Bhaskar Sarma:
I have a quick question on draining process …
If we are planning to drain a node but no resources on other nodes to accommodate the migrating pods ?
can we still drain with --force making the deployment to go into pending state ?
want to understand what actually happens here
Trung Tran:
If there no other ready node, you can still process the node upgrade, but have to accept downtime, because during the upgrade, the pod is not running!
unnivkn:
Hi @Bhaskar Sarma if you don’t have alternate nodes available to handle your application work loads, then it’s not a good idea to drain your nodes.
Bhaskar Sarma:
@unnivkn / @Trung Tran say I have a cluster with 5 worker nodes … but the nodes cannot accommodate additional pods as their resources are full. I am planning to perform an rolling cluster upgrade for node1 in this case what checks I should do to make sure that existing Pods on node1 can be moved to other nodes ?
Assume if they cannot move and I accidentally executed --force command then what happens to pods on node1 will they be evicted from node1 and be in pending state ?
unnivkn:
Hi @Bhaskar Sarma
(i) –> say I have a cluster with 5 worker nodes … but the nodes cannot accommodate additional pods as their resources are full. I am planning to perform an rolling cluster upgrade for node1 in this case what checks I should do to make sure that existing Pods on node1 can be moved to other nodes ?
you have to add an additional node with enough capacity to your cluster & plan for an upgrade.
(ii) Assume if they cannot move and I accidentally executed --force command then what happens to pods on node1 will they be evicted from node1 and be in pending state ?
I haven't tested this scenario, however I believe it may go to pending state.