chore(github-runner): un-park github-runner-sharedpos (Shared.Pos non-root build fixed)
Set replicas back to 1 — the non-root setup-dotnet path issue that drove the park (replicas: 0 on 2026-05-21) is resolved. Shared.Pos #5 landed a step-scoped DOTNET_INSTALL_DIR override and today's Shared.Pos publish succeeded at 01:47 UTC against this runner. Reconciles git with live K8s state: the github-runner-sharedpos pod is already Running on rke2-agent2 (4h12m up, 4 restarts). Without this commit, the next ArgoCD selfHeal would scale it back to 0 and re-stop Shared.Pos publish flow. Follow-on: Sprint 30+ Cx-1 fleet-wide pod-env (DOTNET_INSTALL_DIR set on the Deployment env, not per-workflow) is still queued — once landed, the per-workflow patch can be backed out. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
@@ -233,13 +233,9 @@ metadata:
|
||||
app.kubernetes.io/managed-by: argocd
|
||||
flowercore.io/created-by: argocd
|
||||
spec:
|
||||
# PARKED 2026-05-21: scaled to 0. The Shared.Pos build still fails on the
|
||||
# non-root runner (actions/setup-dotnet -> /usr/share/dotnet denied); a
|
||||
# churning ephemeral runner just retries a doomed build and drives HighCPU
|
||||
# on rke2-agent2. Re-enable (replicas: 1) only as part of the Sprint 30+
|
||||
# Cx-1 Linux-runner-fleet lane, which sets DOTNET_INSTALL_DIR on the runner
|
||||
# pod env fleet-wide. See memory/feedback_flip_sweep_nonroot_runner_workflow_fallout.
|
||||
replicas: 0
|
||||
# UN-PARKED 2026-05-21: Shared.Pos #5 fixed the non-root setup-dotnet path
|
||||
# (DOTNET_INSTALL_DIR step-scoped). Re-enabled to run the now-fixable build.
|
||||
replicas: 1
|
||||
selector:
|
||||
matchLabels:
|
||||
app.kubernetes.io/name: github-runner-sharedpos
|
||||
|
||||
Reference in New Issue
Block a user