CODEWITHSUNDEEP
kubernetes
user9524761
user9524761

Reputation: 851

How do I force delete kubernetes pods?

I have the following pods:

NAME                                                 READY     STATUS        RESTARTS   AGE
xxx-myactivities-79f49cdfb4-nwg22                      1/1       Terminating   0          10h
xxx-mysearch-55864b5c59-6bnwl                          1/1       Terminating   0          1d
xxx-mysearch-55864b5c59-rpn48                          1/1       Terminating   0          13h
xxx-mysearch-6ff9bbb7cb-9qgbb                          1/1       Terminating   0          3d

I am running the following code to forcefully delete those pods:

#
# Clean up dying pods
#
pods=$( kubectl get pods | grep -v Running | tail -n +2 | awk -F " " '{print $1}' )
for pod in $pods;
do
    kubectl delete pod $pod --force
done

Here is the output:

pod "xxx-myactivities-79f49cdfb4-nwg22" deleted
pod "xxx-mysearch-55864b5c59-6bnwl" deleted
pod "xxx-mysearch-55864b5c59-rpn48" deleted
pod "xxx-mysearch-6ff9bbb7cb-9qgbb" deleted

After cleaning up, those pods still hang around.

NAME                                                 READY     STATUS        RESTARTS   AGE
xxx-myactivities-79f49cdfb4-nwg22                      1/1       Terminating   0          10h
xxx-mysearch-55864b5c59-6bnwl                          1/1       Terminating   0          1d
xxx-mysearch-55864b5c59-rpn48                          1/1       Terminating   0          13h
xxx-mysearch-6ff9bbb7cb-9qgbb                          1/1       Terminating   0          3d

How do I clean up those pods?

Upvotes: 82

Views: 211122

Answers (10)

Chris
Chris

Reputation: 5886

I recently had a pod that couldn't be removed by any of the methods listed so far. It turns out this was because of a finalizer on the pod. The operator that handled this finalizer had already been deleted.

To work around this I did kubectl edit pods/{the pod name}, then manually removed the finalizer and saved it. After that it immediately disappeared.

Upvotes: 1

mirekphd
mirekphd

Reputation: 6841

Please wait and reconsider the approach. The force-killed pod in most situations will continue running, so its probably better idea to get inside the pod (using kubectl exec or oc rsh) and kill any offending processes first (e.g. those that exhaust resources). Once force-killed, the pod will no longer be accessible through the k8s API (its zombie processes that continue running despite pod being no longer there will then have to be killed by someone with the root on the host).

Upvotes: 0

RicHincapie
RicHincapie

Reputation: 4003

I use this one to forcefully delete all pods that are not in a Running state:

kubectl get po | grep -v Running | awk 'NR>1 {print $1}' | xargs kubectl delete po --force --grace-period=0

Make sure you are in the namespace you want to work with.

If you want to delete all pods in Running state, remove the -v flag.

Upvotes: 3

Sachin Arote
Sachin Arote

Reputation: 1027

For deleting pod forcefully use this.

kubectl delete pod <Pod_Name> -n <namespace_name> --grace-period=0 --force

OR

kubectl delete pod <Pod_Name> -n <namespace_name> --wait=false

For reference please follow below link. https://kubernetes.io/docs/tasks/run-application/force-delete-stateful-set-pod/

Upvotes: 28

npocmaka
npocmaka

Reputation: 57282

I'm using this:

kubectl get pods | grep part-of-the-pods-name|awk '{split($0,a," "); print a[1]}'|xargs -I {} kubectl delete pods {} --grace-period=0

additional grep can be added if needed.

Upvotes: 1

moshe beeri
moshe beeri

Reputation: 2037

In case you used Helm to install the pod just run

helm delete Your_Deployment_Name -n The_Namespace

Upvotes: 2

eQ19
eQ19

Reputation: 10711

To clean the pods you need to delete their deployments namespace.

First discover that deployments existed:

$ kubectl get deployments --all-namespaces
NAME                               READY     STATUS        RESTARTS   AGE
chetabahana-web-584b95d576-62ccj   1/1       Running       0          20m
tutorial-web-56fbccc56b-wbwjq      1/1       Running       0          1m

Delete the deployment <NAME>-xxxx like this:

$ kubectl delete deployment <NAME>

For example to delete tutorial-web-56fbccc56b-wbwjq run:

$ kubectl delete deployment tutorial

Then all corresponded pods of tutorial-xxxx will terminate by itself.

NAME                               READY     STATUS        RESTARTS   AGE
chetabahana-web-584b95d576-62ccj   1/1       Running       0          20m
tutorial-web-56fbccc56b-wbwjq      0/1       Terminating   0          1m

Upvotes: 97

lanni654321
lanni654321

Reputation: 1087

kubectl get pod --all-namespaces | awk '{if ($4 != "Running") system ("kubectl -n " $1 " delete pods " $2  " --grace-period=0 " " --force ")}'

you can use this command

Upvotes: 5

belabrinel
belabrinel

Reputation: 929

To delete all pods in terminating state with one command do:

for p in $(kubectl get pods | grep Terminating | awk '{print $1}'); do kubectl delete pod $p --grace-period=0 --force;done

Upvotes: 9

James Knott
James Knott

Reputation: 1578

You have these alternatives:

kubectl delete pod xxx --now

Or

SSH into the node the stuck pod was scheduled on Running docker ps | grep {pod name} to get the Docker Container ID Running docker rm -f {container id}

Or

kubectl delete pod NAME --grace-period=0 --force

Upvotes: 104

Related Questions