Let kubelet respect pod's finalizers

[hex108]

What would you like to be added:
If pod's finalizer list is not empty, kubelet should keep it running until its finalizer list is empty even its deletion timestamp is set.

Why is this needed:
If I understand it correctly, finalizer works as a pre-delete hook. We create ingress for pods, and want to remove pod from ingress firstly before terminating it, then we could avoid request traffic are sent to pods that are not running.

/sig node

[liggitt]

Finalizers control the API object lifecycle, not the lifecycle of the pod process. A pod that has had deletion requested is subject to it's process being terminated.

You can use the preStop lifecycle and/or terminationGracePeriodSeconds in combination with SIGTERM handling in your container process to control how long a pod remains running after deletion is requested to give your controller time to observe the deletionTimestamp and react prior to the container process being terminated.

/close

[k8s-ci-robot]

@liggitt: Closing this issue.

In response to this:

Finalizers control the API object lifecycle, not the lifecycle of the pod process. A pod that has had deletion requested is subject to it's process being terminated.

You can use the preStop lifecycle and/or terminationGracePeriodSeconds in combination with SIGTERM handling in your container process to control how long a pod remains running after deletion is requested to give your controller time to observe the deletionTimestamp and react prior to the container process being terminated.

/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.