Healthchecks

All Go services under svc/ expose standardized HTTP probes via the runner. These endpoints are intended for Kubernetes liveness, readiness, and startup probes. Use them instead of service-specific ad hoc endpoints.

Endpoints

The default prefix is /health. Services can override the prefix when registering endpoints.

• GET <prefix>/live returns 200 when the process has started. Returns 503 before startup completes.
• GET <prefix>/ready returns 200 when the process is started and not shutting down, and all readiness checks pass. Returns 503 otherwise.
• GET <prefix>/startup returns 200 after startup completes. Returns 503 before startup completes.

All endpoints return JSON with a status field. The readiness endpoint includes a checks map with per-check results when checks are registered.

Readiness checks

Readiness checks are registered by services with the runner. Each check runs with a timeout (default 500ms). If any check fails or times out, /health/ready returns 503 and includes the error message in checks.

Service coverage

These probes are exposed on every service entrypoint that uses runner:

• API
• Ctrl API
• Ctrl worker
• Frontline (both HTTP and HTTPS listeners)
• Sentinel
• Krane
• Vault
• Preflight

If you add a new service under svc/, call r.RegisterHealth(mux) and ensure r.Wait(ctx) is used to drive lifecycle state.