Open yulianedyalkova opened 1 week ago
Reproduced the issue, the control plane isn't fully working due to the controller-manager crashing at the start.
This is cause by the removal of the aws in-tree cloud provider in kubernetes 1.27: https://github.com/kubernetes/kubernetes/pull/115838
We are correctly setting the cloud-provider as external in the cluster chart, and I think we deploy the aws cloud provider as a separate pod as expected, but we are also setting the external-cloud-volume-plugin which falls back to the in-tree cloud provider, which causes the issue.
Removing that line the controller-manager works as expected. I will test the changes with a test release and see if allocating volumes works correctly. We will also need to check whether any of the other providers actually need that flag.
Merged PR in cluster chart with the fix, removing the setting by default (providers can still define it through helm values) https://github.com/giantswarm/cluster/pull/242
Tested v25, v26, v27 and v28 cluster creation and v25 updated all the way through v28, everything works as expected
As Nikola already tried to create a GS cluster with Kubernetes release v27 and hit some issues, let's try to reproduce that and investigate how we fix the problems ahead of time.
Acceptance criteria: