Skip to content

kubectl scale

Changes the number of pods of a deployment or StatefulSet.

kubectl-scale output

Syntax

kubectl scale TYPE/NAME --replicas=N [-n NAMESPACE]

Options

Option Description
deployment/NAME The deployment (also statefulset/NAME)
--replicas=N Desired number of pods; 0 stops the application without deleting it
--current-replicas=N Only scale if there are exactly N pods right now
-n NAMESPACE Namespace

kubectl autoscale deployment web --min=2 --max=5 --cpu-percent=70 adapts the number automatically to the load.

Examples

Three instead of two pods:

kubectl scale -n status deployment/web --replicas=3
kubectl get pods -n status
deployment.apps/web scaled
NAME                  READY   STATUS    RESTARTS   AGE
web-9bc5f7f5b-8ptw2   1/1     Running   0          2h
web-9bc5f7f5b-m8l96   1/1     Running   0          2h
web-9bc5f7f5b-tgsdt   1/1     Running   0          5s

Stop the application during maintenance:

kubectl scale -n status deployment/web --replicas=0
kubectl get deploy -n status
deployment.apps/web scaled
NAME   READY   UP-TO-DATE   AVAILABLE   AGE
web    0/0     0            0           2h

Scale automatically with the load:

kubectl autoscale -n status deployment web --min=2 --max=5 --cpu-percent=70
kubectl get hpa -n status
horizontalpodautoscaler.autoscaling/web autoscaled
NAME   REFERENCE        TARGETS              MINPODS   MAXPODS   REPLICAS   AGE
web    Deployment/web   cpu: 6%/70%          2         5         2          15s

When do I need it?

  • To handle more load in the short term
  • To stop an application temporarily with --replicas=0, for example during maintenance
  • To test how an application behaves with several copies

apply resets the number

The next kubectl apply sets the number back to replicas: in the manifest. For a lasting change, edit the file.

kubectl rollout · kubectl top · kubectl apply · kubectl get