Starting v4.3, there was a bug identified in the codebase wherein Cluster Autoscaler deployment would be enabled by default for AWS Cloud Provider Clusters even if the same was not opted for during cluster creation.
- Platform9 Managed Kubernetes - v4.3 and Higher
- Because of the way autodiscovery functionality works for AWS Cloud Provider clusters in PMK, even if cluster autoscaler was enabled on a cluster when not opted for initially due to this issue, it will not have any impact on the way cluster functions under increased or decreased load as the numMaxWorkers attribute would be set equal to that of numMinWorkers i.e. the current worker node count from Qbert (PMK Cluster Manager) perspective. So other than having an additional deployment seen to be running in the kube-system namespace, no other impact should be seen.
- Secondly, we have already fixed this bug and have backported it to our latest v4.4 and above build releases. Thus, if you are observing the issue as described above, once you upgrade to v4.4 and beyond, you will see that the unwanted cluster autoscaler deployment has been removed from these clusters on which it wasn't opted to be running on during cluster creation.