Health & Liveness

Endpoints for load balancers, container orchestrators, and monitoring probes. All three probes bypass JWT auth.

/livez — liveness

Static probe. Returns 200 OK as long as the HTTP server is up. Uses no downstream dependencies (no DB, no filesystem, no state). Use as the Kubernetes livenessProbe / systemd watchdog / docker HEALTHCHECK — a 503 here means the process is deadlocked and should be restarted.

GET /livez

OK

/readyz — readiness

Probe that also touches the DB pool. Runs SELECT 1 against the configured backend (Postgres / MySQL / SQLite) and reports the current number of in-memory sessions. Returns 200 when the DB responds and 503 when it does not.

Use as Kubernetes readinessProbe — a 503 here temporarily removes the pod from the load balancer without killing it, letting the DB recover.

GET /readyz

{
  "db": "ok",
  "sessions_known": 500
}

When the DB probe fails:

{
  "db": "fail",
  "sessions_known": 500
}

/health — legacy alias

Same shape as /livez, kept as a compatibility alias for existing Docker HEALTHCHECK / uptime monitors. New deployments should point at /livez for liveness and /readyz for readiness.

GET /health

OK

Metrics

Prometheus text exposition on /metrics (JWT bypass). Counters and gauges for sessions, webhook dispatch, and message throughput. See the prometheus crate defaults for the metric names.