
# not hammer the single-container Prometheus the harness runs.
METRIC_POLL_CONCURRENCY = 8

# Bound on ONE HTTP request to Tempo, Prometheus, Loki or Grafana. aiohttp's
# own default is total=300s, which is longer than any poll window here: a
# single wedged endpoint would blow the shared deadline above and then keep the
# run alive until the CI job's own budget killed it, losing the report and the
# artifacts with it. Failing one request fast and reporting it beats being
# killed with nothing.
#
# Derived from the poll window rather than picked: no single request may outlast
# the phase budget it sits inside, since a request that does can only ever blow
# that deadline. Connecting is held to one poll interval, because an endpoint
# that is absent or wedged should be named immediately, not waited on.
REQUEST_TIMEOUT = aiohttp.ClientTimeout(
    total=METRIC_POLL_TIMEOUT_SEC, sock_connect=METRIC_POLL_INTERVAL_SEC
)

# The Prometheus exporter splits one histogram instrument into three series
# names. Reverse coverage folds them back onto the base family so a contract
# entry (or an accounted_patterns regex) written for the family accounts for
# all three, and so a triple is never reported as three separate gaps.
HISTOGRAM_SUFFIXES = ("_bucket", "_count", "_sum")


# ---------------------------------------------------------------------------
# Data classes
# ---------------------------------------------------------------------------


@dataclass
class CheckResult:
    """Result of a single validation check.

    Attributes:
        name:     Check identifier (e.g., "span.rpc.ws_message").
        category: Validation category (span, metric, log, dashboard).
        passed:   Whether the check passed.
        message:  Human-readable description of the result.
        details:  Optional additional data (counts, values, etc.).
    """

    name: str
    category: str
    passed: bool
    message: str
    details: dict[str, Any] = field(default_factory=dict)

    def to_dict(self) -> dict[str, Any]:
        """Serialize to a JSON-compatible dict."""
        return {
            "name": self.name,
            "category": self.category,
            "passed": self.passed,
            "message": self.message,
            "details": self.details,
        }


@dataclass
class ValidationReport:
    """Aggregated validation report.

    Attributes:
        checks:     List of all individual check results.
        start_time: ISO timestamp when validation started.
        end_time:   ISO timestamp when validation completed.
    """

    checks: list[CheckResult] = field(default_factory=list)
    start_time: str = ""
    end_time: str = ""

    @property
    def total_checks(self) -> int:
        """Total number of checks executed."""
        return len(self.checks)

    @property
    def passed(self) -> int:
        """Number of checks that passed."""
        return sum(1 for c in self.checks if c.passed)

    @property
    def failed(self) -> int:
        """Number of checks that failed."""
        return sum(1 for c in self.checks if not c.passed)

    @property
    def all_passed(self) -> bool:
        """Whether all checks passed."""
        return self.failed == 0

    def add(self, check: CheckResult) -> None:
        """Add a check result to the report."""
        self.checks.append(check)
        status = "PASS" if check.passed else "FAIL"
        logger.info("[%s] %s: %s", status, check.name, check.message)

    def to_dict(self) -> dict[str, Any]:
        """Serialize to a JSON-compatible dict."""
        return {
            "summary": {
                "total": self.total_checks,
                "passed": self.passed,
                "failed": self.failed,
                "all_passed": self.all_passed,
            },
            "start_time": self.start_time,
            "end_time": self.end_time,
            "checks": [c.to_dict() for c in self.checks],
        }


# ---------------------------------------------------------------------------
# Reverse coverage: what is emitted but not accounted for
# ---------------------------------------------------------------------------
#
# The forward checks read the contract and ask Prometheus/Tempo whether each
# listed name exists. That direction cannot see a name the contract omits, and
# omissions are how a 345-family metric gap and 7 unknown spans went unnoticed:
# both emitted inventories were already being fetched, and neither was compared
# back against the contract.
#
# These helpers close the loop. They are deliberately WARN-ONLY. Downstream
# branches legitimately add telemetry that an upstream contract has not seen
# (the sync-diagnostics branch emits 7 spans this contract does not list), so a
# hard failure here would redden every one of them for doing the right thing.
# The value is visibility: name the gaps, in a form a human can read in a CI
# log and diff between runs, and let a person decide.


def _log_name_list(header: str, names: list[str]) -> None:
    """Log a name list one entry per line, sorted.

    A single-line Python list repr of 422 metric families is ~15 kB on one CI
    log line: unreadable, and impossible to diff between runs. One name per line
    makes two runs' logs comparable. Emitted as a single log record so the
    lines cannot be interleaved by another task's output.

    Args:
        header: Line printed above the list; the count is appended to it.
        names:  Names to print. Sorted here, so callers need not be.
    """
    body = "\n".join(f"  {name}" for name in sorted(names)) or "  (none)"
    logger.info("%s (%d total):\n%s", header, len(names), body)


def _reverse_coverage_result(
    check_name: str,
    category: str,
    noun: str,
    emitted_count: int,
    unaccounted: list[str],
) -> CheckResult:
    """Build the always-passing CheckResult for one reverse coverage check.

    ``passed`` is hardcoded True. This is the single place that decides the
    check cannot fail CI, so the guarantee is auditable in one line rather than
    spread over two call sites. The finding travels in ``message`` and in
    ``details["unaccounted"]``, and the names are also logged one per line by
    the caller.

    Args:
        check_name:    Report check name (e.g. "metric.reverse_coverage").
        category:      Report category ("metric" or "span").
        noun:          Plural noun for the message ("metric families", "span names").
        emitted_count: How many names the backend reported.
        unaccounted:   Names the contract does not account for.

    Returns:
        A CheckResult that always passes.
    """
    if emitted_count == 0:
        message = (
            f"reverse coverage not evaluated: no {noun} were reported "
            f"(backend unreachable or empty) — warning only, never fails"
        )
    elif unaccounted:
        message = (
            f"WARNING: {len(unaccounted)} of {emitted_count} emitted {noun} are "
            f"not accounted for by the contract: {', '.join(sorted(unaccounted)[:5])}"
            f"{' …' if len(unaccounted) > 5 else ''} "
            f"(full list logged above; warning only, never fails)"
        )
    else:
        message = f"all {emitted_count} emitted {noun} are accounted for"

    return CheckResult(
        name=check_name,
        category=category,
        passed=True,
        message=message,
        details={
            "emitted": emitted_count,
            "accounted": emitted_count - len(unaccounted),
            "unaccounted_count": len(unaccounted),
            "unaccounted": sorted(unaccounted),
            "enforced": False,
        },
    )


# ---------------------------------------------------------------------------
# Tempo API helpers
# ---------------------------------------------------------------------------


def _log_query_window() -> dict[str, str]:
    """Loki query_range bounds covering a validation run.

    Returns:
        start/end parameters in nanoseconds since the epoch.
    """
    now = time.time()
    return {
        "start": str(int((now - LOG_QUERY_WINDOW_SECONDS) * 1_000_000_000)),
        "end": str(int(now * 1_000_000_000)),
    }


class TempoQueryError(RuntimeError):
    """Tempo answered a query with a failure, as distinct from "nothing found".

    The two must not be conflated. An empty search result and a 404 on a trace
    id are legitimate answers meaning the data is not there; an HTTP 5xx, a 401
    or a 400 mean the question was never answered, and reporting that as "0
    traces" or "0 spans" turns a broken backend into a green-looking negative
    result. Callers that loop over candidates catch absence and move on; this
    exception is what they must NOT swallow.
    """


def _short_body(text: str, limit: int = 300) -> str:
    """Collapse a response body to one bounded line, for a log or a message."""
    return " ".join(text.split())[:limit] or "(empty body)"


async def _tempo_search(
    session: aiohttp.ClientSession,
    tempo_url: str,
    query: str,
    limit: int = 20,
) -> list[dict[str, Any]]:
    """Search traces in Tempo using TraceQL.

    Args:
        session:   aiohttp client session.
        tempo_url: Base URL for Tempo API (e.g., http://localhost:3200).
        query:     TraceQL query string.
        limit:     Maximum number of traces to return.

    Returns:
        List of trace summary dicts from Tempo search results.
    """
    params = {"q": query, "limit": str(limit)}
    async with session.get(f"{tempo_url}/api/search", params=params) as resp:
        # /api/search answers 200 with an empty (or absent) "traces" list when
        # nothing matches, so there is no "legitimately absent" status to
        # tolerate here: every non-200 is a real failure. Left unchecked,
        # resp.json() on an error body yields a dict with no "traces" key and
        # this returns [], which every caller reports as "no traces found" --
        # indistinguishable from a healthy Tempo holding nothing. That is the
        # same silent-failure shape as a span query that returned 200 with an
        # empty list, and it must not be reproduced here.
        if resp.status != 200:
            raise TempoQueryError(
                f"GET /api/search returned HTTP {resp.status}: "
                f"{_short_body(await resp.text())}"
            )
        data = await resp.json()
        return data.get("traces", [])


async def _tempo_get_trace(
    session: aiohttp.ClientSession,
    tempo_url: str,
    trace_id: str,
) -> list[dict[str, Any]]:
    """Fetch a full trace from Tempo by trace ID.

    Returns the list of spans extracted from the OTLP-format response.

    Args:
        session:   aiohttp client session.
        tempo_url: Tempo API base URL.
        trace_id:  Hex trace ID string.

    A 404 means Tempo holds no such trace, which is returned as an empty list
    rather than raised: see the note in the body. Any other non-200 raises
    TempoQueryError.

    Returns:
        Flat list of span dicts as Tempo returned them, carrying at least
        'name', 'attributes', 'spanId' and, for non-root spans, 'parentSpanId'.
        Empty when Tempo has no trace with this id.

    Raises:
        TempoQueryError: Tempo answered with a status other than 200 or 404.
    """
    async with session.get(f"{tempo_url}/api/traces/{trace_id}") as resp:
        # 404 is Tempo's answer for "no trace with that id", and for an id read
        # out of a log line that is an ordinary outcome, not an error: the span
        # may not have been exported yet, or its block may not be searchable.
        # Every caller either loops over candidate ids or over several traces
        # and relies on an empty list meaning "not this one", so a 404 must stay
        # absent -- raising here would turn a normal miss into a check failure
        # and abort the loop over the remaining candidates.
        if resp.status == 404:
            return []
        # Anything else is a real failure. Unchecked, resp.json() on an error
        # body yields a dict with no "batches" key, so this returned [] and the
        # caller reported "0 spans" -- a backend fault laundered into a
        # negative result.
        if resp.status != 200:
            raise TempoQueryError(
                f"GET /api/traces/{trace_id} returned HTTP {resp.status}: "
                f"{_short_body(await resp.text())}"
            )
        data = await resp.json()
        spans: list[dict[str, Any]] = []
        for batch in data.get("batches", []):
            for scope_spans in batch.get("scopeSpans", []):
                spans.extend(scope_spans.get("spans", []))
        return spans


def _otlp_span_attr_keys(span: dict[str, Any]) -> set[str]:
    """Extract all attribute key names from an OTLP span.

    Args:
        span: OTLP span dict with an 'attributes' list.

    Returns:
        Set of attribute key strings.
    """
    return {a["key"] for a in span.get("attributes", []) if "key" in a}


def _traceql_name_predicate(expected_name: str) -> str:
    """Build the TraceQL `name` predicate that selects a contract span name.

    A literal name becomes an equality test. A glob becomes a regex, since
    TraceQL has no glob operator: `rpc.command.*` is sent as
    `name=~"rpc[.]command[.].*"`.

    Dots use the `[.]` character class, not `\\.`: TraceQL's string lexer
    rejects an escape it does not recognise, so `\\.` returns HTTP 400. Bare
    dots would parse but match any character there.

    Args:
        expected_name: Span name or glob from expected_spans.json.

    Returns:
        A TraceQL predicate on the bare `name` intrinsic, without braces.
    """
    if "*" not in expected_name:
        return f'name="{expected_name}"'
    # Span names are lower_snake_case segments joined by dots, so `.` and `*` are
    # the only characters here that mean anything to a regex engine. Anything
    # else appearing would need its own handling rather than silent passthrough.
    unexpected = set(expected_name) - set("abcdefghijklmnopqrstuvwxyz0123456789_.*")
    if unexpected:
        raise ValueError(
            f"span name {expected_name!r} contains {sorted(unexpected)}, which "
            "this predicate builder does not know how to escape for TraceQL"
        )
    pattern = "".join(
        ".*" if ch == "*" else "[.]" if ch == "." else ch for ch in expected_name
    )
    return f'name=~"{pattern}"'


def _span_name_matches(emitted_name: str, expected_name: str) -> bool:
    """Test an emitted span name against a name from expected_spans.json.

    Contract names are either literals or globs containing "*" (for example
    "rpc.command.*"). Literals are compared for exact equality so a longer
    emitted name cannot satisfy a shorter contract: "consensus.accept.apply"
    must not stand in for "consensus.accept".

    Args:
        emitted_name:  Span name as reported by Tempo.
        expected_name: Span name or glob pattern from expected_spans.json.

    Returns:
        True when the emitted name satisfies the expected name.
    """
    if "*" in expected_name:
        return fnmatch.fnmatchcase(emitted_name, expected_name)
    return emitted_name == expected_name


def _unaccounted_span_names(emitted: list[str], expected: dict[str, Any]) -> list[str]:
    """Names Tempo reports that no expected_spans.json entry accounts for.

    "Accounted for" is the same relation the forward check uses, so a contract
    glob such as "rpc.command.*" covers every command it expands to, and an
    entry marked ``optional`` still accounts for its name — being unasserted is
    not the same as being unknown. No separate pattern list is needed for spans
    because the contract already carries globs.

    Args:
        emitted:  Span names as reported by Tempo's span.name tag values.
        expected: The parsed expected_spans.json contract.

    Returns:
        Sorted list of emitted names with no matching contract entry.
    """
    contract = [span_def["name"] for span_def in expected.get("spans", [])]
    return sorted(
        name
        for name in emitted
        if name
        and not any(
            _span_name_matches(name, expected_name) for expected_name in contract
        )
    )


# ---------------------------------------------------------------------------
# Span Validation (Tempo API)
# ---------------------------------------------------------------------------


async def validate_spans(
    session: aiohttp.ClientSession,
    tempo_url: str,
    report: ValidationReport,
) -> None:
    """Validate that all expected spans appear in Tempo.

    Queries the Tempo TraceQL API for each expected span name and checks
    that traces exist. Also validates required attributes on spans and
    parent-child relationships.

    Args:
        session:   aiohttp client session.
        tempo_url: Base URL for Tempo API (e.g., http://localhost:3200).
        report:    ValidationReport to accumulate results.
    """
    logger.info("--- Span Validation (Tempo) ---")

    # Load expected spans.
    with open(EXPECTED_SPANS_FILE) as f:
        expected = json.load(f)

    # Check service registration.
    try:
        async with session.get(
            f"{tempo_url}/api/v2/search/tag/resource.service.name/values"
        ) as resp:
            data = await resp.json()
            tag_values = data.get("tagValues", [])
            services = [tv.get("value", "") for tv in tag_values]
            has_xrpld = "xrpld" in services
            report.add(
                CheckResult(
                    name="span.service_registration",
                    category="span",
                    passed=has_xrpld,
                    message=(
                        f"Service 'xrpld' registered (found: {services})"
                        if has_xrpld
                        else f"Service 'xrpld' NOT found (found: {services})"
                    ),
                )
            )
    except Exception as exc:
        report.add(
            CheckResult(
                name="span.service_registration",
                category="span",
                passed=False,
                message=f"Tempo API unreachable: {exc}",
            )
        )
        return

    # List every span name Tempo holds. This is both a diagnostic (it makes a
    # missing-span failure debuggable without reproducing the stack locally)
    # and the input to the reverse coverage check added at the end of this
    # function. Note the tag-values API is not service-scoped, so in a stack
    # where something other than xrpld also sent traces this list would be a
    # superset; on the harness only xrpld exports spans.
    #
    # The tag is the bare intrinsic `name`, NOT `span.name`. A span's name is a
    # TraceQL intrinsic, not a span-scoped attribute, so `span.name` resolves to
    # an attribute nothing sets and Tempo answers 200 with an empty tagValues
    # list -- indistinguishable from an empty backend, which is how the wrong
    # tag went unnoticed. Verified against tempo 2.9.4 holding one span:
    # `span.name` -> {"tagValues":[]}, `name` -> that span's name.
    emitted_span_names: list[str] = []
    try:
        async with session.get(f"{tempo_url}/api/v2/search/tag/name/values") as resp:
            ops_data = await resp.json()
            tag_values = ops_data.get("tagValues", [])
            emitted_span_names = [tv.get("value", "") for tv in tag_values]
            _log_name_list("Tempo span names", emitted_span_names)
    except Exception as exc:
        logger.warning("Failed to fetch Tempo operations: %s", exc)

    # Concrete probe names for wildcard span entries. Exact-match TraceQL can't
    # match a literal "*", so a representative operation name is substituted.
    # Wildcards without a known concrete example (e.g. grpc.<MethodName> when no
    # gRPC client runs) are skipped when marked optional.
    wildcard_probes = {"rpc.command.*": "rpc.command.server_info"}

    # Check each expected span.
    for span_def in expected["spans"]:
        span_name = span_def["name"]
        is_optional = span_def.get("optional", False)
        check_name = f"span.{span_name}"

        if "*" in span_name:
            operation = wildcard_probes.get(span_name)
            if operation is None:
                # No concrete probe. Optional wildcards (e.g. grpc.*) are skipped;
                # a required one would be a config error worth surfacing.
                if is_optional:
                    logger.info(
                        "[SKIP] %s: optional wildcard span with no concrete "
                        "probe (not exercised by the workload)",
                        check_name,
                    )
                    continue
                report.add(
                    CheckResult(
                        name=check_name,
                        category="span",
                        passed=False,
                        message=f"{span_name}: required wildcard has no probe name",
                    )
                )
                continue
        else:
            operation = span_name

        try:
            query = '{resource.service.name="xrpld" && name="' + operation + '"}'
            traces = await _tempo_search(session, tempo_url, query, limit=5)
            count = len(traces)
            # Optional spans only fire under specific traffic (mode changes,
            # missing-ledger fetch, fee escalation). Absence is not a failure —
            # mirror the parent-child "skip" handling so CI stays green.
            if count == 0 and is_optional:
                logger.info(
                    "[SKIP] %s: optional span not emitted under this workload",
                    check_name,
                )
                report.add(
                    CheckResult(
                        name=check_name,
                        category="span",
                        passed=True,
                        message=f"{span_name}: optional, not emitted (skipped)",
                        details={"trace_count": 0, "optional": True},
                    )
                )
                continue
            report.add(
                CheckResult(
                    name=check_name,
                    category="span",
                    passed=count > 0,
                    message=(
                        f"{span_name}: {count} traces found"
                        if count > 0
                        else f"{span_name}: 0 traces (expected > 0)"
                    ),
                    details={"trace_count": count},
                )
            )

            # Validate required attributes on first trace.
            if count > 0 and span_def.get("required_attributes"):
                await _check_attributes_on_first_trace(
                    session, tempo_url, traces, span_def, report
                )
        except Exception as exc:
            report.add(
                CheckResult(
                    name=check_name,
                    category="span",
                    passed=False,
                    message=f"{span_name}: query failed ({exc})",
                )
            )

    # Validate parent-child relationships.
    for rel in expected.get("parent_child_relationships", []):
        # Skip relationships marked with "skip: true" (e.g., cross-thread
        # parent-child that requires a C++ fix to propagate span context).
        if rel.get("skip", False):
            reason = rel.get("skip_reason", "marked skip in expected_spans.json")
            logger.info(
                "[SKIP] span.hierarchy.%s->%s: %s",
                rel["parent"],
                rel["child"],
                reason,
            )
            continue
        await _validate_parent_child(session, tempo_url, rel, report)

    # Reverse direction: span names Tempo holds that the contract never mentions.
    # Added last so no existing check's position in the report moves.
    unaccounted_spans = _unaccounted_span_names(emitted_span_names, expected)
    if unaccounted_spans:
        _log_name_list(
            "Span names emitted but NOT accounted for by expected_spans.json "
            "(warning only, does not fail CI)",
            unaccounted_spans,
        )
    report.add(
        _reverse_coverage_result(
            check_name="span.reverse_coverage",
            category="span",
            noun="span names",
            emitted_count=len(emitted_span_names),
            unaccounted=unaccounted_spans,
        )
    )


async def _check_attributes_on_first_trace(
    session: aiohttp.ClientSession,
    tempo_url: str,
    traces: list[dict[str, Any]],
    span_def: dict[str, Any],
    report: ValidationReport,
) -> None:
    """Fetch the first trace and check the span's required attributes.

    Fetching the trace is a second network call, so it carries its own error
    handling. Letting it fall through to the caller's handler would add a
    second result under the span's own check name, which has already recorded
    the trace as found -- one entry passing and one failing for the same name,
    inflating the check total and blaming the trace-existence check for an
    attribute-fetch failure.

    Args:
        session:    aiohttp client session.
        tempo_url:  Base URL for the Tempo API.
        traces:     Traces returned for this span, most recent first.
        span_def:   The span's entry from expected_spans.json.
        report:     ValidationReport to accumulate results.
    """
    span_name = span_def["name"]
    try:
        trace_id = traces[0].get("traceID", "")
        if not trace_id:
            # Recorded rather than returned on silently. Returning with no
            # result would drop this span's attribute contract out of the
            # report and shrink the check total, so the surface would look
            # smaller with nothing saying why. Reported the same way as a
            # fetched trace holding no matching span, below: being unable to
            # verify is itself the finding. The caller only reaches here for a
            # span that declares required_attributes, so this adds no check
            # where none was expected.
            report.add(
                CheckResult(
                    name=f"span.attrs.{span_name}",
                    category="span",
                    passed=False,
                    message=(
                        f"{span_name}: newest trace carried no traceID, cannot "
                        "verify its attributes"
                    ),
                )
            )
            return
        spans = await _tempo_get_trace(session, tempo_url, trace_id)
        await _validate_span_attributes_otlp(spans, span_def, report)
    except Exception as exc:
        report.add(
            CheckResult(
                name=f"span.attrs.{span_name}",
                category="span",
                passed=False,
                message=f"{span_name}: attribute check failed ({exc})",
            )
        )


async def _validate_span_attributes_otlp(
    spans: list[dict[str, Any]],
    span_def: dict[str, Any],
    report: ValidationReport,
) -> None:
    """Check that the contract's own span carries its required attributes.

    Only spans whose name matches ``span_def["name"]`` are inspected.
    Attributes are never borrowed from siblings: many span types share keys
    such as ledger_seq or tx_hash, so a trace-wide scan would satisfy every
    one of those contracts from a single carrier span and make the per-span
    contract unenforceable.

    A span type passes when at least one instance of it carries every required
    attribute. When none does, the closest instance's missing keys are
    reported.

    Args:
        spans:    Every OTLP span dict in the fetched trace.
        span_def: Span definition from expected_spans.json.
        report:   ValidationReport to accumulate results.
    """
    required_attrs = span_def.get("required_attributes", [])
    if not required_attrs:
        return

    span_name = span_def["name"]
    check_name = f"span.attrs.{span_name}"
    matching = [s for s in spans if _span_name_matches(s.get("name", ""), span_name)]

    if not matching:
        report.add(
            CheckResult(
                name=check_name,
                category="span",
                passed=False,
                message=(
                    f"{span_name}: no span named '{span_name}' in the fetched "
                    "trace, cannot verify its attributes"
                ),
                details={"required": required_attrs, "instances": 0},
            )
        )
        return

    # Keep the instance that is missing the fewest required attributes, so the
    # failure message names the closest witness rather than an arbitrary one.
    best_found: set[str] = set()
    best_missing: list[str] = list(required_attrs)
    for span in matching:
        found = _otlp_span_attr_keys(span)
        missing = [a for a in required_attrs if a not in found]
        if len(missing) < len(best_missing):
            best_found, best_missing = found, missing
        if not best_missing:
            break

    report.add(
        CheckResult(
            name=check_name,
            category="span",
            passed=not best_missing,
            message=(
                f"{span_name}: all {len(required_attrs)} attributes present"
                if not best_missing
                else f"{span_name}: no '{span_name}' span carried all "
                f"{len(required_attrs)} required attributes; closest of "
                f"{len(matching)} instance(s) missing {best_missing}"
            ),
            details={
                "required": required_attrs,
                "found": sorted(best_found),
                "missing": best_missing,
                "instances": len(matching),
            },
        )
    )


# Verdicts _span_ancestry can return, most conclusive first. A single trace
# proving "under" settles the relationship, so it wins outright. A definite
# negative outranks an indefinite one: "not_under" means a chain was walked to a
# root and the parent was not on it, while "broken_chain" only means ancestry
# could not be established, which is a different thing to go and look at.
# "no_child" ranks last because it is the starting value: a relationship with no
# candidate trace at all reports it, and every other verdict must be able to
# replace it.
_ANCESTRY_PRIORITY = ("under", "not_under", "broken_chain", "no_parent", "no_child")


def _span_ancestry(
    spans: list[dict[str, Any]],
    parent_name: str,
    child_name: str,
) -> tuple[str, str]:
    """Decide whether a matching child hangs under a matching parent in one trace.

    Walks each candidate child's parentSpanId chain upwards. Ancestry rather
    than a direct edge, because every relationship in expected_spans.json is
    worded as the parent "containing" the child: a scope appearing in between is
    a refactor, not a broken relationship.

    Span ids are compared as opaque strings and never decoded. Both fields come
    from the same Tempo response and so share whatever encoding it uses, which
    keeps this independent of whether that is hex or base64.

    Args:
        spans:       Every OTLP span dict in one fetched trace.
        parent_name: Parent name from the contract; may be a glob.
        child_name:  Child name from the contract; may be a glob.

    Returns:
        ``(verdict, detail)``. Verdict is one of ``_ANCESTRY_PRIORITY``. For
        ``broken_chain`` the detail is the span id the chain ran into.
    """
    by_id = {s["spanId"]: s for s in spans if s.get("spanId")}
    parent_ids = {
        s["spanId"]
        for s in spans
        if s.get("spanId") and _span_name_matches(s.get("name", ""), parent_name)
    }
    children = [s for s in spans if _span_name_matches(s.get("name", ""), child_name)]
    if not children:
        return "no_child", ""
    if not parent_ids:
        return "no_parent", ""

    broken = ""
    walked_to_a_root = False
    for child in children:
        current = child.get("parentSpanId", "")
        # Guards against a cycle in malformed data, which would otherwise spin
        # here forever rather than failing the check.
        seen: set[str] = set()
        while current and current not in seen:
            if current in parent_ids:
                return "under", ""
            seen.add(current)
            if (next_span := by_id.get(current)) is None:
                # This child proves nothing either way. Keep the first gap seen,
                # in case no other child yields a definite answer.
                broken = broken or current
                break
            current = next_span.get("parentSpanId", "")
        else:
            # Ran out of chain rather than hitting a gap, so this child really is
            # not under the parent. One such child is a definite negative and
            # outranks any other child's gap.
            walked_to_a_root = True
    if walked_to_a_root or not broken:
        return "not_under", ""
    return "broken_chain", broken


def _hierarchy_message(
    parent_name: str, child_name: str, verdict: str, detail: str
) -> str:
    """Phrase one ancestry verdict for the report.

    Each verdict gets its own wording because they send the reader somewhere
    different: a missing child is a workload or instrumentation gap, a child
    that is present but not under the parent is a hierarchy bug, and a broken
    chain is a span that never reached Tempo.
    """
    if verdict == "under":
        return f"Found {child_name} under {parent_name}"
    if verdict == "not_under":
        return (
            f"{child_name} shares a trace with {parent_name} but is not under "
            f"it -- co-occurrence is not a hierarchy"
        )
    if verdict == "broken_chain":
        return (
            f"{child_name}'s parent chain runs into span {detail}, which the "
            f"trace does not contain, so ancestry under {parent_name} cannot "
            f"be verified"
        )
    if verdict == "no_parent":
        return f"{parent_name} absent from traces selected for containing it"
    return f"{child_name} not found in {parent_name} traces"


async def _best_ancestry_verdict(
    session: aiohttp.ClientSession,
    tempo_url: str,
    traces: list[dict[str, Any]],
    parent_name: str,
    child_name: str,
) -> tuple[str, str]:
    """Fetch each candidate trace and keep the most conclusive verdict.

    Every candidate is examined rather than stopping at the first negative,
    because a conditional child may hang correctly in one trace while another
    carries the parent alone. Stops early once one trace proves "under", which
    nothing later can improve on.

    Names are re-checked per trace rather than trusted from the query that
    selected it: it keeps the glob handling in one place, so a wrongly built
    query cannot pass silently.

    Args:
        session:     aiohttp client session.
        tempo_url:   Base URL for the Tempo API.
        traces:      Trace summaries that matched parent and child.
        parent_name: Parent name from the contract; may be a glob.
        child_name:  Child name from the contract; may be a glob.

    Returns:
        The winning ``(verdict, detail)`` from _span_ancestry.
    """
    verdict, detail = "no_child", ""
    for trace_summary in traces:
        trace_id = trace_summary.get("traceID", "")
        if not trace_id:
            continue
        spans = await _tempo_get_trace(session, tempo_url, trace_id)
        candidate, candidate_detail = _span_ancestry(spans, parent_name, child_name)
        if _ANCESTRY_PRIORITY.index(candidate) < _ANCESTRY_PRIORITY.index(verdict):
            verdict, detail = candidate, candidate_detail
        if verdict == "under":
            break
    return verdict, detail


async def _validate_parent_child(
    session: aiohttp.ClientSession,
    tempo_url: str,
    relationship: dict[str, Any],
    report: ValidationReport,
) -> None:
    """Validate a parent-child span relationship in Tempo traces.

    Co-occurrence in a trace is the search filter, not the assertion: the check
    then walks the child's parentSpanId chain to confirm it really hangs under
    the parent. See _span_ancestry.

    Args:
        session:      aiohttp client session.
        tempo_url:    Base URL for Tempo API.
        relationship: Dict with 'parent' and 'child' span names.
        report:       ValidationReport to accumulate results.
    """
    parent_name = relationship["parent"]
    child_name = relationship["child"]

    try:
        # Query traces for the parent span. Kept as its own query so that "the
        # parent stopped being emitted" stays distinguishable from "the parent is
        # there but the child never co-occurs" — they mean different things to
        # whoever reads the report.
        query = '{resource.service.name="xrpld" && name="' + parent_name + '"}'
        traces = await _tempo_search(session, tempo_url, query, limit=3)

        if not traces:
            report.add(
                CheckResult(
                    name=f"span.hierarchy.{parent_name}->{child_name}",
                    category="span",
                    passed=False,
                    message=f"No {parent_name} traces to check hierarchy",
                )
            )
            return

        # Ask for traces holding BOTH rather than the newest parent traces. A
        # parent that fires on every close has newest traces least likely to
        # carry a conditional child. `{A} && {B}` matches at trace level, so
        # co-occurrence is found wherever it happened.
        both_query = query + " && {" + _traceql_name_predicate(child_name) + "}"
        traces = await _tempo_search(session, tempo_url, both_query, limit=3)

        verdict, detail = await _best_ancestry_verdict(
            session, tempo_url, traces, parent_name, child_name
        )
        report.add(
            CheckResult(
                name=f"span.hierarchy.{parent_name}->{child_name}",
                category="span",
                passed=verdict == "under",
                message=_hierarchy_message(parent_name, child_name, verdict, detail),
            )
        )
    except Exception as exc:
        report.add(
            CheckResult(
                name=f"span.hierarchy.{parent_name}->{child_name}",
                category="span",
                passed=False,
                message=f"Hierarchy check failed: {exc}",
            )
        )


# ---------------------------------------------------------------------------
# Trace-join Validation (Tempo API)
# ---------------------------------------------------------------------------


async def assert_trace_join_groups(
    session: aiohttp.ClientSession,
    tempo_url: str,
    report: ValidationReport,
) -> None:
    """Assert each declared trace-join group really lands in ONE trace.

    A trace-join group is a set of spans that share one trace id with NO
    parent/child link between them: each derives its trace id deterministically
    from the same hash (SpanGuard::hashSpan over a ledger hash), which is how
    spans produced on unrelated threads are joined without propagating any
    context. The parent/child check above cannot express that -- there is no
    parent to look for -- so the assertion here is co-occurrence: search for the
    group's anchor span, then require at least one of its traces to also contain
    every span in required_members.

    A regression this catches: if the join key or the deterministic-root
    mechanism breaks, each span reverts to its own single-span trace and no trace
    contains the members together, so a slow ledger is no longer readable as one
    unit. That is invisible to every other check in this harness -- the spans are
    all still emitted with all their attributes.

    An absent "trace_join_groups" key is a genuine no-op (nothing declared yet),
    not a failure: unlike the sync_diagnostics metric group, this key is optional
    and older expected_spans.json files predate it.

    Args:
        session:   aiohttp client session.
        tempo_url: Base URL for the Tempo API.
        report:    ValidationReport to accumulate results.
    """
    logger.info("--- Trace-Join Validation (Tempo) ---")

    with open(EXPECTED_SPANS_FILE) as f:
        expected = json.load(f)

    groups = expected.get("trace_join_groups", {}).get("groups", [])
    if not groups:
        logger.info("[SKIP] span.trace_join: no join groups declared")
        return

    for group in groups:
        await _validate_trace_join_group(session, tempo_url, group, report)


async def _validate_trace_join_group(
    session: aiohttp.ClientSession,
    tempo_url: str,
    group: dict[str, Any],
    report: ValidationReport,
) -> None:
    """Validate one trace-join group: anchor and members share a trace.

    Args:
        session:   aiohttp client session.
        tempo_url: Tempo API base URL.
        group:     One entry from expected_spans.json trace_join_groups.groups.
        report:    ValidationReport to accumulate results.
    """
    name = group.get("name", "<unnamed>")
    anchor = group["anchor"]
    required = list(group.get("required_members", []))
    check_name = f"span.trace_join.{name}"

    try:
        query = '{resource.service.name="xrpld" && name="' + anchor + '"}'
        traces = await _tempo_search(session, tempo_url, query, limit=10)

        if not traces:
            report.add(
                CheckResult(
                    name=check_name,
                    category="span",
                    passed=False,
                    message=f"{name}: no {anchor} traces to check the join against",
                    details={"anchor": anchor, "required_members": required},
                )
            )
            return

        # Walk the anchor's traces and keep the best result: the join holds as
        # soon as ONE trace carries every required member. Several traces are
        # examined because a given ledger may legitimately be missing an
        # optional member (e.g. a self-built ledger has no arriving validation).
        best_missing = required
        for summary in traces:
            trace_id = summary.get("traceID", "")
            if not trace_id:
                continue
            spans = await _tempo_get_trace(session, tempo_url, trace_id)
            present = {s.get("name", "") for s in spans}
            missing = [m for m in required if m not in present]
            if len(missing) < len(best_missing):
                best_missing = missing
            if not missing:
                break

        report.add(
            CheckResult(
                name=check_name,
                category="span",
                passed=not best_missing,
                message=(
                    f"{name}: {anchor} shares a trace with {required}"
                    if not best_missing
                    else (
                        f"{name}: no {anchor} trace contained {best_missing} "
                        f"-- the per-{group.get('join_key', 'hash')} trace join "
                        f"is broken (spans are landing in separate traces)"
                    )
                ),
                details={
                    "anchor": anchor,
                    "join_key": group.get("join_key"),
                    "required_members": required,
                    "missing": best_missing,
                    "traces_examined": len(traces),
                },
            )
        )
    except Exception as exc:
        report.add(
            CheckResult(
                name=check_name,
                category="span",
                passed=False,
                message=f"{name}: trace-join check failed ({exc})",
            )
        )


# ---------------------------------------------------------------------------
# Metric Validation (Prometheus API)
# ---------------------------------------------------------------------------

# Top-level keys of expected_metrics.json that validate_metrics() must not walk
# with its generic group loop: "description" is prose, "grafana_dashboards"
# holds dashboard UIDs (checked by validate_dashboards), and
# "sync_diagnostics" has its own validator, assert_sync_diagnostics_metrics().
#
# MERGE HAZARD, for whoever brings phase-10 forward into this branch. phase-10
# replaces the flatten below with _metric_check_targets(), which selects every
# group satisfying isinstance(category_data, dict) and has no equivalent of this
# tuple. Resolving that merge in phase-10's favour therefore lets
# validate_metrics walk "sync_diagnostics" as well, while
# assert_sync_diagnostics_metrics still walks it -- every metric in the group
# polled twice and reported twice, inflating the check total. The resolution must
# reinstate this exclusion inside _metric_check_targets, or fold
# assert_sync_diagnostics_metrics into it and drop the separate pass. Either is
# fine; keeping both walkers without an exclusion is not.
SKIPPED_METRIC_GROUPS = ("description", "grafana_dashboards", SYNC_DIAGNOSTICS_GROUP)


async def _log_prometheus_metric_names(
    session: aiohttp.ClientSession, prometheus_url: str
) -> list[str]:
    """Log every metric family name Prometheus knows, and return the list.

    Two jobs. It is the diagnostic that makes a name mismatch between
    expected_metrics.json and actual emissions debuggable from a CI log, and it
    is the emitted inventory the reverse coverage check compares the contract
    against. Failures are warnings, never check failures; an empty return means
    "could not determine", which the reverse check reports as not evaluated
    rather than as a clean run.

    Deliberately unfiltered. This used to keep only names matching 19
    hard-coded prefixes, which made it useless for the job it exists to do: on
    the last CI run it printed 147 of 422 families, and none of the prefixes
    covered state_accounting_*, node_family_*, overlay_peer_disconnects or the
    pathfind_* histograms, so a coverage gap in exactly those families could
    not be seen here. An allow-list can only ever show names someone already
    thought to look for, which is the opposite of what a discovery aid needs
    to do. The whole list is a few kilobytes of CI log, so there is nothing to
    save by truncating it.

    Sorted and printed one name per line so two runs' output can be diffed
    directly; the Prometheus API does not promise an order, and the whole list
    on one log line was neither readable nor diffable.

    Args:
        session:        aiohttp client session.
        prometheus_url: Prometheus base URL.

    Returns:
        Sorted metric family names, or an empty list if the fetch failed.
    """
    try:
        async with session.get(
            f"{prometheus_url}/api/v1/label/__name__/values"
        ) as resp:
            label_data = await resp.json()
            all_metrics = sorted(label_data.get("data", []))
            _log_name_list("Prometheus metric families", all_metrics)
            return all_metrics
    except Exception as exc:
        logger.warning("Failed to fetch Prometheus metric names: %s", exc)
        return []


def _selector_metric_name(selector: str) -> str:
    """Strip any label matcher from a contract selector, leaving the name.

    Contract entries are usually bare names but some carry a matcher, e.g.
    ``ledger_economy{metric="base_fee_xrp"}``. Reverse coverage compares family
    names, and Prometheus reports one ``__name__`` per family regardless of how
    many label combinations it has, so the matcher must come off first.

    Args:
        selector: A metric name, optionally followed by a brace matcher group.

    Returns:
        The bare metric family name.
    """
    return selector.split("{", 1)[0].strip()


def _metric_family_candidates(name: str) -> list[str]:
    """The names a contract entry could use to account for an emitted series.

    One histogram instrument reaches Prometheus as three names. Folding the
    ``_bucket``/``_count``/``_sum`` suffix back off gives the base family, so a
    contract entry or a regex written for the family covers all three.

    The unfolded name is always kept as a candidate too, and a name is
    accounted for if *any* candidate matches. That is what makes folding
    ``_count`` safe even though a plain gauge can legitimately end in it
    (``jobq_job_count`` does): the gauge matches on its own full name, and the
    extra ``jobq_job`` candidate can only ever add coverage, never remove it.

    Args:
        name: Emitted Prometheus metric family name.

    Returns:
        ``[name]``, plus the base family if ``name`` carries a histogram suffix.
    """
    candidates = [name]
    for suffix in HISTOGRAM_SUFFIXES:
        if name.endswith(suffix) and len(name) > len(suffix):
            candidates.append(name[: -len(suffix)])
            break
    return candidates


def _accounted_metric_names(
    expected: dict[str, Any],
) -> tuple[set[str], list[re.Pattern[str]]]:
    """Everything expected_metrics.json accounts for, in family-name form.

    Three sources, all of which count as "the contract has considered this
    name" — which is what reverse coverage asks, not "is this name asserted":

    * every selector under a group's ``metrics`` (asserted),
    * every key of ``not_asserted.metrics_excluded`` (deliberately unasserted
      with a written reason),
    * every regex under the top-level ``accounted_patterns`` (bulk families).

    Literal histogram entries are expanded to the whole triple plus the base
    family, so listing only ``foo_bucket`` still accounts for ``foo_count`` and
    ``foo_sum``. The expansion is keyed off ``_bucket``/``_sum`` only, never off
    ``_count`` alone, so a gauge named ``..._count`` does not silently claim a
    shorter base family that nothing emits.

    Args:
        expected: The parsed expected_metrics.json contract.

    Returns:
        A (literal family names, compiled patterns) pair.
    """
    literals: set[str] = set()
    for group in expected.values():
        if not isinstance(group, dict):
            continue
        for selector in group.get("metrics", []):
            literals.add(_selector_metric_name(selector))
        # not_asserted records its entries under metrics_excluded, keyed by name.
        literals.update(group.get("metrics_excluded", {}))

    for name in list(literals):
        for suffix in ("_bucket", "_sum"):
            if name.endswith(suffix) and len(name) > len(suffix):
                base = name[: -len(suffix)]
                literals.add(base)
                literals.update(base + s for s in HISTOGRAM_SUFFIXES)
                break

    patterns = [
        re.compile(entry["pattern"])
        for entry in expected.get("accounted_patterns", [])
        if entry.get("pattern")
    ]
    return literals, patterns


def _unaccounted_metric_names(
    emitted: list[str], expected: dict[str, Any]
) -> list[str]:
    """Metric families Prometheus holds that the contract never mentions.

    Args:
        emitted:  Metric family names from the Prometheus __name__ label.
        expected: The parsed expected_metrics.json contract.

    Returns:
        Sorted list of emitted family names nothing in the contract accounts for.
    """
    literals, patterns = _accounted_metric_names(expected)
    unaccounted = []
    for name in emitted:
        candidates = _metric_family_candidates(name)
        accounted = any(
            candidate in literals
            or any(pattern.fullmatch(candidate) for pattern in patterns)
            for candidate in candidates
        )
        if not accounted:
            unaccounted.append(name)
    return sorted(unaccounted)


def _metric_check_targets(
    expected: dict[str, Any],
) -> tuple[list[tuple[str, str]], list[tuple[str, str, list[str]]]]:
    """Flatten expected_metrics.json into the two lists of check targets.

    Args:
        expected: The parsed expected_metrics.json contract.

    Returns:
        A (metric targets, label targets) pair. Metric targets are
        (group, metric selector) for every name under a group's "metrics".
        Label targets are (group, label, that group's metric selectors) for
        every name under a group's "required_labels" — read for every group
        that declares it, not just spanmetrics. Nothing read the key at all
        until this was added, so the four labels the spanmetrics group
        documented as required had never actually been checked.

    Note:
        A group is any top-level object EXCEPT those in SKIPPED_METRIC_GROUPS.
        The contract also carries a string ("description"), a list
        ("accounted_patterns") and an object that declares no metrics
        ("grafana_dashboards"); the isinstance test skips the first two
        structurally rather than by name, so adding another non-group key cannot
        break this function, and the third contributes nothing because it has no
        "metrics" key.

        The name-based exclusion is still required for one group, so the
        isinstance test alone is NOT equivalent to it. "sync_diagnostics" is an
        object and does declare "metrics", but it has its own validator,
        assert_sync_diagnostics_metrics, which polls and reports the same names.
        Selecting it here as well would poll every metric in the group twice and
        report each check twice, inflating the totals. It is excluded by name
        because the reason is ownership, which no structural test can express.
    """
    groups = [
        (category_key, category_data)
        for category_key, category_data in expected.items()
        if isinstance(category_data, dict) and category_key not in SKIPPED_METRIC_GROUPS
    ]
    targets = [
        (category_key, metric_name)
        for category_key, category_data in groups
        for metric_name in category_data.get("metrics", [])
    ]
    label_targets = [
        (category_key, label, category_data.get("metrics", []))
        for category_key, category_data in groups
        for label in category_data.get("required_labels", [])
    ]
    return targets, label_targets


async def validate_metrics(
    session: aiohttp.ClientSession,
    prometheus_url: str,
    report: ValidationReport,
) -> None:
    """Validate that expected metrics appear in Prometheus with non-zero values.

    Two kinds of check come out of expected_metrics.json: every name under a
    group's "metrics" must have at least one series, and every label under a
    group's "required_labels" must reach at least one of that group's series
    with a non-empty value.

    Args:
        session:        aiohttp client session.
        prometheus_url: Base URL for Prometheus API (e.g., http://localhost:9090).
        report:         ValidationReport to accumulate results.
    """
    logger.info("--- Metric Validation (Prometheus) ---")

    emitted_metric_names = await _log_prometheus_metric_names(session, prometheus_url)

    with open(EXPECTED_METRICS_FILE) as f:
        expected = json.load(f)

    # Flatten the contract, then poll every target concurrently against ONE
    # shared deadline. Polling them serially made each metric own its own
    # timeout, so the waits were additive: 58 metrics x 45 s = 43.5 min, which
    # overran the CI job budget and lost the artifact-upload and summary
    # diagnostics. Sharing the deadline bounds the whole phase to a single
    # poll window.
    #
    # _metric_check_targets applies SKIPPED_METRIC_GROUPS, which is what keeps
    # "sync_diagnostics" out of this walk: it has its own validator,
    # assert_sync_diagnostics_metrics, and letting both walk it would poll and
    # report every metric in the group twice.
    targets, label_targets = _metric_check_targets(expected)

    deadline = time.monotonic() + METRIC_POLL_TIMEOUT_SEC
    sem = asyncio.Semaphore(METRIC_POLL_CONCURRENCY)
    # Both kinds of check share the one deadline and the one concurrency bound,
    # so the label checks cost no extra poll window and add no extra load.
    metric_checks, label_checks = await asyncio.gather(
        asyncio.gather(
            *(
                _check_prometheus_metric(
                    session, prometheus_url, metric_name, category, deadline, sem
                )
                for category, metric_name in targets
            )
        ),
        asyncio.gather(
            *(
                _check_metric_label(
                    session,
                    prometheus_url,
                    category,
                    label,
                    metric_selectors,
                    deadline,
                    sem,
                )
                for category, label, metric_selectors in label_targets
            )
        ),
    )

    # Add in contract order, not completion order, so the report and its log
    # lines stay deterministic across runs. Existence checks keep their place
    # ahead of the label checks, so no existing check's position moves.
    for check in [*metric_checks, *label_checks]:
        report.add(check)

    # Reverse direction: metric families Prometheus holds that the contract
    # never mentions. Added last so no existing check's position moves.
    unaccounted_metrics = _unaccounted_metric_names(emitted_metric_names, expected)
    if unaccounted_metrics:
        _log_name_list(
            "Metric families emitted but NOT accounted for by expected_metrics.json "
            "(warning only, does not fail CI)",
            unaccounted_metrics,
        )
    report.add(
        _reverse_coverage_result(
            check_name="metric.reverse_coverage",
            category="metric",
            noun="metric families",
            emitted_count=len(emitted_metric_names),
            unaccounted=unaccounted_metrics,
        )
    )


def _selector_with_label(metric_selector: str, label: str) -> str:
    """Add a "label is present and non-empty" matcher to a metric selector.

    ``<label>!=""`` is the only matcher that expresses the requirement. An
    absent Prometheus label is indistinguishable from an empty-string one, so
    ``<label>=~".*"`` matches series that never carried the label at all and a
    check written that way would pass on a node that lost the attribute
    entirely. ``!=""`` rejects both the absent and the blank case.

    Selectors in expected_metrics.json are usually bare names, but some already
    carry a matcher (``ledger_economy{metric="base_fee_xrp"}``), so the matcher
    is merged into an existing brace group rather than appended after it.

    Args:
        metric_selector: A metric name, optionally with a brace matcher group.
        label:           Prometheus label name that must be present, non-empty.

    Returns:
        The selector with the label matcher merged in.
    """
    matcher = label + '!=""'
    if metric_selector.endswith("}"):
        head = metric_selector[:-1].rstrip()
        separator = "" if head.endswith("{") else ", "
        return head + separator + matcher + "}"
    return metric_selector + "{" + matcher + "}"


async def _poll_series_count(
    session: aiohttp.ClientSession,
    prometheus_url: str,
    selectors: list[str],
    deadline: float,
    sem: asyncio.Semaphore,
) -> int:
    """Poll Prometheus until a selector has series or the deadline passes.

    Uses the /api/v1/series endpoint instead of an instant query.
    Beast::insight StatsD gauges only mark dirty on value *changes*, so a gauge
    that stabilizes (e.g. peer count stays at 1) may go stale in Prometheus and
    disappear from instant queries.  The series endpoint returns any metric
    that existed in the window, regardless of staleness.

    Polls rather than querying once: late-populating gauges/counters may not
    have completed the export+scrape pipeline when this runs, so a single query
    races. A metric that never appears still fails once the deadline passes.

    Args:
        session:        aiohttp client session.
        prometheus_url: Prometheus base URL.
        selectors:      One or more Prometheus selectors. /api/v1/series takes
                        repeated match[] parameters and unions their results,
                        so several selectors answer "does ANY of these have a
                        series" in a single request.
        deadline:       Monotonic deadline shared by every metric in the run.
        sem:            Bounds how many requests reach Prometheus at once. It
                        is held only across the request, never across the
                        sleep, so one absent metric cannot starve the others.

    Returns:
        Number of series found, or 0 if nothing ever appeared.
    """
    params: list[tuple[str, str]] = [("match[]", s) for s in selectors]
    while True:
        async with sem:
            async with session.get(
                f"{prometheus_url}/api/v1/series", params=params
            ) as resp:
                data = await resp.json()
                series_count = len(data.get("data", []))
        if series_count > 0 or time.monotonic() >= deadline:
            return series_count
        # Never sleep past the shared deadline.
        await asyncio.sleep(min(METRIC_POLL_INTERVAL_SEC, deadline - time.monotonic()))


async def _check_prometheus_metric(
    session: aiohttp.ClientSession,
    prometheus_url: str,
    metric_name: str,
    category: str,
    deadline: float,
    sem: asyncio.Semaphore,
) -> CheckResult:
    """Query Prometheus for a specific metric and check it exists.

    Args:
        session:        aiohttp client session.
        prometheus_url: Prometheus base URL.
        metric_name:    Prometheus metric name.
        category:       Metric category for the report.
        deadline:       Monotonic deadline shared by every metric in the run.
        sem:            Bounds how many requests reach Prometheus at once.

    Returns:
        The CheckResult for this metric. The caller adds it to the report so
        report order follows the contract file rather than completion order.
    """
    try:
        series_count = await _poll_series_count(
            session, prometheus_url, [metric_name], deadline, sem
        )
        return CheckResult(
            name=f"metric.{category}.{metric_name}",
            category="metric",
            passed=series_count > 0,
            message=(
                f"{metric_name}: {series_count} series"
                if series_count > 0
                else f"{metric_name}: 0 series (expected > 0)"
            ),
            details={"series_count": series_count},
        )
    except Exception as exc:
        return CheckResult(
            name=f"metric.{category}.{metric_name}",
            category="metric",
            passed=False,
            message=f"{metric_name}: query failed ({exc})",
        )


async def assert_sync_diagnostics_metrics(
    session: aiohttp.ClientSession,
    prometheus_url: str,
    report: ValidationReport,
) -> None:
    """Assert every metric in the 'sync_diagnostics' group is present.

    Fresh-node sync-diagnostics work packages append native metric names to the
    "sync_diagnostics" group in expected_metrics.json; this check fails the run
    if any listed metric regresses to absent, so a dropped signal cannot pass
    CI silently. An empty group is a genuine no-op: nothing is queried and no
    check is recorded, which is the state before any signal has landed.

    A missing group key is itself a failure — the key is the anchor the sync
    work packages append to, so its absence means the harness lost the contract
    rather than that there is nothing to check.

    Args:
        session:        aiohttp client session.
        prometheus_url: Prometheus API base URL.
        report:         ValidationReport to accumulate results.
    """
    logger.info("--- Fresh-Node Sync Diagnostics Metrics ---")

    with open(EXPECTED_METRICS_FILE) as f:
        expected = json.load(f)

    group = expected.get(SYNC_DIAGNOSTICS_GROUP)
    if group is None:
        report.add(
            CheckResult(
                name=f"metric.{SYNC_DIAGNOSTICS_GROUP}.group_present",
                category="metric",
                passed=False,
                message=(
                    f"'{SYNC_DIAGNOSTICS_GROUP}' group missing from "
                    f"{EXPECTED_METRICS_FILE.name}"
                ),
            )
        )
        return

    metrics = group.get("metrics", [])
    if not metrics:
        logger.info(
            "[SKIP] metric.%s: group is empty (no sync-diagnostics signals "
            "declared yet)",
            SYNC_DIAGNOSTICS_GROUP,
        )
        return

    # Same fan-out shape as validate_metrics(): ONE shared deadline for the
    # whole group, so the phase costs a single poll window instead of one per
    # metric, and a semaphore so the fan-out cannot exhaust the Prometheus
    # connection pool.
    #
    # _check_prometheus_metric RETURNS its CheckResult rather than recording it,
    # so each result must be added here. Discarding them would let the group
    # pass silently whatever Prometheus holds, which is the more dangerous half
    # of the defect this replaces: the original call also passed `report` where
    # `deadline` belongs and omitted `sem` entirely, raising TypeError before any
    # check ran and taking the validation phases ordered after it down with it --
    # three of them as CI invokes this (dashboards, parity span attrs, parity
    # value sanity), since the workflow passes --skip-loki; four when the
    # log-correlation phase is enabled.
    deadline = time.monotonic() + METRIC_POLL_TIMEOUT_SEC
    sem = asyncio.Semaphore(METRIC_POLL_CONCURRENCY)
    checks = await asyncio.gather(
        *(
            _check_prometheus_metric(
                session,
                prometheus_url,
                metric_name,
                SYNC_DIAGNOSTICS_GROUP,
                deadline,
                sem,
            )
            for metric_name in metrics
        )
    )
    for check in checks:
        report.add(check)


async def _check_metric_label(
    session: aiohttp.ClientSession,
    prometheus_url: str,
    category: str,
    label: str,
    metric_selectors: list[str],
    deadline: float,
    sem: asyncio.Semaphore,
) -> CheckResult:
    """Check that a group's required label reaches Prometheus non-empty.

    A "required_labels" entry names a label that at least one series of that
    group must carry with a non-empty value. It is checked per group rather
    than per metric because the requirement is about the label reaching
    Prometheus at all, and one series carrying it proves the pipeline works.

    The failure this guards against is a label going missing while every metric
    in the group still exists, so the existence checks stay green. xrpl_node_id
    is the case that motivated it: losing it makes Grafana Cloud trace ingest
    fold distinct nodes into one.

    Args:
        session:          aiohttp client session.
        prometheus_url:   Prometheus base URL.
        category:         Group key from expected_metrics.json.
        label:            Label name that must be present and non-empty.
        metric_selectors: The group's asserted metric selectors.
        deadline:         Monotonic deadline shared by every metric in the run.
        sem:              Bounds how many requests reach Prometheus at once.

    Returns:
        The CheckResult for this (group, label) pair. The caller adds it to the
        report so report order follows the contract file.
    """
    check_name = f"metric.{category}.label.{label}"

    if not metric_selectors:
        # required_labels on a group with no asserted metrics would be a check
        # that silently never runs — the exact failure mode it exists to catch.
        return CheckResult(
            name=check_name,
            category="metric",
            passed=False,
            message=(
                f"{label}: group '{category}' declares required_labels but no "
                "metrics, so there is nothing to check the label against"
            ),
        )

    selectors = [_selector_with_label(m, label) for m in metric_selectors]
    try:
        series_count = await _poll_series_count(
            session, prometheus_url, selectors, deadline, sem
        )
        return CheckResult(
            name=check_name,
            category="metric",
            passed=series_count > 0,
            message=(
                f"{label}: present and non-empty on {series_count} "
                f"{category} series"
                if series_count > 0
                else f"{label}: no {category} series carries it with a "
                "non-empty value (an absent label and an empty one are the "
                "same thing in Prometheus, so this covers both). If this "
                "group's own metric checks also failed, the metrics are "
                "missing rather than the label."
            ),
            details={"series_count": series_count, "selectors": selectors},
        )
    except Exception as exc:
        return CheckResult(
            name=check_name,
            category="metric",
            passed=False,
            message=f"{label}: query failed ({exc})",
        )


# ---------------------------------------------------------------------------
# Log-Trace Correlation Validation (Loki API)
# ---------------------------------------------------------------------------


async def _loki_json(
    session: aiohttp.ClientSession,
    loki_url: str,
    path: str,
    params: dict[str, str] | None = None,
) -> dict[str, Any] | None:
    """GET one Loki API path and decode it, or None if that failed.

    Diagnostic-only helper: every failure is a warning, so one unreachable
    endpoint does not suppress the rest of the diagnostic.

    Args:
        session:  aiohttp client session.
        loki_url: Loki API base URL.
        path:     API path, including its leading slash.
        params:   Optional query parameters.

    Returns:
        The decoded response body, or None when the request or decode failed.
    """
    try:
        async with session.get(f"{loki_url}{path}", params=params or {}) as resp:
            if resp.status != 200:
                # Loki reports a rejected query as text/plain, so resp.json()
                # would raise ContentTypeError and the log would carry a
                # mimetype complaint instead of Loki's own explanation. Read
                # the body and print it: a diagnostic that swallows the
                # server's reason is worse than no diagnostic.
                body = " ".join((await resp.text()).split())
                logger.warning(
                    "Loki diagnostic: GET %s returned HTTP %d: %s",
                    path,
                    resp.status,
                    body[:500] or "(empty body)",
                )
                return None
            return await resp.json()
    except Exception as exc:  # noqa: BLE001 - diagnostic must never raise
        logger.warning("Loki diagnostic: GET %s failed: %s", path, exc)
        return None


async def _loki_count(
    session: aiohttp.ClientSession, loki_url: str, query: str
) -> int | None:
    """Sum an instant LogQL count over every stream it returns.

    Returns None rather than 0 when Loki could not be queried. "Loki holds
    none" and "Loki did not answer" are different findings, and printing the
    same number for both is how a diagnostic misleads.

    Args:
        session:  aiohttp client session.
        loki_url: Loki API base URL.
        query:    An instant LogQL metric query.

    Returns:
        Total entry count, or None when the query could not be evaluated.
    """
    params = {"query": query, "time": str(int(time.time()))}
    data = await _loki_json(session, loki_url, "/loki/api/v1/query", params)
    if data is None:
        return None
    try:
        results = data.get("data", {}).get("result", [])
        return sum(int(float(entry["value"][1])) for entry in results)
    except Exception as exc:  # noqa: BLE001 - diagnostic must never raise
        logger.warning("Loki diagnostic: could not read count for %s: %s", query, exc)
        return None


async def _log_loki_diagnostics(session: aiohttp.ClientSession, loki_url: str) -> None:
    """Log what Loki holds for the stream the correlation checks select.

    Same role as _log_prometheus_metric_names on the metric side, and the same
    contract: diagnostic only, every failure a warning, never a check result.
    A failed correlation check says a log line was not found; these numbers say
    which of the three Loki-side reasons applies — nothing was ingested at all,
    entries exist but none carry a trace id, or entries arrived under a label
    set this selector cannot match (which the label inventory exposes).

    Counted over exactly LOG_QUERY_WINDOW_SECONDS, the window the checks use.
    A wider window here would report entries the checks cannot see, including a
    previous run's, and describe a failure that is not the one at hand.

    Args:
        session:  aiohttp client session.
        loki_url: Loki API base URL.
    """
    window = f"[{LOG_QUERY_WINDOW_SECONDS}s]"
    logger.info("Loki diagnostic: the checks below query %s", LOG_CORRELATION_QUERY)

    labels = await _loki_json(session, loki_url, "/loki/api/v1/labels")
    if labels is not None:
        names = labels.get("data") or []
        logger.info(
            "Loki diagnostic: stream labels Loki knows: %s",
            ", ".join(names) or "(none)",
        )

    values = await _loki_json(
        session, loki_url, "/loki/api/v1/label/service_name/values"
    )
    if values is not None:
        found = values.get("data") or []
        logger.info(
            "Loki diagnostic: service_name values: %s", ", ".join(found) or "(none)"
        )

    # sum() is load-bearing, not cosmetic. The file_log receiver's regex_parser
    # leaves message, timestamp, trace_id and span_id as log-record attributes,
    # and Loki's OTLP path stores those as structured metadata, which joins the
    # label set of a metric query. Because `message` and `timestamp` are unique
    # per line, an unaggregated count_over_time returns ONE SERIES PER LOG LINE
    # and blows past limits_config.max_query_series (500) at a few hundred
    # lines: Loki answers HTTP 400 "maximum number of series (500) reached for
    # a single query", both legs report "unavailable", and the diagnostic loses
    # the one distinction it exists to draw. Aggregating collapses the result to
    # a single series regardless of line count. An empty result still decodes to
    # 0, so the deliberate "0 entries" versus "could not answer" split above is
    # preserved.
    for label, query in (
        ("selector only", f"sum(count_over_time({LOG_STREAM_SELECTOR}{window}))"),
        (
            "selector + line filter",
            f"sum(count_over_time({LOG_CORRELATION_QUERY} {window}))",
        ),
    ):
        total = await _loki_count(session, loki_url, query)
        logger.info(
            "Loki diagnostic: %s = %s entries in the last %ds  [%s]",
            label,
            "unavailable" if total is None else total,
            LOG_QUERY_WINDOW_SECONDS,
            query,
        )


async def _poll_tempo_for_logged_ids(
    session: aiohttp.ClientSession,
    tempo_url: str,
    unique_ids: list[str],
) -> tuple[str | None, int, int, str | None]:
    """Resolve one of ``unique_ids`` in Tempo, retrying until a deadline.

    Every id is tried on each pass, not just the first: one unexported trace
    must not fail the check while correlation demonstrably works for another.

    The pass is then repeated until the shared poll window closes, because a
    trace id reaches a log line the instant its span is created but only becomes
    queryable after the exporter batches it, Tempo ingests it and its block is
    searchable. A single pass therefore races that pipeline exactly the way the
    metric checks used to race the export-plus-scrape pipeline, and it is fixed
    the same way -- with METRIC_POLL_TIMEOUT_SEC and METRIC_POLL_INTERVAL_SEC,
    the constants that already bound every other poll in this file, rather than
    a second timeout of its own.

    Retrying does not weaken the assertion. The check still passes only on a
    logged id that Tempo really returns spans for; what changes is that a
    failure now means the id was absent for the whole window, which is a
    confirmed absence rather than a snapshot taken at one moment.

    A Tempo request that FAILS is not the same as a trace that is absent, and
    the two are reported differently. A 404 is absence and returns no spans, so
    the next candidate is tried. Any other failure is remembered and polling
    continues -- a transient blip mid-window should not fail a check that would
    otherwise pass -- but if the window closes with nothing resolved, the last
    failure is returned so the caller can say the query broke instead of
    claiming the spans were never exported. Swallowing it would reproduce the
    silent failure this change exists to remove.

    Args:
        session:    aiohttp client session.
        tempo_url:  Base URL for the Tempo API.
        unique_ids: Candidate trace ids read from Loki, already deduplicated.

    Returns:
        ``(trace_id, span_count, attempts, error)``. ``error`` is None unless a
        Tempo request failed and nothing resolved; ``(None, 0, n, None)`` means
        every id was genuinely absent for the whole window.
    """
    deadline = time.monotonic() + METRIC_POLL_TIMEOUT_SEC
    attempts = 0
    last_error: str | None = None
    while True:
        attempts += 1
        for candidate in unique_ids:
            try:
                spans = await _tempo_get_trace(session, tempo_url, candidate)
            except Exception as exc:  # noqa: BLE001 - a poll must not abort here
                last_error = f"{type(exc).__name__}: {exc}"
                continue
            if spans:
                # Resolved, so whatever failed earlier in the window did not
                # matter -- do not report a stale error alongside a pass.
                return candidate, len(spans), attempts, None
        remaining = deadline - time.monotonic()
        if remaining <= 0:
            return None, 0, attempts, last_error
        # Never sleep past the deadline, matching _poll_series_count.
        await asyncio.sleep(min(METRIC_POLL_INTERVAL_SEC, remaining))


async def validate_log_trace_correlation(
    session: aiohttp.ClientSession,
    loki_url: str,
    tempo_url: str,
    report: ValidationReport,
) -> None:
    """Validate that Loki logs contain trace_id/span_id for correlation.

    Checks:
      1. Logs with trace_id= field exist in Loki.
      2. A random trace_id from Tempo can be found in Loki logs.

    Args:
        session:   aiohttp client session.
        loki_url:  Base URL for Loki API (e.g., http://localhost:3100).
        tempo_url: Base URL for Tempo API.
        report:    ValidationReport to accumulate results.
    """
    logger.info("--- Log-Trace Correlation Validation (Loki) ---")

    # Numbers first, so a failure below is read next to the state that caused
    # it. Never raises and never adds a check (see _log_loki_diagnostics).
    await _log_loki_diagnostics(session, loki_url)

    # Check 1: Any logs with trace_id exist.
    try:
        params = {
            "query": LOG_CORRELATION_QUERY,
            "limit": 5,
            "direction": "backward",
            **_log_query_window(),
        }
        async with session.get(
            f"{loki_url}/loki/api/v1/query_range", params=params
        ) as resp:
            data = await resp.json()
            streams = data.get("data", {}).get("result", [])
            total_entries = sum(len(s.get("values", [])) for s in streams)
            report.add(
                CheckResult(
                    name="log.trace_id_present",
                    category="log",
                    passed=total_entries > 0,
                    message=(
                        f"Found {total_entries} log entries with trace_id"
                        if total_entries > 0
                        else "No log entries with trace_id found"
                    ),
                    details={"log_count": total_entries},
                )
            )
    except Exception as exc:
        report.add(
            CheckResult(
                name="log.trace_id_present",
                category="log",
                passed=False,
                message=f"Loki query failed: {exc}",
            )
        )

    # Check 2: Cross-reference a trace_id from a log line back to Tempo.
    #
    # Driven from the log side on purpose. A trace_id only reaches a log line
    # when that line is emitted inside a sampled span, and at `warning` level
    # most spans produce no log output at all — so picking an arbitrary trace
    # from Tempo and expecting it in Loki fails even when correlation works.
    # Starting from a logged trace_id tests the invariant that matters: an id
    # written to a log must resolve to a trace that was actually exported.
    try:
        loki_params = {
            "query": LOG_CORRELATION_QUERY,
            "limit": 5,
            "direction": "backward",
            **_log_query_window(),
        }
        async with session.get(
            f"{loki_url}/loki/api/v1/query_range", params=loki_params
        ) as resp:
            data = await resp.json()
            streams = data.get("data", {}).get("result", [])

        logged_ids = [
            match.group(1)
            for stream in streams
            for _, line in stream.get("values", [])
            if (match := re.search(r"trace_id=([0-9a-f]{32})", line))
        ]

        if not logged_ids:
            report.add(
                CheckResult(
                    name="log.trace_id_cross_reference",
                    category="log",
                    passed=False,
                    message=(
                        "No logged trace_id to cross-reference. Log lines carry one only "
                        "when emitted inside a sampled span; raise the log level or widen "
                        "the workload if this persists."
                    ),
                )
            )
        else:
            unique_ids = list(dict.fromkeys(logged_ids))
            resolved, span_count, attempts, poll_error = (
                await _poll_tempo_for_logged_ids(session, tempo_url, unique_ids)
            )
            if resolved:
                message = (
                    f"logged trace_id {resolved[:16]}... resolves to "
                    f"{span_count} spans in Tempo"
                )
            elif poll_error:
                # Distinct from the absence message below on purpose: "Tempo
                # never answered" and "the spans were not exported" call for
                # different investigations, and printing the second for the
                # first sends the reader to the wrong subsystem.
                message = (
                    f"could not verify {len(unique_ids)} logged trace_id(s): "
                    f"Tempo queries failed over {METRIC_POLL_TIMEOUT_SEC:g}s "
                    f"({attempts} attempt(s)); last error was {poll_error}"
                )
            else:
                message = (
                    f"none of {len(unique_ids)} logged trace_id(s) resolve in "
                    f"Tempo after {attempts} attempt(s) over "
                    f"{METRIC_POLL_TIMEOUT_SEC:g}s; the spans they name were "
                    "not exported"
                )
            report.add(
                CheckResult(
                    name="log.trace_id_cross_reference",
                    category="log",
                    passed=resolved is not None,
                    message=message,
                    details={
                        "trace_id": resolved,
                        "span_count": span_count,
                        "candidates": len(unique_ids),
                        "poll_attempts": attempts,
                        "poll_timeout_sec": METRIC_POLL_TIMEOUT_SEC,
                        "poll_error": poll_error,
                    },
                )
            )
    except Exception as exc:
        report.add(
            CheckResult(
                name="log.trace_id_cross_reference",
                category="log",
                passed=False,
                message=f"Cross-reference check failed: {exc}",
            )
        )


# ---------------------------------------------------------------------------
# Dashboard Validation (Grafana API)
# ---------------------------------------------------------------------------


def _leaf_panel_count(dashboard: dict[str, Any]) -> int:
    """Count the panels a dashboard actually renders.

    Grafana models a row as an entry of ``type: "row"`` in the top-level
    ``panels`` list, and a collapsed row carries its children in its own
    nested ``panels`` list. So ``len(dashboard["panels"])`` counts rows as
    though they were panels and misses everything inside a collapsed one --
    on ``node-health`` that reads 55 where the true figure is 51, and a
    dashboard consisting only of collapsed rows would report a positive
    count while rendering nothing.

    Args:
        dashboard: The ``dashboard`` object from the Grafana API response.

    Returns:
        The number of non-row panels, including those nested inside rows.
    """
    total = 0
    for panel in dashboard.get("panels", []):
        if panel.get("type") == "row":
            total += len(panel.get("panels", []))
        else:
            total += 1
    return total


async def validate_dashboards(
    session: aiohttp.ClientSession,
    grafana_url: str,
    report: ValidationReport,
) -> None:
    """Validate that all Grafana dashboards are accessible and return data.

    For each expected dashboard UID, queries the Grafana API to verify
    the dashboard exists and is loadable.

    Args:
        session:     aiohttp client session.
        grafana_url: Base URL for Grafana API (e.g., http://localhost:3000).
        report:      ValidationReport to accumulate results.
    """
    logger.info("--- Dashboard Validation (Grafana) ---")

    with open(EXPECTED_METRICS_FILE) as f:
        expected = json.load(f)

    dashboard_uids = expected.get("grafana_dashboards", {}).get("uids", [])

    for uid in dashboard_uids:
        try:
            async with session.get(f"{grafana_url}/api/dashboards/uid/{uid}") as resp:
                if resp.status == 200:
                    data = await resp.json()
                    dashboard = data.get("dashboard", {})
                    panel_count = _leaf_panel_count(dashboard)
                    report.add(
                        CheckResult(
                            name=f"dashboard.{uid}",
                            category="dashboard",
                            passed=panel_count > 0,
                            message=(
                                f"{uid}: loaded ({panel_count} panels)"
                                if panel_count
                                else f"{uid}: loaded but renders no panels"
                            ),
                            details={"panel_count": panel_count},
                        )
                    )
                else:
                    report.add(
                        CheckResult(
                            name=f"dashboard.{uid}",
                            category="dashboard",
                            passed=False,
                            message=f"{uid}: HTTP {resp.status}",
                        )
                    )
        except Exception as exc:
            report.add(
                CheckResult(
                    name=f"dashboard.{uid}",
                    category="dashboard",
                    passed=False,
                    message=f"{uid}: query failed ({exc})",
                )
            )


# ---------------------------------------------------------------------------
# Span duration validation
# ---------------------------------------------------------------------------


async def validate_span_durations(
    session: aiohttp.ClientSession,
    tempo_url: str,
    report: ValidationReport,
) -> None:
    """Validate that span durations are within reasonable bounds.

    Checks that spans have duration > 0 and < 60s, flagging any anomalies.

    Args:
        session:   aiohttp client session.
        tempo_url: Base URL for Tempo API.
        report:    ValidationReport to accumulate results.
    """
    logger.info("--- Span Duration Validation ---")

    try:
        traces = await _tempo_search(
            session,
            tempo_url,
            '{resource.service.name="xrpld"}',
            limit=5,
        )

        if not traces:
            report.add(
                CheckResult(
                    name="span.duration_bounds",
                    category="span",
                    passed=False,
                    message="No traces available for duration check",
                )
            )
            return

        total_spans = 0
        invalid_spans = 0
        max_duration_ns = 0

        for trace_summary in traces:
            trace_id = trace_summary.get("traceID", "")
            if not trace_id:
                continue
            spans = await _tempo_get_trace(session, tempo_url, trace_id)
            for span in spans:
                start_ns = int(span.get("startTimeUnixNano", "0"))
                end_ns = int(span.get("endTimeUnixNano", "0"))
                duration_ns = end_ns - start_ns
                total_spans += 1
                max_duration_ns = max(max_duration_ns, duration_ns)
                # Invalid if negative or > 60 seconds.
                if duration_ns < 0 or duration_ns > 60_000_000_000:
                    invalid_spans += 1

        max_duration_ms = max_duration_ns / 1_000_000

        report.add(
            CheckResult(
                name="span.duration_bounds",
                category="span",
                passed=invalid_spans == 0,
                message=(
                    f"All {total_spans} spans have valid durations "
                    f"(max: {max_duration_ms:.1f}ms)"
                    if invalid_spans == 0
                    else f"{invalid_spans}/{total_spans} spans have invalid "
                    "durations (<0 or >60s)"
                ),
                details={
                    "total_spans": total_spans,
                    "invalid_spans": invalid_spans,
                    "max_duration_ms": round(max_duration_ms, 2),
                },
            )
        )
    except Exception as exc:
        report.add(
            CheckResult(
                name="span.duration_bounds",
                category="span",
                passed=False,
                message=f"Duration check failed: {exc}",
            )
        )


# ---------------------------------------------------------------------------
# External Dashboard Parity Validation
# ---------------------------------------------------------------------------

# Span attributes that external dashboards (validator-health, peer-quality,
# node-health) depend on.  Each entry maps a span name to the
# attributes that must be present for external dashboard panels to render.
# Keys follow the 2026-05-13 span-attr naming redesign (bare/underscore form;
# dotted xrpl.* reserved for resource attributes). The amendment_blocked,
# server_state, and proposers_validated values that earlier external-dashboard
# work tracked are NOT span attributes — they exist only as MetricsRegistry
# metrics (validator_health{metric="amendment_blocked"},
# state_tracking{metric="state_value"}, etc.), so they are validated by
# PARITY_VALUE_SANITY below rather than as span attributes here.
PARITY_SPAN_ATTRS: list[dict[str, str]] = [
    {"span": "tx.receive", "attr": "peer_version"},
    {"span": "consensus.validation.send", "attr": "ledger_hash"},
    {"span": "consensus.validation.send", "attr": "full_validation"},
    # peer.validation.receive shares the ledger_hash / full_validation keys with
    # consensus.validation.send (same keys, told apart by span name).
    {"span": "peer.validation.receive", "attr": "ledger_hash"},
    {"span": "peer.validation.receive", "attr": "full_validation"},
    {"span": "consensus.accept", "attr": "quorum"},
]

# Value sanity bounds for external-parity metrics.  Each entry specifies a
# Prometheus query and the acceptable range [lo, hi] for the returned value.
PARITY_VALUE_SANITY: list[dict[str, Any]] = [
    {
        "name": "validation_agreement_pct_1h",
        "query": 'validation_agreement{metric="agreement_pct_1h"}',
        "lo": 0,
        "hi": 100,
    },
    {
        "name": "unl_expiry_days",
        "query": 'validator_health{metric="unl_expiry_days"}',
        # Four reading classes are all legitimate: days remaining, a negative
        # count once the list has expired, -1 when no published list has been
        # fetched, and +inf for a config-listed list that never expires. Only a
        # reading past a century is nonsense, and that means a broken clock.
        # The wrap this floor used to hide is caught deterministically by the
        # MetricsRegistry::daysUntil unit tests instead.
        "lo": -36500,
        "hi": None,
    },
    {
        "name": "peer_latency_p90_ms",
        "query": 'peer_quality{metric="peer_latency_p90_ms"}',
        "lo": 0,
        "hi": None,
    },
    {
        "name": "state_value",
        "query": 'state_tracking{metric="state_value"}',
        "lo": 0,
        "hi": 7,
    },
]


async def validate_parity_span_attrs(
    session: aiohttp.ClientSession,
    tempo_url: str,
    report: ValidationReport,
) -> None:
    """Validate span attributes required by external dashboard panels.

    For each (span, attribute) pair in PARITY_SPAN_ATTRS, queries Tempo
    for the span and checks that the attribute key exists on at least one
    span in the returned traces.

    Args:
        session:   aiohttp client session.
        tempo_url: Base URL for Tempo API.
        report:    ValidationReport to accumulate results.
    """
    logger.info("--- External Parity: Span Attribute Checks ---")

    for entry in PARITY_SPAN_ATTRS:
        span_name = entry["span"]
        attr_name = entry["attr"]
        check_name = f"parity.span_attr.{span_name}.{attr_name}"

        try:
            query = '{resource.service.name="xrpld" && name="' + span_name + '"}'
            traces = await _tempo_search(session, tempo_url, query, limit=5)

            if not traces:
                report.add(
                    CheckResult(
                        name=check_name,
                        category="parity",
                        passed=False,
                        message=(
                            f"{span_name}: no traces found, "
                            f"cannot verify attr {attr_name}"
                        ),
                    )
                )
                continue

            # Fetch full trace and search spans for the attribute.
            found = False
            for trace_summary in traces:
                trace_id = trace_summary.get("traceID", "")
                if not trace_id:
                    continue
                spans = await _tempo_get_trace(session, tempo_url, trace_id)
                for span in spans:
                    if attr_name in _otlp_span_attr_keys(span):
                        found = True
                        break
                if found:
                    break

            report.add(
                CheckResult(
                    name=check_name,
                    category="parity",
                    passed=found,
                    message=(
                        f"{span_name}: attribute '{attr_name}' present"
                        if found
                        else f"{span_name}: attribute '{attr_name}' missing"
                    ),
                )
            )
        except Exception as exc:
            report.add(
                CheckResult(
                    name=check_name,
                    category="parity",
                    passed=False,
                    message=f"{span_name}: attr check failed ({exc})",
                )
            )


def _series_label(series: dict[str, Any]) -> str:
    """Name a Prometheus series for use in a failure message.

    Args:
        series: One entry from a Prometheus query result.

    Returns:
        The series' service_instance_id when it carries one (the label that
        tells harness cluster nodes apart), else its full label set.
    """
    metric = series.get("metric", {})
    instance = metric.get("service_instance_id")
    if instance:
        return f"service_instance_id={instance}"
    return str(metric) if metric else "<unlabelled series>"


def _value_in_bounds(
    value: float, lo: float, hi: float | None, exclusive_lo: bool
) -> bool:
    """Test one sample against a sanity range.

    Args:
        value:        Sample value.
        lo:           Lower bound.
        hi:           Upper bound, or None when unbounded above.
        exclusive_lo: True when the lower bound is exclusive.

    Returns:
        True when the value is inside the range.
    """
    lo_ok = value > lo if exclusive_lo else value >= lo
    return lo_ok and (hi is None or value <= hi)


def _bounds_description(lo: float, hi: float | None, exclusive_lo: bool) -> str:
    """Build the human-readable bound text used in check messages.

    Args:
        lo:           Lower bound.
        hi:           Upper bound, or None when unbounded above.
        exclusive_lo: True when the lower bound is exclusive.

    Returns:
        A phrase such as "> 0 and <= 100".
    """
    desc = f"{'>' if exclusive_lo else '>='} {lo}"
    if hi is not None:
        desc += f" and <= {hi}"
    return desc


async def _poll_instant_query(
    session: aiohttp.ClientSession,
    prometheus_url: str,
    query: str,
    deadline: float,
) -> list[dict[str, Any]]:
    """Run an instant query, retrying until it returns series or time runs out.

    A bounds check needs the sample value, so it cannot use the /api/v1/series
    endpoint the metric checks poll. An instant query answers from the last
    scrape, and a gauge that stops changing can fall out of it, so one attempt
    is not enough to call the series absent.

    Args:
        session:        aiohttp client session.
        prometheus_url: Prometheus API base URL.
        query:          PromQL instant query.
        deadline:       Monotonic deadline. Never slept past.

    Returns:
        The result list, empty if nothing appeared before the deadline.
    """
    while True:
        async with session.get(
            f"{prometheus_url}/api/v1/query", params={"query": query}
        ) as resp:
            data = await resp.json()
            # An error is not "not yet": a bad query never becomes good, so
            # retrying it only burns the whole deadline. Raise instead, and let
            # the caller report it against the check's own name.
            if data.get("status") != "success":
                raise RuntimeError(
                    "Prometheus rejected the query: "
                    f"{data.get('error') or data.get('status')}"
                )
            results = data.get("data", {}).get("result", [])
        remaining = deadline - time.monotonic()
        if results or remaining <= 0:
            return results
        await asyncio.sleep(min(METRIC_POLL_INTERVAL_SEC, remaining))


async def _check_parity_value(
    session: aiohttp.ClientSession,
    prometheus_url: str,
    entry: dict[str, Any],
) -> CheckResult:
    """Bounds-check every series returned by one parity sanity query.

    Args:
        session:        aiohttp client session.
        prometheus_url: Prometheus API base URL.
        entry:          One PARITY_VALUE_SANITY entry.

    Returns:
        A CheckResult that fails if any series is out of bounds, naming each
        offending series.
    """
    name = entry["name"]
    lo = entry["lo"]
    hi = entry["hi"]
    exclusive_lo = entry.get("exclusive_lo", False)
    check_name = f"parity.value_sanity.{name}"

    try:
        deadline = time.monotonic() + METRIC_POLL_TIMEOUT_SEC
        results = await _poll_instant_query(
            session, prometheus_url, entry["query"], deadline
        )

        if not results:
            return CheckResult(
                name=check_name,
                category="parity",
                passed=False,
                message=(
                    f"{name}: no data returned from Prometheus after "
                    f"{METRIC_POLL_TIMEOUT_SEC:g}s"
                ),
            )

        values: list[float] = []
        offenders: list[str] = []
        for series in results:
            value = float(series["value"][1])
            values.append(value)
            if not _value_in_bounds(value, lo, hi, exclusive_lo):
                offenders.append(f"{_series_label(series)} value {value}")

        bound_desc = _bounds_description(lo, hi, exclusive_lo)
        return CheckResult(
            name=check_name,
            category="parity",
            passed=not offenders,
            message=(
                f"{name}: all {len(values)} series within bounds ({bound_desc})"
                if not offenders
                else f"{name}: {len(offenders)} of {len(values)} series out of "
                f"bounds (expected {bound_desc}): " + "; ".join(offenders)
            ),
            details={
                "values": values,
                "series_count": len(values),
                "out_of_bounds": offenders,
                "lo": lo,
                "hi": hi,
            },
        )
    except Exception as exc:
        return CheckResult(
            name=check_name,
            category="parity",
            passed=False,
            message=f"{name}: sanity check failed ({exc})",
        )


async def validate_parity_value_sanity(
    session: aiohttp.ClientSession,
    prometheus_url: str,
    report: ValidationReport,
) -> None:
    """Validate that external-parity metric values fall within sane bounds.

    For each entry in PARITY_VALUE_SANITY, queries Prometheus and checks
    *every* returned series against the specified [lo, hi] range. These
    queries are bare selectors with no aggregation, so a multi-node harness
    cluster returns one series per service_instance_id; checking only the
    first would let an out-of-range node pass silently.

    Args:
        session:        aiohttp client session.
        prometheus_url: Prometheus API base URL.
        report:         ValidationReport to accumulate results.
    """
    logger.info("--- External Parity: Value Sanity Checks ---")

    for entry in PARITY_VALUE_SANITY:
        report.add(await _check_parity_value(session, prometheus_url, entry))


# ---------------------------------------------------------------------------
# Main validation orchestrator
# ---------------------------------------------------------------------------


async def run_validation(
    tempo_url: str,
    prometheus_url: str,
    loki_url: str,
    grafana_url: str,
    skip_loki: bool = False,
) -> ValidationReport:
    """Run all validation checks and return a report.

    Args:
        tempo_url:      Tempo API base URL.
        prometheus_url: Prometheus API base URL.
        loki_url:       Loki API base URL.
        grafana_url:    Grafana API base URL.
        skip_loki:      If True, skip log-trace correlation checks.

    Returns:
        ValidationReport with all check results.
    """
    report = ValidationReport()
    report.start_time = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())

    async with aiohttp.ClientSession(timeout=REQUEST_TIMEOUT) as session:
        await validate_spans(session, tempo_url, report)
        await validate_span_durations(session, tempo_url, report)
        await assert_trace_join_groups(session, tempo_url, report)
        await validate_metrics(session, prometheus_url, report)
        await assert_sync_diagnostics_metrics(session, prometheus_url, report)
        if not skip_loki:
            await validate_log_trace_correlation(session, loki_url, tempo_url, report)
        await validate_dashboards(session, grafana_url, report)
        await validate_parity_span_attrs(session, tempo_url, report)
        await validate_parity_value_sanity(session, prometheus_url, report)

    report.end_time = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
    return report


# ---------------------------------------------------------------------------
# CLI entry point
# ---------------------------------------------------------------------------


def parse_args() -> argparse.Namespace:
    """Parse command-line arguments."""
    parser = argparse.ArgumentParser(
        description="Telemetry Validation Suite for xrpld",
        formatter_class=argparse.RawDescriptionHelpFormatter,
        epilog="""
Examples:
  # Run all validations with defaults:
  python3 validate_telemetry.py

  # Write report to file:
  python3 validate_telemetry.py --report /tmp/validation-report.json

  # Custom endpoints:
  python3 validate_telemetry.py \\
      --tempo http://tempo:3200 --prometheus http://prom:9090

  # Skip Loki checks (if log-trace correlation is not set up):
  python3 validate_telemetry.py --skip-loki
        """,
    )
    parser.add_argument(
        "--tempo",
        type=str,
        default=DEFAULT_TEMPO,
        help=f"Tempo API URL (default: {DEFAULT_TEMPO})",
    )
    parser.add_argument(
        "--prometheus",
        type=str,
        default=DEFAULT_PROMETHEUS,
        help=f"Prometheus API URL (default: {DEFAULT_PROMETHEUS})",
    )
    parser.add_argument(
        "--loki",
        type=str,
        default=DEFAULT_LOKI,
        help=f"Loki API URL (default: {DEFAULT_LOKI})",
    )
    parser.add_argument(
        "--grafana",
        type=str,
        default=DEFAULT_GRAFANA,
        help=f"Grafana API URL (default: {DEFAULT_GRAFANA})",
    )
    parser.add_argument(
        "--skip-loki",
        action="store_true",
        help="Skip log-trace correlation validation",
    )
    parser.add_argument(
        "--report",
        type=str,
        default=None,
        help="Write JSON report to this file path",
    )
    parser.add_argument(
        "--verbose",
        action="store_true",
        help="Enable debug logging",
    )
    return parser.parse_args()


def main() -> None:
    """Main entry point for the telemetry validation suite."""
    args = parse_args()

    logging.basicConfig(
        level=logging.DEBUG if args.verbose else logging.INFO,
        format="%(asctime)s [%(name)s] %(levelname)s %(message)s",
    )

    report = asyncio.run(
        run_validation(
            tempo_url=args.tempo,
            prometheus_url=args.prometheus,
            loki_url=args.loki,
            grafana_url=args.grafana,
            skip_loki=args.skip_loki,
        )
    )

    # Print summary.
    print("")
    print("=" * 60)
    print("  TELEMETRY VALIDATION REPORT")
    print("=" * 60)
    print(f"  Total checks: {report.total_checks}")
    print(f"  Passed:       {report.passed}")
    print(f"  Failed:       {report.failed}")
    print("=" * 60)
    print("")

    # Print failures.
    if report.failed > 0:
        print("FAILED CHECKS:")
        for check in report.checks:
            if not check.passed:
                print(f"  [{check.category}] {check.name}: {check.message}")
        print("")

    # Write report file.
    report_dict = report.to_dict()
    if args.report:
        with open(args.report, "w") as f:
            json.dump(report_dict, f, indent=2)
        logger.info("Report written to %s", args.report)
    else:
        print(json.dumps(report_dict, indent=2))

    # Exit with appropriate code for CI.
    sys.exit(0 if report.all_passed else 1)


if __name__ == "__main__":
    main()