Vcluster 0.21 seems to not respect --kubernetes-version

[felipecrs]

What happened?

If I try to create a vcluster with --kubernetes-version 1.27 in a host cluster running K8s 1.30.4, the --kubernetes-version 1.27 flag seems ignored. Nothing is shown in the logs either. The final vcluster k8s version ends up as 1.30.4.

What did you expect to happen?

This was working fine on 0.20. I was expecting it to work the same?

How can we reproduce it (as minimally and precisely as possible)?

$ k3d cluster create # right now it creates a 1.30.4 cluster

$ vcluster create --kubernetes-version 1.27 felipe
17:02:45 info Creating namespace vcluster-felipe
17:02:45 info Chart not embedded: "open chart/vcluster-0.21.1.tgz: file does not exist", pulling from helm repository.
17:02:45 info Create vcluster felipe...
17:02:45 info execute command: helm upgrade felipe https://charts.loft.sh/charts/vcluster-0.21.1.tgz --create-namespace --kubeconfig /tmp/899677600 --namespace vcluster-felipe --install --repository-config='' --values /tmp/3145409311
17:02:47 done Successfully created virtual cluster felipe in namespace vcluster-felipe
17:02:55 info Waiting for vcluster to come up...
17:03:04 info vcluster is waiting, because vcluster pod felipe-0 has status: Init:1/3
17:03:14 info vcluster is waiting, because vcluster pod felipe-0 has status: Init:2/3
17:03:34 done vCluster is up and running
17:03:34 info Starting background proxy container...
17:03:46 done Switched active kube context to vcluster_felipe_vcluster-felipe_k3d-k3s-default
- Use `vcluster disconnect` to return to your previous kube context
- Use `kubectl get namespaces` to access the vcluster

$ vcluster connect felipe -- kubectl version
17:05:08 done vCluster is up and running
Client Version: v1.31.3
Kustomize Version: v5.4.2
Server Version: v1.30.4

Anything else we need to know?

No response

Host cluster Kubernetes version

$ kubectl version
Client Version: v1.31.3
Kustomize Version: v5.4.2
Server Version: v1.30.4+k3s1
Kubecolor Version: 0.4.0

vcluster version

$ vcluster --version
vcluster version 0.21.1

VCluster Config

N/A

[cbron]

Confirmed with both 1.27 and v1.29. Thanks @felipecrs for the report.

[deniseschannon]

@felipecrs For v0.20+, we should be using the vcluster.yaml (not CLI flags) to set the k8s version.

This is only supported for the k8s distro as of today.
https://www.vcluster.com/docs/vcluster/configure/vcluster-yaml/control-plane/components/distro/k8s

controlPlane:
  distro:
    k8s:
      # Version specifies k8s components (scheduler, kube-controller-manager & apiserver) version.
      # It is a shortcut for controlPlane.distro.k8s.apiServer.image.tag,
      # controlPlane.distro.k8s.controllerManager.image.tag and
      # controlPlane.distro.k8s.scheduler.image.tag
      # If e.g. controlPlane.distro.k8s.version is set to v1.30.1 and
      # controlPlane.distro.k8s.scheduler.image.tag
      # (or controlPlane.distro.k8s.controllerManager.image.tag or controlPlane.distro.k8s.apiServer.image.tag)
      # is set to v1.31.0,
      # value from controlPlane.distro.k8s.(controlPlane-component).image.tag will be used
      # (where controlPlane-component is apiServer, controllerManager and scheduler).
      version: ""

[felipecrs]

Thank you, I'll try that.

I suppose you forgot to remove the --kubernetes-version CLI flag then?

https://github.com/loft-sh/vcluster/blob/main/docs/pages/cli/vcluster_create.md#L43

[deniseschannon]

Yes, we most likely forgot to deprecate/remove it and it's something we are following up on.

Please let me know if you have any issues using the version in the vcluster.yaml.

[felipecrs]

It worked. :)