User Story

As a user i want to ensure that i know what the cluster autoscaler is doing with respect to ignoring labels when balancing node groups. Removing the automatically ignored labels for the cluster-api specific cloud providers from the autoscaler and replacing that with documentation to instruct me on how to use the ignored label flag would help me to have a more explicit usage of the autoscaler.

Background

Based on the discussion with the upstream cluster-api community about how the autoscaler should handle the various cloud specific zone labels, we should remove all the nodegroupset processors for cluster-api. in their place we should create documentation that instructs the user about how to ignore labels and some of the common labels we know about.

this video recording gives more context
 

Steps

• remove nodegroupset processors for clusterapi
• add documentation about how to use the ignore label flag
• add documentation about known labels

Stakeholders

• cluster api community
• cloud team

Definition of Done

• the autoscaler does not automatically ignore any labels when balancing on cluster-api

• Docs

• users can see how to use the --balancing-ignore-label flag and have a sample of known labels to ignore

• Testing

• this shouldn't require extra testing on its own as it is removing automatic behavior that is replicated through flags.
• we should not bring this change back into our ci until we have also updated the CAO to pass the ignore labels as it will break our tests for balancing similar node groups.