Skip to content

scripts: drop clusterapi step while k3s disabled#2927

Merged
berendt merged 1 commit into
mainfrom
disable-clusterapi-step
Jul 9, 2026
Merged

scripts: drop clusterapi step while k3s disabled#2927
berendt merged 1 commit into
mainfrom
disable-clusterapi-step

Conversation

@ideaship

@ideaship ideaship commented Jul 9, 2026

Copy link
Copy Markdown
Contributor

Problem

testbed-update-stable-current-ubuntu-24.04 started failing on the
2026-07-09 midnight run (the first run after the commit below merged).
The Deploy services task fails rc=2 in 510-clusterapi.sh: the
cert_manager role's first task Deploy cert-manager crds aborts with

Could not find or access '/share/kubeconfig' on the Ansible Controller.

/share/kubeconfig is the manager-local k3s / cluster-api management
cluster kubeconfig, produced by scripts/deploy/500-kubernetes.sh.

Root cause

Incomplete removal in:

That commit disabled the k3s tech preview by dropping the
500-kubernetes.sh invocation from deploy-services.sh and
upgrade-services.sh, but left the 510-clusterapi.sh invocation in
place. 510-clusterapi.sh runs osism apply clusterapi and the
cert_manager play, both of which need the k3s cluster and its
/share/kubeconfig. With k3s no longer deployed the kubeconfig is never
written, so cert_manager fails on every run — this is persistent, not a
flake.

Fix

Remove the 510-clusterapi.sh invocation from both deploy-services.sh
and upgrade-services.sh so the k8s tech preview is fully disabled as
intended. It can be restored together with 500-kubernetes.sh once the
k3s upgrade path is fixed.

🤖 Generated with Claude Code

Commit 75b9d6c temporarily disabled the k3s tech preview by removing
the 500-kubernetes.sh invocation from deploy-services.sh and
upgrade-services.sh, but left the 510-clusterapi.sh invocation in
place. That step runs "osism apply clusterapi" and the cert_manager
play, both of which require the manager-local k3s / cluster-api
management cluster and its kubeconfig at /share/kubeconfig, which
500-kubernetes.sh is what produces.

With k3s no longer deployed the kubeconfig is never written, so the
cert_manager role's first task "Deploy cert-manager crds" aborts and
fails the Deploy services / Upgrade services step (rc=2):

  Could not find or access '/share/kubeconfig' on the Ansible
  Controller.

This first hit testbed-update-stable-current-ubuntu-24.04 on the
2026-07-09 midnight run (the first run after 75b9d6c merged) and is
persistent, not transient. Remove the clusterapi invocation from both
scripts so the k8s tech preview is fully disabled as intended; it can
be restored together with 500-kubernetes.sh once the k3s upgrade path
is fixed.

Assisted-by: Claude:claude-opus-4-8
Signed-off-by: Roger Luethi <luethi@osism.tech>
@ideaship ideaship marked this pull request as ready for review July 9, 2026 05:01
@ideaship ideaship requested a review from berendt July 9, 2026 05:01
@berendt berendt merged commit ec7c712 into main Jul 9, 2026
3 checks passed
@berendt berendt deleted the disable-clusterapi-step branch July 9, 2026 05:09
@github-project-automation github-project-automation Bot moved this from New to Done in Human Board Jul 9, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: Done

Development

Successfully merging this pull request may close these issues.

3 participants