E2E Tests failing on main: runner SA lacks PVC access in kube-system #77
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Problem
The E2E Tests workflow is failing on
mainafter PR #76 merged. The ARC runner SA (system:serviceaccount:arc-runners:runners-privilegedescalation-gha-rs-no-permission) does not have permission to access PersistentVolumeClaims in thekube-systemnamespace.Error:
Run: https://github.com/privilegedescalation/headlamp-polaris-plugin/actions/runs/23344829743
Root cause
The current E2E tests on
mainusekube-systemand require RBAC permissions that the runner SA does not have. This is the known issue that PR #73 (E2E redesign) addresses by moving to a dedicatedheadlamp-e2enamespace.Fix
Merge PR #73 (refactor: redesign E2E to use ConfigMap volume mount with stock Headlamp image) — this redesign avoids
kube-systementirely.The
headlamp-e2enamespace already exists in the cluster. The RBAC for the runner SA inheadlamp-e2eis pending cluster admin action (tracked separately).Workaround (temporary)
The E2E workflow could be disabled on
mainuntil #73 merges, but the preferred fix is merging #73.This is resolved by PR #73 which redesigns E2E to use a ConfigMap volume mount in a dedicated
headlamp-e2enamespace instead of PVC access inkube-system. Once #73 merges and the namespace RBAC is applied (tracked in PRI-420/PRI-404), E2E tests will no longer need PVC permissions.Keeping open until #73 is merged and validated.
QA Note\n\nVerified: E2E is indeed failing on with the reported error — runner SA cannot access PVCs in . This is a real CI regression.\n\nFix pathway: PR #73 (E2E redesign using ConfigMap) is the intended fix — it avoids entirely. However:\n- PR #73 is currently BEHIND main and needs rebase\n- E2E CI on PR #73 also fails (missing namespace — tracked in PRI-385)\n- My review on PR #73 is still a COMMENT (not APPROVE) because E2E is not green\n\nImmediate action needed: Cluster admin must create namespace and apply RBAC (PRI-385). Once that is done:\n1. E2E will pass on PR #73 → I can APPROVE → Nancy can merge\n2. Main branch E2E can be updated to use the new approach\n\nFlagging to Nancy for prioritization.
PR #73 has been merged, which redesigns E2E to use ConfigMap volume mount with a stock Headlamp image in a dedicated
headlamp-e2enamespace. The PVC-based approach inkube-systemis removed.This issue's root cause (runner SA lacking PVC access in kube-system) is no longer relevant. E2E will work once the cluster admin applies the new RBAC per #78.
Closing — tracked by #78 going forward.
Closing: this PR is stale — the branch has already been merged (head SHA = base SHA). No changes remain.