Open schlichtanders opened 1 month ago
is there a k3s 1.30 autoscaler alternative which works with kube-hetzner?
also why is upstream not compatible with kube-hetzner. Is this intended?
Using
cluster_autoscaler_image = "ghcr.io/kube-hetzner/autoscaler/cluster-autoscaler"
cluster_autoscaler_version = "20240227"
instead does not show this error, but also is not successful in creating new nodes on hetzner. It just does not appear. Maybe it is because of a mismatch in k3s versions - I'd like to upgrade to 1.30
@Silvest89 You, are the autoscaler king man, any ideas on this? 🙏
@schlichtanders Have you tried messing around with the name?
I see that it is ca-${var.default_location}-cax41
what about changing it to just ca-cax41 to see if it works. This is actually hetzner giving error on the name input.
When I reverted to k3s 1.29 and the hetzner autoscaler branch, it works without me changing the name. Hence I think this is not the issue.
I am currently back to k3s, however because of the validating admission policy feature which got stable in 1.30, I really like to go to 1.30.
Will report back if I tried it again on 1.30. Thanks for the hint
Do you have a commit/tag of cluster-autoscaler that was still working for you?
You can enable debug logs for the Hetzner Cloud API calls by raising the verbosity to 5
: --v=5
. That will show the actual name being sent to the API.
instead does not show this error, but also is not successful in creating new nodes on hetzner.
I think this happens because we changed the error reporting in https://github.com/kubernetes/autoscaler/pull/6750. Previously it just ignored any errors while creating new nodes. Now it returns the error to cluster-autoscaler and it is being shown.
This was released in 1.28.6, 1.29.4 and 1.30.2.
Description
I've set
which gives the following error in the logs:
Kube.tf file
Screenshots
No response
Platform
Linux