Health & Liveness Probes

Configure proper startup, liveness, and readiness probes for zero-downtime operations.

Language / Topic

Kubernetes

Rules

Health & Liveness Probes · balanced balanced

- Implement both `readinessProbe` and `livenessProbe` for HTTP/TCP/exec checks

- Use `startupProbe` for legacy apps with extremely long or unpredictable initialization times

- Do not couple readiness checks with backend database latency—check only if the app itself can accept traffic

- `livenessProbe` decides if a pod is dead and should be restarted; restrict its logic to purely application-level deadlocks

- `readinessProbe` decides if a pod should receive ingress traffic; this can temporarily fail without causing restarts

- Avoid cascading failures by ensuring your readiness/liveness checks do not hit congested DBs or rate-limited external APIs