Cluster autoscaler appears unable to scale nodes down. 

In cluster-autoscaler-default pod logs, we can see that the nodes are tainted for deletion and then repeatedly see "Skipping {node} from delete consideration - the node is currently being deleted" followed by "Nodegroup is nil for azure:///subscriptions/blah/blah/samenodename"

Customer tried draining the nodes in question to give the autoscaler a hand and to see if any stray pods holding on to storage on the nodes or something like that but no change.

According to customer, this morning (after several days of issues and reportedly without any change by customer), autoscaler worked fine and was successful in scaling down nodes.

More logs / info to follow and in support case 03528995

1.
2.
3.