I have a container with some SYS_ADMIN capabilities, and I'm having trouble killing the pod, and am wondering what is the best way to deal with these reluctant containers?
- image: headless-chrome
securityContext:
capabilities:
add:
- SYS_ADMINSometimes a pod will zombify, even without privileged status. In order to kill a zombified pod, first try the following: kubectl delete pod <<PODNAME>> --grace-period=0 --force ...where <<PODNAME>> is the name of the offending pod. I have to do that from time to time, even with v1.7.X.
If that doesn't work, then try to first kubectl drain the node, find the Docker container (docker ps), delete it (docker rm), restart the Docker service and the Kubelet service, then kubectl uncordon the node. I've only had to do that once, and not since v1.6.X.