kubermatic / machine-controller

Apache License 2.0
297 stars 124 forks source link

Future API group conflict with upstream CAPI #1780

Open schlakob opened 2 months ago

schlakob commented 2 months ago

Hi,

we are currently using the machine-controller as an addon of Kubeone, in its default configuration.

I was wondering why the CRDs of CAPI (machinedeployments, machinesets, machines) are of the cluster.k8s.io api group and not of the upstream CAPI cluster.x-k8s.io api group? I noticed, that CAPI switched from k8s.io to x-k8s.io a while ago.

Are there plans using the same api group as upstream CAPI in future?

For us this is important, because we are thinking of deploying a CAPI in the Kubeone cluster running the machine-controller, therefore we would like to ensure, that no conflicts regarding the CRDs occur.

kron4eg commented 2 months ago

Hello. Because up until now we preferred to stay on older CAPI (v1alpha1) version. The rumors say maybe we want to update, but that's not currently on the table.

schlakob commented 2 months ago

This would then result in conflicts when running a CAPI and machine-controller in the same cluster. Is or will there be a way to overwrite the group for the machine-controller.

embik commented 1 week ago

If the Kubermatic stack ever upgrades to newer CAPI versions and adopts newer API groups, it would probably integrate into CAPI as a set of providers. I don't think this would really create a conflict since CAPI is pluggable by design.