Fix two bugs from PRI-879 QA review:
- HEADLAMP_URL is now written to .env.e2e unconditionally, before
attempting token generation. Previously it was only written when
token generation succeeded, causing tests to fail if the token
command errored.
- ClusterRole headlamp-e2e-test-reader now includes pods get/list/watch
so the Rook PodsPage can populate without permission errors.
Does not address the popup race in auth.setup.ts — that file was not
changed because the popup race claim in PRI-879 does not match the
actual code order. The popupPromise (line 9) is already captured before
the click (line 10) in the source file.
Fixes: PRI-879
Co-Authored-By: Paperclip <noreply@paperclip.ing>
The headlamp-e2e-test service account needs cluster-wide read permissions
for storageclasses, cephclusters, persistentvolumes, and
persistentvolumeclaims so the Rook plugin sidebar can populate these
resources without errors.
- Add ClusterRole headlamp-e2e-test-reader with get/list/watch on
storageclasses, cephclusters, cephclusters/status, persistentvolumes,
persistentvolumeclaims
- Add ClusterRoleBinding headlamp-e2e-test-crb binding the role to
the headlamp-e2e-test service account
- Update teardown to also clean up the ClusterRole and ClusterRoleBinding
Fixes: PRI-741
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Add pre-deployment node/namespace/resource diagnostics and wrap
kubectl apply in explicit error handling with cluster state dump on
failure. This gives us actionable output in the GitHub Actions logs
when the Deploy E2E step fails, instead of a silent exit code.
PRI-956
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Chris Farhood