Question

You administer a Kubernetes cluster with 12 GPU nodes running long-lived training jobs. You need to apply a new GPU driver version to one node, but you want the node to stop accepting NEW pods immediately while allowing the currently running training pods to finish naturally over the next several hours before you reboot. Which action best achieves this?

Accepted Answer

Run 'kubectl cordon' on the node to mark it unschedulable without evicting running pods.

Answer

Run 'kubectl drain' on the node to immediately evict all pods and block scheduling.

Answer

Add a NoExecute taint to the node so the scheduler routes new pods elsewhere.

Answer

Delete the node object from the cluster and re-add it after the driver upgrade.