1. how to stop pod that still terminating #

Sometimes there was a pod keep in terminating status. For example, the terminationGracePeriodSeconds being set to a very large value.

Do not force delete pod through kubectl delete pod -f option, this will cause container process residual on node unless you make sure the container on node already terminated.

You can stopping by following ways.

attact to pod container, find which PID that block container exit, then kill it;
if the block process is init process(PID is 1), then do the following steps.

if k8s using containerd as runtime, try using ctr command.

ctr -n k8s.io container ls | grep $imageName
#find taskID
ctr -n k8s.io task kill taskID

Or using crictl

crictl ps -a  | grep $podName 
crictl stop $containerID

Why ctr c ls shows nothing? All k8s pod containerd run in k8s.io namespace, containerd namespace is not POD namespace in k8s, containerd use default as default namespace
containerd's container is different from docker container. when contaienrd create container, the container just be created but not started. ctr start will launch container process, this calls task.