DeploymentsRolloutsTroubleshootingCKAD

Rollout Stuck? How to Unstick a Kubernetes Deployment

Progress deadline exceeded, old pods never terminate, new ReplicaSet at zero — a calm order of operations before rollback or delete.

2026-06-091 min read

You shipped a new image. The rollout never finishes.

kubectl rollout status deployment/web -n prod

Waiting for deployment "web" rollout to finish: 1 old replicas are pending termination...

Or:

error: deployment "web" exceeded its progress deadline

What Deployment is waiting for

A Deployment rollout completes when new ReplicaSet pods become Available and old pods terminate according to strategy:

Stuck usually means: new pods not Ready, old pods won't drain, or PDB / quotas block progress.

kubectl get rs -n prod -l app=web
kubectl get pods -n prod -l app=web -o wide
kubectl describe deployment web -n prod

Compare old vs new ReplicaSet pod counts and Ready state.

Then for failing new pods:

kubectl describe pod <new-pod> -n prod
kubectl logs <new-pod> -n prod --previous

Common blockers: probe failure on new image, insufficient CPU, PVC still attaching, ProgressDeadlineSeconds too aggressive.

Situation	Sensible move
Bad new image	`kubectl rollout undo deployment/web -n prod`
Need history	`kubectl rollout history` then undo to revision
Paused by human	`kubectl rollout resume deployment/web`
Single bad pod	Fix manifest; let controller reconcile

Avoid first: deleting the Deployment (destroys revision history), random scale without reading Events, or --force on everything.

Exams often combine rollout status + describe new pod + undo. Speed comes from knowing get rs / get pods before undo.

Drill Deployment Rollouts scenarios on Decision Trainer — timed exam sprints included in CKAD Exam Prep.