Observability

Health

Endpoint	Meaning
GET /healthz	Liveness: the process is up
GET /readyz	Readiness: dependencies reachable (database, NATS, cache, blob)

Every component exposes them on its listener. Use slipway healthcheck --url http://127.0.0.1:8080/healthz for container health checks in distroless images (no shell or curl).

Metrics

Prometheus metrics are served at GET /metrics on every component (otel.prometheus, default on). With the Helm chart:

metrics:
  serviceMonitor: { enabled: true, interval: 30s }

Signals worth alerting on:

Signal	Why
Queue depth and oldest queued job age	Runners are saturated or missing
Runner count by status	offline runners; pool capacity
Lease expirations and runner_lost jobs	Runner instability
API error rate and latency per method	Control-plane health
Database pool saturation	Connection limits
Webhook delivery failures	Customer integrations failing
Deployment verification failures and rollbacks	Release quality

Traces

OpenTelemetry traces and metrics can be exported to any OTLP/HTTP collector:

SLIPWAY_OTEL_EXPORTER=otlp
SLIPWAY_OTEL_ENDPOINT=http://otel-collector:4318
SLIPWAY_OTEL_SAMPLE_RATIO=0.1

Requests carry trace context through the interceptor chain, so a slow API call can be followed into the database.

Logs

Structured JSON on stdout (log.format: json, level via log.level). Secrets are never logged: configuration printing is redacted and the runner masks secrets before shipping job logs. Ship stdout with your usual agent (Fluent Bit, Vector, Promtail).

Job logs are a separate thing: they stream from runners to object storage and are served through the API, subject to your plan’s retention.

Audit log streaming

Business and above can stream the audit log to S3, Splunk or Datadog. Every plan has it in-app (30 days on Free). See Security and compliance.

Slipway observing your deployments

Distinct from the above: Slipway can send deploy markers and read canary metrics from your observability tools. See Observability integrations.