kube-safe-drain

This script can be used to safely drain all the cordoned nodes in a Kubernetes cluster, by performing all the necessary rollout restarts.

Usage

Important: this has been tested with Python 3.11.2 on Debian 12 (bookworm).

First you have to cordon all the nodes you want to drain. Example:

1

kubectl cordon mynode

Then generate the list of rollout restart commands:

1

kubectl get -A node,pod,cj,deploy,ds,job,rs,sts -ojson | python3 main.py

Review the list and run the commands. The rollout restarts may take some time, even after the commands have completed the execution.

Finally, check that the nodes have been drained successfully. Example:

1

kubectl drain --ignore-daemonsets --delete-emptydir-data --dry-run=client mynode