Startup Probes and Probe Configuration

Your liveness probe fires at second 5. Your JVM app hasn't finished loading its Spring context yet. Kubernetes kills the container. The pod restarts. The liveness probe fires at second 5 again. Repeat until CrashLoopBackOff.

This is the canonical slow-start failure mode, and the usual fix (padding initialDelaySeconds until the probe stops firing) trades one problem for another. Set it to 60 seconds and you won't detect a real deadlock for a full minute after every container restart. Set it to 120 and you've effectively disabled liveness for two minutes on every rollout.