Skip to content

kubectl rollout

Watches, restarts and undoes updates of a deployment.

kubectl-rollout output

Syntax

kubectl rollout SUBCOMMAND TYPE/NAME [-n NAMESPACE]

Options

Option Description
status Wait until the update is finished and report the result
history List the previous versions (revisions)
undo Go back to the previous version; --to-revision=N to a specific one
restart Replace all pods one after the other, for example to load a new ConfigMap
pause, resume Pause and continue an update

A new image version is set with kubectl set image deployment/web nginx=nginx:1.29 or by changing the manifest.

Examples

Roll out a new image and wait for it:

kubectl set image -n status deployment/web nginx=nginx:1.29
kubectl rollout status -n status deployment/web
deployment.apps/web image updated
Waiting for deployment "web" rollout to finish: 1 out of 3 new replicas have been updated...
Waiting for deployment "web" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "web" rollout to finish: 1 old replicas are pending termination...
deployment "web" successfully rolled out

The previous versions:

kubectl rollout history -n status deployment/web
deployment.apps/web
REVISION  CHANGE-CAUSE
1         <none>
2         <none>

Back to the previous version:

kubectl rollout undo -n status deployment/web
deployment.apps/web rolled back

Restart all pods one after the other:

kubectl rollout restart -n status deployment/web
deployment.apps/web restarted

When do I need it?

  • After a change, to know when all new pods are running
  • When a new version causes errors and you have to go back quickly
  • To restart all pods of an application without downtime

Updates without downtime

A deployment replaces pods one at a time and only removes an old one once the new one is ready. Add a readinessProbe so that Kubernetes knows when your application is really ready.

undo and your manifest

kubectl rollout undo changes the cluster, not your YAML file. Correct the file too, otherwise the next kubectl apply brings back the faulty version.

kubectl apply · kubectl scale · kubectl get · helm