Delete and Cleanup

You have finished testing a Deployment and its Pods. How do you cleanly remove everything? Deleting a Pod alone does not work if a ReplicaSet immediately recreates it. And does deleting a Deployment also remove its Pods? This lesson answers both questions and covers the different ways to delete resources in Kubernetes.

Cascade deletion: ownership matters

When you delete a Deployment, Kubernetes does not stop at the Deployment object itself. It follows the chain of ownership: the Deployment owns its ReplicaSet, and the ReplicaSet owns its Pods. Each level is deleted in order.

This ownership is stored in each object’s ownerReferences field. The garbage collector reads those references and triggers deletions downstream. Because of this, deleting a Pod that belongs to a Deployment is pointless: the ReplicaSet notices the count is wrong and creates a replacement immediately.

kubectl create deployment web --image=nginx:1.28 --replicas=2
kubectl get pods
kubectl delete deployment web
kubectl get pods

After the last command, the Pods are gone too. You did not delete them explicitly: the cascade took care of it.

Graceful termination

Kubernetes does not kill Pods immediately. By default, it sends a SIGTERM signal to the container’s main process and waits up to 30 seconds. That window gives the application time to finish ongoing requests, flush buffers, or close connections cleanly. After the grace period, Kubernetes sends SIGKILL and the container is force-stopped.

kubectl create deployment web --image=nginx:1.28 --replicas=2
kubectl get pods

Copy the name of one of the Pods, then:

kubectl delete pod <pod-name>

The Pod enters Terminating status during the grace period, then disappears. On a real cluster with a well-written application, no request is dropped during that window.

To skip the graceful shutdown and delete immediately:

kubectl delete pod <pod-name> --force --grace-period=0

Deleting by label

Labels make it easy to target a group of resources without listing them individually.

kubectl delete pods -l app=web

This deletes every Pod whose app label equals web in the current namespace. Since the Deployment is still alive, the ReplicaSet will recreate those Pods. To permanently remove them, delete the Deployment itself.

Deleting from a file

If you applied a resource from a YAML file, you can use the same file to delete it.

kubectl delete -f web-deployment.yaml

kubectl reads the kind and name from the file and issues the corresponding delete. This is convenient when you want to undo exactly what you applied, without remembering resource names.

Observing cascade in action

kubectl create deployment demo --image=nginx:1.28 --replicas=2
kubectl get pods

kubectl delete pod <pod-name>
kubectl get pods

kubectl delete deployment demo
kubectl delete deployment web

Deletion in Kubernetes respects ownership. Once you understand cascade and graceful termination, cleanups become predictable. The next lesson moves to the other end of the workflow: extracting exactly the information you need from kubectl get using output formatting flags.