Skip to content

Add support for specifying schedulerName on deployment podspec.

What does this MR do?

Adds support for specifying schedulerName on deployment podspec.

Why was this MR needed?

Runner pods are long-lived and it is preferable to mark them as non-evictable with regards to cluster autoscaler. However this prevent nodes on which runners are scheduled from scaling down. A simple solution to this is to configure a scheduler to ensure that runners are grouped together.

What's the best way to test this MR?

Render chart templates with --set schedulerName=default-scheduler and apply to a cluster. This should create a runner which will end up in in Running state.

Render chart templates with --set schedulerName=some-nonexistent-scheduler and apply to a cluster. This should create a runner which will stay in Pending state.

Merge request reports