Opscraft
Kubernetes troubleshooting · Lesson 4 of 11 · 3 min read

Probes, readiness and traffic

Running but not Ready

READY 0/1 with Running status: the readiness probe is failing, so the pod is excluded from Service endpoints.

kubectl describe pod orders-7d9f-abc -n shop | grep -A3 -i readiness
kubectl get endpointslices -n shop -l kubernetes.io/service-name=orders

Common mistakes: probe on the wrong port or path; probe expects 200 but the app redirects (301/302); timeout too short; readiness checks a dependency that's down (sometimes intended — removes the pod from traffic — but if all pods share the dependency, the whole Service goes to 0 endpoints and you get 503s everywhere).

Probe design summary

Probe Question Failure action Should check
startup Has the app finished starting? Keep waiting (kill after threshold) Startup complete
liveness Is the process alive/unstuck? Restart the container Cheap internal health — no dependencies
readiness Can it serve traffic now? Remove from endpoints Ready to serve; critical local dependencies
Enlarge diagramstartupfinished starting?keeps the others waitinglivenessprocess stuck?restarts the containerreadinesscan serve now?removes from the Service
Figure 1. Three probes, three questions, three different actions. Liveness should never check dependencies.
Swipe sideways to see all of it, or tap to open it full screen.

Rollouts that get stuck

kubectl rollout status hangs; the Deployment shows Progressing=False, ProgressDeadlineExceeded. New pods aren't becoming Ready (crash, failing readiness, image pull, Pending for capacity) — and because maxUnavailable protects availability, old pods stay. Diagnose the new ReplicaSet's pods, then fix forward or roll back (revert in Git under Flux).