Observability

A facilitator has a small set of numbers worth watching. This page names them, says why each matters, and what to alert on.

​

Metrics

All metrics are Prometheus-scrapable at :$METRICS_PORT/metrics.

​

Request-level

Alerting: settle_duration p95 > 15s for 5 minutes is an incident. Verify p95 > 500ms is a warning.

​

Sponsor account

Alerting: x402_sponsor_balance_xlm < SPONSOR_MIN_BALANCE_XLM is a page.

​

RPC health

​

Bazaar

​

Tracing

OpenTelemetry OTLP export via OTEL_EXPORTER_OTLP_ENDPOINT. Spans emitted:

• x402.verify (parent for a verify request)
• x402.settle (parent for a settle request)
• x402.rpc.submit (Soroban submit)
• x402.rpc.get_transaction (Soroban polling)
• x402.bazaar.index (async catalog write)

Attach x402.tx_hash on settle spans as soon as it is known.

​

Dashboards

A reference Grafana dashboard ships in dashboards/facilitator.json. It has four rows:

1. Traffic — verify/settle rate and error rate.
2. Latency — verify and settle latency percentiles.
3. Sponsor — balance and burn rate.
4. Bazaar — listing count, index lag, search QPS.

​

Recommended SLOs

• Verify success rate > 99.5% (excluding client-side codes like insufficient_funds).
• Settle success rate > 99% given verify succeeded.
• Settle p95 latency < 10s on mainnet.
• Bazaar index lag p95 < 30s.