AWS Autoscaling: How to specify min/max nodes with autodiscovery?

[trondhindenes]

From what I can see in the docs, I can either specify a "node group" (which takes min/max nodes) or use autodiscovery tags. It's not clear from the documentation how to specify min/max nodes when autodiscovery is used. It would be awesome if that was clarified.

[aleksandra-malinowska]

See usage described in #462, which added support for it on GCE.

[trondhindenes]

Does that mean AWS support was already in place, or that there isn't support for it on AWS? It's not clear to me from looking at #462.

[aleksandra-malinowska]

The feature was originally developed for AWS case. I'm not sure what's the current status, but there were several issues tracking fixes for setting min/max limits in that env. You can also try asking on sig-autoscaling slack channel

[trondhindenes]

I tried the same parameter structure (wow did I just write that) as in the PR, and that's not working at all:
--node-group-auto-discovery=asg:tag=k8s.io/cluster-autoscaler/enabled,kubernetes.io/cluster/rtv-dev-cluster02,minNodes=1,maxNodes=10
It works fine if I removed the minNodes/maxNodes thingy.

[johanneswuerbach]

If you are using autodiscovery the min/max values are taken from the discovered ASGs and can be adjusted on the fly by changing the values of the AWS ASG itself.

You can override this by manually specifying an ASG and providing min/max as autoscaler parameters. This always takes precedence over autodiscovered values.

[trondhindenes]

Thanks @johanneswuerbach . So what you're saying is that as long as auto-discovery is enabled, cluster-autoscaler will always respect the limits set in the ASG.

[johanneswuerbach]

Correct, the autoscaler is only modfiying the desired size value of an asg, never min or max.

[mf-lit]

This confused me too, PR opened.

[aleksandra-malinowska]

Closed via #1042