k8s version outside supported range of CAPI

Summary

main (as of Oct 24th) defaults to k8s 1.37, but the CAPI providers deployed by Sylva (capi, cabpk, capd) are still on v1.12.8, which supports up to 1.36 only:

  • Management cluster: v1.31.x -> v1.36.x
  • Workload cluster: v1.29.x -> v1.36.x

So both management and workload clusters on the default k8s version run outside the CAPI support matrix.

Nothing catches this today: k8s versions and CAPI versions are bumped independently, and nothing checks one against the other.

Proposal

We would need some kind of validation that would not block an unsupported upgrade (see #4897 (comment 4106649387)), but at least to warn if this happens.

Also this can be extended to other (bootstrap) providers where possible.

Edited by Cristian Manda