question: is it possible to run cluster-autoscaler in dry-run mode

[rajatjindal]

Hi, Its more of a question if we can run cluster-autoscaler in dry-run mode. We want to evaluate when autoscaler would have triggered the scale-down/up actions without actually taking the action (to avoid disruption during the evaluation)

Thanks
Rajat Jindal

[aleksandra-malinowska]

There's no support for this right now, and I don't think it'd be trivial change. Current logic assumes all requests for modifying cluster state are processed, so just injecting a version of cloud provider code which does nothing won't work out of the box for this - CA will enter back-offs, report node groups as unhealthy etc. You'd have to ensure all relevant parts of the logic are disabled, not just the calls to cloud provider.

Additionally, its usefulness would be limited as it can only provide a single decision at any point in time. Let's say there are 3 nodes, each of utilized at 30% and otherwise removable. We know 2 of those can be safely removed, resulting in the third being 90% utilized. However, Cluster Autoscaler:

1. Drains and deletes nodes running pods one-by-one.
2. Doesn't choose the node to drain in a deterministic way.

So the output of dry run would probably be sth like this, in random order:

• remove node 1
• remove node 2
• remove node 3
• remove node 1
  etc.

But this doesn't mean all 3 nodes would be deleted if it was enabled. Without dry run, after removing the first node, one or both of the others may become unremovable. A second node may then be removed, but the last one won't.

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle rotten

[fejta-bot]

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/close

[k8s-ci-robot]

@fejta-bot: Closing this issue.

In response to this:

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/close

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.