As an HCP Karpenter management cluster admin, I want to expose events and metrics to improve observability and monitoring when deleting a hosted cluster with autonode on.

Events:

Requires a bit more research/experimentation on what could actually cause deletion to get stuck, but some very general reasons off the bat are probably:

• Node/NodeClaim cannot be deleted, and it's been stuck for a long time
• Nodepool cannot be deleted, and its been stuck for a long time

Most likely, we won't be able to know why a provisioned instance cannot be deleted (if we do know, that would be better!), so waiting an reasonably long time without updates is probably a good indicator something went wrong.

Metrics:

This story should also cover the metrics that we could probably expose in the operator when initiating a deletion.

Some common ones are probably:

• How long a Karpenter tear down is taking
• How many nodes are being deleted in the hosted cluster
• Current progress of node/nodeclaims deleted e.g. 4/10 deleted

Alerts:

There might also be potential to include emiting of OpenShift alerts when a deletion has stalled or been completed.