[Bug]: cluster-autoscaler does not wait long enough for new server to become available

[UBaggeler]

Description

Since midnight our autoscaler is having issues launching new nodes (from 1 --> 3). The creation of the new servers seems to take longer than the expected (default) timeout of 5min:

failed to create error: failed to start server hcloud-autoscaled-xyz error: timeout waiting for server hcloud-autoscaled-xyz

In this case cluster-autoscaler removes/deletes the nodes and repeatedly tries to spin up new servers, without success.

Manually increasing the timeout (for example to 15min) by setting the env variable HCLOUD_SERVER_CREATION_TIMEOUT on the autoscaler deployment resolves the issue.

Unfortunately the autoscaler.yaml.tpl file does not allow to set this environment variable.

Kube.tf file

n/a

Screenshots

No response

Platform

Linux

[zarevavasyl]

Tnx!)

[mysticaltech]

@UBaggeler Your fix was just released as part of v2.13.4 🚀