Kubernetes Pod Lifecycle and Readiness Probes

A pod goes Pending (waiting to schedule + pull images) → Running (containers up) → Succeeded or Failed. Init containers run to completion first — good for migrations and config fetch. The two probes solve different problems and both matter. Liveness answers 'is the process hung?' — if it fails, Kubernetes kills the container. Use it for genuine deadlocks only; overly aggressive liveness is how you crash-loop production. Readiness answers 'should I send traffic to this pod?' — failure removes the pod from the Service endpoint set without killing it. This is what you want during warmup, during dependency outages, or when the pod is temporarily degraded. For apps with long startup (JVM, ML models), add a startup probe so liveness doesn't fire before the app is even up. For graceful shutdown: SIGTERM → preStop hook → drain connections → exit. If your app exits faster than load balancers notice, you drop requests — preStop sleep is the common fix.