Draining a Kubernetes Node

Problem

• Load-average on the master node goes high causing Kubernetes related services to fail 
• The node may become unstable in terms of Kubernetes workloads and customer may decide to evict the node. 

Environment

• Platform9 Managed Kubernetes

Resolution

1. Run the command to drain the node:

   $ kubectl drain <node_name>

2. The above command may fail if the pods are using local storage or if daemon sets are running on the node. In that case, be aware that local storage associated with the pods will be deleted and then use the above command with switches --delete-local-data=false and --ignore-daemonsets=false.

   $ kubectl drain <node_name> --delete-local-data=false --ignore-daemonsets=false

Additional Information

Kubernetes Docs - Safely Drain Node