rsql|

Metrics and Profiling

Monitor rsql with Prometheus metrics, namespace overview data, and optional pprof.

1 min read Updated 2026-07-27 #prometheus#pprof#operations

rsql exposes process-wide Prometheus metrics and namespace-scoped operational overview data. They serve different purposes.

Prometheus

GET /metrics is public and includes:

  • rsql_http_requests_total
  • rsql_http_request_duration_seconds
  • standard Go runtime collectors
  • standard process collectors

HTTP metric labels are method, normalized route path, and status. Namespace and table names are replaced with route placeholders to keep cardinality bounded.

Because /metrics has no authentication, expose it only on a trusted network or restrict it at the ingress.

Namespace telemetry

GET /v1/{namespace}/overview is authenticated and designed for tenant database dashboards. It contains namespace storage, traffic, latency, contention, schema, realtime, and growth data without tenant-data scans.

Telemetry writes are aggregated in memory and flushed to telemetry.db. They are intentionally not exported as high-cardinality Prometheus labels.

pprof

Enable pprof only while diagnosing a trusted environment:

bash
rsql serve \
  --api-token=dev-token \
  --pprof-enabled \
  --pprof-listen=127.0.0.1:6060

Profiles are served from the dedicated listener under /debug/pprof/. Keep the listener on loopback or a protected operations network.

bash
go tool pprof http://127.0.0.1:6060/debug/pprof/profile?seconds=30
go tool pprof http://127.0.0.1:6060/debug/pprof/heap

Capture profiles under representative load and compare them with endpoint latency, process CPU/RSS, and database/WAL growth.

PreviousConfiguration NextArchitecture