CAPI autoscaler does not support per NodeGroup autoscaling options

[mweibel]

Which component are you using?:
cluster-autoscaler

Is your feature request designed to solve a problem? If so describe the problem this feature should solve.:
Different NodeGroups have different scaling up/down behavior and timings. The clusterapi provider doesn't provide an option to configure this per NodeGroup but instead relies on the autoscaling defaults which are populated using cluster-autoscaler flags.

In other cloudprovider implementation this dilemma is solved by parsing specialized annotations on each NodeGroup and using those for overriding the global configuration values.

Describe the solution you'd like.:
The same implementation as e.g. cloud provider Azure or GCE has: Annotations on each type of NodeGroup the clusterapi provides (MachinePool/-Set/-Deployment).

Describe any alternative solutions you've considered.:
Certain properties might be covered by existing infra provider implementations. There might be a way to introduce certain kind of configuration like that. However that would mean encoding autoscaling behavior specific to cluster-autoscaler in a place where it doesn't belong to.

[mweibel]

/label area/provider/cluster-api

[k8s-ci-robot]

@mweibel: The label(s) /label area/provider/cluster-api cannot be applied. These labels are supported: api-review, tide/merge-method-merge, tide/merge-method-rebase, tide/merge-method-squash, team/katacoda, refactor. Is this label configured under labels -> additional_labels or labels -> restricted_labels in plugin.yaml?

In response to this:

/label area/provider/cluster-api

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[elmiko]

/area provider/cluster-api