E2E CI runner service account lacks RBAC permissions #86
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 CI workflow adds an
Apply RBACstep that runskubectl apply -f deployment/e2e-ci-runner-rbac.yaml. However, the ARC runners service account lacks permissions to GET roles/rolebindings in the namespace.kubectl applyperforms a GET first to check if the resource exists before deciding between create/update. Without GET permissions, the command fails with Forbidden.Reproduction
Run the E2E workflow on PR #85 — it fails at the Apply RBAC step.
Required Fix
The runner ClusterRole needs permissions:
This is an infra issue affecting the ARC runner configuration. Per org policy, CI/CD infra changes must go through Hugh Hackman.
Impact
Product Triage
This is a symptom of the same root cause tracked in #79 and #87: Flux hasn't been bootstrapped to apply CI runner RBAC to the cluster.
The runner can't self-apply RBAC because it doesn't have
getonroles/rolebindings— and it can't get those permissions without someone applying the RBAC first (circular dependency documented in #87).Resolution path: Cluster admin applies the Flux bootstrap manifests (one-time
kubectl apply). Once Flux reconciles, all RBAC is applied and E2E unblocks.Closing as duplicate of #87, which has the complete fix instructions. #79 remains the parent tracking issue for E2E failures on main.
Duplicate of #87 (same root cause). #79 is the parent tracking issue.