User Story

As a user I want to ensure that scale from zero cluster autoscaling works well when using the upstream scaling hint annotations so that I can follow the community best practices. Having the cluster autoscaler operator monitor the scale from zero annotations, and correct them when incorrect, will confirm the correct behavior.

Background

As part of migrating the OpenShift scale from zero annotations to use the upstream annotations keys, the cluster autoscaler operator should be updated to look for these annotations on MachineSets that it is monitoring.

Currently, we use annotations with prefix "machinie.openshift.io", in the upstream the prefix is "capacity.cluster-autoscaler.kubernetes.io". The CAO should be updated to recognize when a MachineSet has either set of annotations, and then ensure that both sets exist.

Adding both sets of annotations will help us during the transition to using the upstream set, and will also ensure backward compatibility with our published API.

Please note that care must be taken with the suffixes as well. Some of the OpenShift suffixes are different from upstream, and in specific the memory suffix uses a different type of calculation. As we convert our autoscaler implementation to use the upstream annotations we must make sure that any conversions will conform to upstream.

<Describes the context or background related to this story>

Steps

• Add ability to check annotations on MachineSets
• Add conversion/update to scale from zero annotations
• Add unit tests to confirm the proper behavior

Stakeholders

• openshift cloud team

Definition of Done

• CAO is properly adding both sets of scale from zero annotations

• Docs

• developer docs in the CAO repo will need updating
• if there are product docs around the CAO, they should be updated as well

• Testing

• the CAO needs full unit testing to confirm the behavior described above