oam/knowledge base/kubernetes/drain a node.md

Drain a K8S cluster node

Table of contents

1. TL;DR
2. Further readings
3. Sources

TL;DR

1. mark the node as unschedulable (cordon):

   $ kubectl cordon 'kworker-rj2'
   node/kworker-rj2 cordoned
   

2. remove pods running on the node:

   $ kubectl drain 'kworker-rj2' --grace-period=300 --ignore-daemonsets=true
   node/kworker-rj2 already cordoned
   WARNING: ignoring DaemonSet-managed Pods: kube-system/calico-node-fl8dl, kube-system/kube-proxy-95vdf
   evicting pod default/my-dep-557548758d-d2pmd
   pod/my-dep-557548758d-d2pmd evicted
   node/kworker-rj2 evicted
   

3. do to the node what you need to do

4. make the node available again:

   $ kubectl uncordon 'kworker-rj2'
   node/kworker-rj2 uncordoned
   

Further readings

• Kubernetes
• kubectl

Sources

All the references in the further readings section, plus the following:

• How to drain a node in Kubernetes