rippled/docker/telemetry/workload/run-full-validation.sh

#!/usr/bin/env bash
# run-full-validation.sh — Orchestrates the full telemetry validation pipeline.
#
# Sequence:
#   1. Start the observability stack (OTel Collector, Tempo, Prometheus, Loki, Grafana)
#   2. Start a multi-node rippled cluster with full telemetry enabled
#   3. Wait for consensus
#   4. Run workload orchestrator (RPC load, TX submission, propagation wait)
#   5. Run the telemetry validation suite
#   6. Capture OTel timings and compare against committed baseline
#   7. (Optional) Run the performance overhead benchmark
#
# Usage:
#   ./run-full-validation.sh --xrpld /path/to/xrpld
#   ./run-full-validation.sh --xrpld /path/to/xrpld --with-benchmark
#   ./run-full-validation.sh --xrpld /path/to/xrpld --skip-regression
#   ./run-full-validation.sh --cleanup
#
# Exit codes:
#   0 — All validation checks and the regression gate passed
#   1 — Validation checks failed OR the regression gate detected a regression
#       OR the benchmark exceeded its overhead thresholds
#   2 — Infrastructure error (cluster/stack failed to start, workload
#       orchestration failed, timing capture failed, overhead could not be
#       measured)
#
# Every step below records its status and folds it into FINAL_EXIT; the first
# non-zero status in pipeline order is the one returned, so the earliest
# failure — the one that explains the later ones — is what the caller sees.

set -euo pipefail

# ---------------------------------------------------------------------------
# Colored output helpers
# ---------------------------------------------------------------------------
log() { printf "\033[1;34m[VALIDATE]\033[0m %s\n" "$*"; }
ok() { printf "\033[1;32m[VALIDATE]\033[0m %s\n" "$*"; }
warn() { printf "\033[1;33m[VALIDATE]\033[0m %s\n" "$*"; }
fail() { printf "\033[1;31m[VALIDATE]\033[0m %s\n" "$*"; }
die() {
    printf "\033[1;31m[VALIDATE]\033[0m %s\n" "$*" >&2
    exit 2
}

# Overall run status, folded step by step (see the exit-code table above).
FINAL_EXIT=0

# fold_exit STATUS — record a step's status in FINAL_EXIT.
# First non-zero wins, so FINAL_EXIT names the earliest failing step.
fold_exit() {
    if [ "$1" -ne 0 ] && [ "$FINAL_EXIT" -eq 0 ]; then
        FINAL_EXIT="$1"
    fi
}

# ---------------------------------------------------------------------------
# Configuration
# ---------------------------------------------------------------------------
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
TELEMETRY_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
REPO_ROOT="$(cd "$TELEMETRY_DIR/../.." && pwd)"
COMPOSE_FILE="$TELEMETRY_DIR/docker-compose.workload.yaml"
WORKDIR="/tmp/xrpld-validation"

XRPLD="${XRPLD:-$REPO_ROOT/.build/xrpld}"
NUM_NODES=5
RPC_PORT_BASE=5005
WS_PORT_BASE=6006
PEER_PORT_BASE=51235
# Inert: parsed from --rpc-rate/--rpc-duration/--tx-tps/--tx-duration and never
# read again. Load shape comes from the workload profile instead. Kept because
# the CI workflow still passes the four flags.
RPC_RATE=50
RPC_DURATION=120
TX_TPS=5
TX_DURATION=120
WITH_BENCHMARK=false
SKIP_LOKI=false
SKIP_REGRESSION=false
WORKLOAD_PROFILE="full-validation"
REPORT_DIR="$WORKDIR/reports"
# Rate window handed to Prometheus `rate()` when capturing timings. Keep
# this close to the active workload duration so histogram buckets cover
# the measurement window; longer windows dilute short-lived regressions.
REGRESSION_WINDOW="${REGRESSION_WINDOW:-3m}"
BASELINE_FILE="${BASELINE_FILE:-$SCRIPT_DIR/baselines/baseline-timings.json}"
THRESHOLDS_FILE="${THRESHOLDS_FILE:-$SCRIPT_DIR/regression-thresholds.json}"
METRICS_FILE="${METRICS_FILE:-$SCRIPT_DIR/regression-metrics.json}"

GENESIS_ACCOUNT="rHb9CJAWyB4rj91VRWn96DkukG4bwdtyTh"
GENESIS_SEED="snoPBrXtMeMyMHUVTgbuqAfg1SUTb"

# ---------------------------------------------------------------------------
# Argument parsing
# ---------------------------------------------------------------------------
usage() {
    echo "Usage: $0 [OPTIONS]"
    echo ""
    echo "Options:"
    echo "  --xrpld PATH         Path to xrpld binary"
    echo "  --nodes NUM          Number of validator nodes (default: 5)"
    echo "  --profile NAME       Workload profile (default: full-validation)"
    echo "  --with-benchmark     Also run performance overhead benchmark (telemetry off vs on)"
    echo "  --skip-loki          Skip Loki log-trace correlation checks"
    echo "  --skip-regression    Skip the OTel-baseline regression gate"
    echo "  --cleanup            Tear down everything and exit"
    echo "  -h, --help           Show this help"
    echo ""
    echo "Accepted but INERT (parsed for compatibility, then ignored):"
    echo "  --rpc-rate RPS       no effect"
    echo "  --rpc-duration SECS  no effect"
    echo "  --tx-tps TPS         no effect"
    echo "  --tx-duration SECS   no effect"
    echo ""
    echo "  Load shape comes from the workload profile (--profile), which sets"
    echo "  the rate and duration of every phase in workload-profiles.json."
    echo "  These four flags stay accepted because the CI workflow passes them."
    exit 0
}

while [ $# -gt 0 ]; do
    case "$1" in
        --xrpld)
            XRPLD="$2"
            shift 2
            ;;
        --nodes)
            NUM_NODES="$2"
            shift 2
            ;;
        # The next four are inert — see the RPC_RATE default above.
        --rpc-rate)
            RPC_RATE="$2"
            shift 2
            ;;
        --rpc-duration)
            RPC_DURATION="$2"
            shift 2
            ;;
        --tx-tps)
            TX_TPS="$2"
            shift 2
            ;;
        --tx-duration)
            TX_DURATION="$2"
            shift 2
            ;;
        --profile)
            WORKLOAD_PROFILE="$2"
            shift 2
            ;;
        --with-benchmark)
            WITH_BENCHMARK=true
            shift
            ;;
        --skip-loki)
            SKIP_LOKI=true
            shift
            ;;
        --skip-regression)
            SKIP_REGRESSION=true
            shift
            ;;
        --cleanup) # Cleanup mode
            log "Cleaning up..."
            # Match the node config path, not the bare workdir: a plain
            # "$WORKDIR" pattern also matches any shell, editor or log tail
            # whose command line merely mentions that path.
            pkill -f "$WORKDIR/node[0-9]+/xrpld\.cfg" 2>/dev/null || true
            docker compose -f "$COMPOSE_FILE" down 2>/dev/null || true
            rm -rf "$WORKDIR"
            ok "Cleanup complete."
            exit 0
            ;;
        -h | --help) usage ;;
        *) die "Unknown option: $1" ;;
    esac
done

# ---------------------------------------------------------------------------
# Prerequisites
# ---------------------------------------------------------------------------
log "Checking prerequisites..."
[ -x "$XRPLD" ] || die "xrpld binary not found: $XRPLD"
command -v docker >/dev/null 2>&1 || die "docker not found"
docker compose version >/dev/null 2>&1 || die "docker compose (v2) not found"
command -v python3 >/dev/null 2>&1 || die "python3 not found"
command -v curl >/dev/null 2>&1 || die "curl not found"
command -v jq >/dev/null 2>&1 || die "jq not found"
[ -f "$COMPOSE_FILE" ] || die "docker-compose.workload.yaml not found"

# Install Python dependencies.
log "Installing Python dependencies..."
pip3 install -q -r "$SCRIPT_DIR/requirements.txt" 2>/dev/null ||
    pip install -q -r "$SCRIPT_DIR/requirements.txt" 2>/dev/null ||
    warn "Could not install Python dependencies — they may already be present"

ok "Prerequisites verified."

# ---------------------------------------------------------------------------
# Cleanup previous run
# ---------------------------------------------------------------------------
log "Cleaning up previous run..."
# Narrowed for the same reason as the --cleanup branch above.
pkill -f "$WORKDIR/node[0-9]+/xrpld\.cfg" 2>/dev/null || true
sleep 2
rm -rf "$WORKDIR"
mkdir -p "$WORKDIR" "$REPORT_DIR"

# ---------------------------------------------------------------------------
# Step 1: Start observability stack
# ---------------------------------------------------------------------------
log "Step 1: Starting observability stack..."
# Point the collector's log mount at this run's workdir so the filelog
# receiver tails the per-node debug.log files generated below.
XRPLD_LOG_DIR="$WORKDIR" docker compose -f "$COMPOSE_FILE" up -d

log "Waiting for OTel Collector..."
for attempt in $(seq 1 30); do
    status=$(curl -so /dev/null -w '%{http_code}' http://localhost:4318/ 2>/dev/null || echo 000)
    if [ "$status" != "000" ]; then
        ok "OTel Collector ready (attempt $attempt)"
        break
    fi
    [ "$attempt" -eq 30 ] && die "OTel Collector not ready after 30s"
    sleep 1
done

log "Waiting for Tempo..."
for attempt in $(seq 1 30); do
    if curl -sf "http://localhost:3200/ready" >/dev/null 2>&1; then
        ok "Tempo ready (attempt $attempt)"
        break
    fi
    [ "$attempt" -eq 30 ] && die "Tempo not ready after 30s"
    sleep 1
done

log "Waiting for Prometheus..."
for attempt in $(seq 1 30); do
    if curl -sf "http://localhost:9090/-/healthy" >/dev/null 2>&1; then
        ok "Prometheus ready (attempt $attempt)"
        break
    fi
    [ "$attempt" -eq 30 ] && die "Prometheus not ready after 30s"
    sleep 1
done

# ---------------------------------------------------------------------------
# Step 2: Generate validator keys and start cluster
# ---------------------------------------------------------------------------
log "Step 2: Starting $NUM_NODES-node validator cluster..."

bash "$SCRIPT_DIR/generate-validator-keys.sh" "$XRPLD" "$NUM_NODES" "$WORKDIR"

for i in $(seq 1 "$NUM_NODES"); do
    NODE_DIR="$WORKDIR/node$i"
    mkdir -p "$NODE_DIR/nudb" "$NODE_DIR/db"

    RPC_PORT=$((RPC_PORT_BASE + i - 1))
    WS_PORT=$((WS_PORT_BASE + i - 1))
    PEER_PORT=$((PEER_PORT_BASE + i - 1))
    SEED=$(jq -r ".[$((i - 1))].seed" "$WORKDIR/validator-keys.json")

    # Build ips_fixed.
    IPS_FIXED=""
    for j in $(seq 1 "$NUM_NODES"); do
        if [ "$j" -ne "$i" ]; then
            IPS_FIXED="${IPS_FIXED}127.0.0.1 $((PEER_PORT_BASE + j - 1))
"
        fi
    done

    cat >"$NODE_DIR/xrpld.cfg" <<EOCFG
[server]
port_rpc
port_ws
port_peer

[port_rpc]
port = $RPC_PORT
ip = 127.0.0.1
admin = 127.0.0.1
protocol = http

[port_ws]
port = $WS_PORT
ip = 127.0.0.1
admin = 127.0.0.1
protocol = ws

[port_peer]
port = $PEER_PORT
ip = 0.0.0.0
protocol = peer

[node_db]
type=NuDB
path=$NODE_DIR/nudb
online_delete=256

[database_path]
$NODE_DIR/db

[debug_logfile]
$NODE_DIR/debug.log

[validation_seed]
$SEED

[validators_file]
$WORKDIR/validators.txt

[ips]
${IPS_FIXED}

[telemetry]
enabled=1
service_instance_id=validator-${i}
endpoint=http://localhost:4318/v1/traces
batch_size=512
batch_delay_ms=2000
max_queue_size=2048
trace_rpc=1
trace_transactions=1
trace_consensus=1
trace_peer=1
trace_ledger=1

[insight]
# Native OTel metrics via OTLP/HTTP. The collector has no StatsD receiver
# (its metrics pipeline is [otlp, spanmetrics]), so beast::insight must export
# over OTLP for system metrics to reach Prometheus at all. server=otel is the
# only load-bearing key here -- it selects the OTel collector.
#
# endpoint below is parsed but inert, exactly like prefix: CollectorManager
# reads it and hands it to OTelCollector, which uses it only in its startup log
# line. The URL the metric exporter actually posts to is derived in
# Telemetry::initMetrics() from [telemetry] endpoint, by swapping the trailing
# /v1/traces for /v1/metrics. It is kept here so the log line names the right
# URL; changing it would not redirect a single metric.
#
# No prefix is set on purpose. It would be inert: OTelCollector applies no
# prefix to instrument names, so exported names are the lowercased raw names
# (jobq_job_count, rpc_requests_total, total_bytes_in). The service is
# identified by the OTel resource service.name, not by a name prefix.
server=otel
endpoint=http://localhost:4318/v1/metrics

[rpc_startup]
# info is required here, not cosmetic, and is the minimum that works. The
# log-trace correlation checks need a log line emitted while a span is current
# on the thread. A span becomes current either as a ScopedSpanGuard or by
# activating a plain SpanGuard via activate()/activateIfLive(); a plain
# SpanGuard that is never activated yields no trace_id.
#
# The guaranteed correlated line at info is the consensus accept pair in
# RCLConsensus.cpp (an if/else, so exactly one fires every accepted round).
# doAccept activates the accept span as ambient over its whole body, so both
# branches carry a trace_id. At ~4s per round that is dozens of correlated
# lines per run.
#
# Do NOT raise this to debug to get more coverage. debug puts synchronous log
# I/O inside ledger.build, consensus.accept and tx.apply -- the same spans
# whose p50/p95/p99 latencies regression-metrics.json gates -- so a baseline
# captured at debug would bake log I/O into the numbers permanently.
{ "command": "log_level", "severity": "info" }

[signing_support]
true

[ssl_verify]
0
EOCFG

    "$XRPLD" --conf "$NODE_DIR/xrpld.cfg" --start >"$NODE_DIR/stdout.log" 2>&1 &
    echo $! >"$NODE_DIR/xrpld.pid"
    log "  Node $i: RPC=$RPC_PORT WS=$WS_PORT Peer=$PEER_PORT PID=$!"
done

# ---------------------------------------------------------------------------
# Step 3: Wait for consensus
# ---------------------------------------------------------------------------
# Report whether a node process is still alive.
#
# A child that has exited but not yet been waited on still answers `kill -0`,
# because the zombie keeps its pid until someone collects it. Checking only
# `kill -0` therefore reads a dead node as alive for the whole readiness
# window, which is how a crashed node came to look like a slow one.
node_running() {
    local pid="$1" state
    kill -0 "$pid" 2>/dev/null || return 1
    if [ -r "/proc/$pid/stat" ]; then
        state=$(awk '{print $3}' "/proc/$pid/stat" 2>/dev/null || echo "?")
        [ "$state" != "Z" ] || return 1
    fi
    return 0
}

# Print why each stopped node stopped: its wait status, then its last output.
#
# The status is the discriminator this harness was missing -- 137 is SIGKILL
# (the kernel reclaiming memory), 139 a segfault, 134 an abort, anything under
# 128 a deliberate exit. The nodes are direct children of this script, so their
# status is still retrievable until something waits on them.
#
# stdout is printed inline rather than left to the artifact upload because a
# node that dies before its debug log opens writes nothing else, and a
# cancelled run uploads nothing at all.
report_stopped_nodes() {
    local i pid status
    for i in $(seq 1 "$NUM_NODES"); do
        pid=$(cat "$WORKDIR/node$i/xrpld.pid" 2>/dev/null || echo "")
        [ -n "$pid" ] || continue
        node_running "$pid" && continue
        status=0
        wait "$pid" 2>/dev/null || status=$?
        warn "node$i (pid $pid) is not running — wait status $status"
        if [ -s "$WORKDIR/node$i/stdout.log" ]; then
            warn "node$i last output:"
            tail -n 15 "$WORKDIR/node$i/stdout.log" | sed 's/^/      /' >&2
        else
            warn "node$i wrote no stdout at all"
        fi
    done
}

log "Step 3: Waiting for consensus..."
for attempt in $(seq 1 120); do
    ready=0
    # Reset each attempt so a timeout reports the final state, not a history.
    laggards=""
    for i in $(seq 1 "$NUM_NODES"); do
        port=$((RPC_PORT_BASE + i - 1))
        state=$(curl -sf "http://localhost:$port" \
            -d '{"method":"server_info"}' 2>/dev/null |
            jq -r '.result.info.server_state' 2>/dev/null || echo "")
        if [ "$state" = "proposing" ]; then
            ready=$((ready + 1))
        else
            # Name the node and what it last reported. A bare count says a
            # node is missing but not which one, which leaves nothing to grep
            # for in the artifacts. An empty state means the RPC port did not
            # answer at all, which usually means the process is gone.
            laggards="$laggards node$i=${state:-unreachable}"
        fi
    done
    if [ "$ready" -ge "$NUM_NODES" ]; then
        ok "All $NUM_NODES nodes proposing (attempt $attempt)"
        break
    fi
    # A stopped process will never reach proposing. Waiting out the rest of the
    # window only delays the same failure and buries its cause under two
    # minutes of progress output.
    stopped=0
    for n in $(seq 1 "$NUM_NODES"); do
        p=$(cat "$WORKDIR/node$n/xrpld.pid" 2>/dev/null || echo "")
        if [ -n "$p" ] && ! node_running "$p"; then
            stopped=$((stopped + 1))
        fi
    done
    if [ "$stopped" -gt 0 ]; then
        echo ""
        report_stopped_nodes
        die "$stopped of $NUM_NODES node(s) stopped during startup; only $ready reached proposing. Not proposing:${laggards}. Per-node status is above, then '$0 --cleanup'."
    fi
    if [ "$attempt" -eq 120 ]; then
        # Fatal, not a warning. A partial cluster still answers queries, so the
        # run would complete and report unrelated span/metric failures: series
        # counts scale with the number of live nodes, and spans that need a
        # quorum are simply never emitted. One infrastructure error here is
        # worth more than a pile of misleading assertion failures later.
        echo ""
        # Every node is still running but not proposing, so this is a genuine
        # convergence problem rather than a crash. Run the reporter anyway: it
        # is a no-op when nothing stopped, and it costs nothing to be sure.
        report_stopped_nodes
        die "Consensus timeout — only $ready/$NUM_NODES nodes proposing after ${attempt}s. Not proposing:${laggards}. Check $WORKDIR/node*/debug.log and $WORKDIR/node*/stdout.log (a node that died before its log sink opened writes only the latter), then '$0 --cleanup'."
    fi
    printf "\r  %d/%d nodes proposing..." "$ready" "$NUM_NODES"
    sleep 1
done
echo ""

# Wait for first validated ledger.
log "Waiting for validated ledger..."
for attempt in $(seq 1 60); do
    val_seq=$(curl -sf "http://localhost:$RPC_PORT_BASE" \
        -d '{"method":"server_info"}' 2>/dev/null |
        jq -r '.result.info.validated_ledger.seq // 0' 2>/dev/null || echo 0)
    if [ "$val_seq" -gt 2 ] 2>/dev/null; then
        ok "Validated ledger: seq $val_seq"
        break
    fi
    # Fatal for the same reason as the consensus timeout above, and because
    # several assertions are gated on a validated ledger existing at all:
    # ledger_economy{metric="base_fee_xrp"} is only observed from a validated
    # ledger, and complete_ledgers stays absent while the range is empty.
    if [ "$attempt" -eq 60 ]; then
        die "No validated ledger after ${attempt}s (last seq: $val_seq). Check $WORKDIR/node*/debug.log, then '$0 --cleanup'."
    fi
    sleep 1
done

# ---------------------------------------------------------------------------
# Step 4: Run workload orchestrator
# ---------------------------------------------------------------------------
log "Step 4: Running workload orchestrator (profile: $WORKLOAD_PROFILE)..."

WS_ENDPOINTS=""
for i in $(seq 1 "$NUM_NODES"); do
    WS_ENDPOINTS="$WS_ENDPOINTS ws://localhost:$((WS_PORT_BASE + i - 1))"
done

ORCHESTRATOR_EXIT=0
python3 "$SCRIPT_DIR/workload_orchestrator.py" \
    --profile "$WORKLOAD_PROFILE" \
    --endpoints $WS_ENDPOINTS \
    --report "$REPORT_DIR/workload-report.json" \
    --report-dir "$REPORT_DIR" || ORCHESTRATOR_EXIT=$?

if [ "$ORCHESTRATOR_EXIT" -eq 0 ]; then
    ok "Workload orchestration complete."
else
    # Treated as an infrastructure error: the span and metric assertions below
    # would be graded against traffic that was never generated.
    fail "Workload orchestrator failed (exit $ORCHESTRATOR_EXIT) — the checks below run against incomplete traffic"
    fold_exit 2
fi

# ---------------------------------------------------------------------------
# Step 5: Run telemetry validation suite
# ---------------------------------------------------------------------------
log "Step 5: Running telemetry validation suite..."

VALIDATION_ARGS="--report $REPORT_DIR/validation-report.json"
if [ "$SKIP_LOKI" = true ]; then
    VALIDATION_ARGS="$VALIDATION_ARGS --skip-loki"
fi

VALIDATION_EXIT=0
python3 "$SCRIPT_DIR/validate_telemetry.py" $VALIDATION_ARGS || VALIDATION_EXIT=$?

if [ "$VALIDATION_EXIT" -eq 0 ]; then
    ok "All telemetry validation checks passed!"
else
    fail "Some telemetry validation checks failed (exit $VALIDATION_EXIT)"
fi
fold_exit "$VALIDATION_EXIT"

# ---------------------------------------------------------------------------
# Step 6: Capture OTel timings and run the regression comparison
# ---------------------------------------------------------------------------
# This step ALWAYS captures timings (so CI always has an artifact from which
# to bootstrap/refresh the committed baseline). The comparator then either:
#   - prints the paste-me JSON when the baseline is a placeholder, or
#   - enforces thresholds and fails the run on regression.
# Use --skip-regression to opt out (e.g. for ad-hoc local exploration).
TIMINGS_FILE="$REPORT_DIR/timings.json"
REGRESSION_REPORT="$REPORT_DIR/regression-report.json"
REGRESSION_EXIT=0

if [ "$SKIP_REGRESSION" != true ]; then
    log "Step 6: Capturing OTel timings from Prometheus..."
    if python3 "$SCRIPT_DIR/capture_timings.py" \
        --prometheus "http://localhost:9090" \
        --metrics "$METRICS_FILE" \
        --output "$TIMINGS_FILE" \
        --window "$REGRESSION_WINDOW" \
        --profile "$WORKLOAD_PROFILE"; then
        ok "Timings captured: $TIMINGS_FILE"
    else
        fail "Failed to capture timings — skipping regression comparison."
        REGRESSION_EXIT=2
        SKIP_REGRESSION=true
    fi
fi

if [ "$SKIP_REGRESSION" != true ]; then
    log "Comparing against baseline $BASELINE_FILE..."
    python3 "$SCRIPT_DIR/compare_to_baseline.py" \
        --timings "$TIMINGS_FILE" \
        --baseline "$BASELINE_FILE" \
        --thresholds "$THRESHOLDS_FILE" \
        --report "$REGRESSION_REPORT" || REGRESSION_EXIT=$?
    if [ "$REGRESSION_EXIT" -eq 0 ]; then
        ok "Regression gate passed (or baseline placeholder — paste JSON printed above)."
    elif [ "$REGRESSION_EXIT" -eq 1 ]; then
        fail "Regression detected — see $REGRESSION_REPORT"
    else
        fail "Regression comparator internal error (exit $REGRESSION_EXIT)"
    fi
else
    warn "Regression gate skipped."
fi
fold_exit "$REGRESSION_EXIT"

# ---------------------------------------------------------------------------
# Step 7: (Optional) Run overhead benchmark
# ---------------------------------------------------------------------------
BENCHMARK_EXIT=0
if [ "$WITH_BENCHMARK" = true ]; then
    log "Step 7: Running performance benchmark..."
    bash "$SCRIPT_DIR/benchmark.sh" \
        --xrpld "$XRPLD" \
        --duration 120 \
        --nodes 3 \
        --output "$REPORT_DIR" || BENCHMARK_EXIT=$?

    if [ "$BENCHMARK_EXIT" -eq 0 ]; then
        ok "Benchmark within overhead thresholds."
    elif [ "$BENCHMARK_EXIT" -eq 1 ]; then
        # A measured threshold breach — same class as a failed check.
        fail "Benchmark exceeded overhead thresholds (exit 1)"
        fold_exit 1
    else
        # benchmark.sh could not produce a usable measurement (e.g. incomplete
        # system metrics). Reported as an infrastructure error, not a perf
        # regression: nothing was measured, so nothing was breached.
        fail "Benchmark could not measure overhead (exit $BENCHMARK_EXIT) — treated as an infrastructure error"
        fold_exit 2
    fi
fi

# ---------------------------------------------------------------------------
# Summary
# ---------------------------------------------------------------------------
echo ""
echo "==========================================================="
echo "  FULL VALIDATION RESULTS"
echo "==========================================================="
echo ""
echo "  Reports directory: $REPORT_DIR"
echo ""
ls -la "$REPORT_DIR/" 2>/dev/null || true
echo ""
echo "  Observability stack is running:"
echo "    Tempo:         http://localhost:3200"
echo "    Grafana:       http://localhost:3000"
echo "    Prometheus:    http://localhost:9090"
echo ""
echo "  xrpld nodes ($NUM_NODES) are running:"
for i in $(seq 1 "$NUM_NODES"); do
    rpc=$((RPC_PORT_BASE + i - 1))
    ws=$((WS_PORT_BASE + i - 1))
    pid=$(cat "$WORKDIR/node$i/xrpld.pid" 2>/dev/null || echo 'unknown')
    echo "    Node $i: RPC=$rpc WS=$ws PID=$pid"
done
echo ""
echo "  To tear down:"
echo "    $0 --cleanup"
echo ""
echo "  Step statuses (0 = ok):"
echo "    Workload orchestration: $ORCHESTRATOR_EXIT"
echo "    Telemetry validation:   $VALIDATION_EXIT"
echo "    Regression gate:        $REGRESSION_EXIT"
echo "    Overhead benchmark:     $BENCHMARK_EXIT"
echo ""
echo "==========================================================="

# FINAL_EXIT already holds the first non-zero step status (see fold_exit).
exit "$FINAL_EXIT"