Appendices

Full logs, raw output, and troubleshooting notes for the OCPBUGS-88531 test verification.


A. Test Environment

Management ClusterAWS (via KUBECONFIG=/Users/brcox/aws_dev_kubeconfig)
HostedClusterbrcox-sm-dev-hc (namespace: clusters)
HCP Namespaceclusters-brcox-sm-dev-hc
PlatformAzure
OCP Version4.22.0-rc.3
CNO Image(fill in after image build completes)
PR#3030
JIRAOCPBUGS-88531

B. CNO Operator Logs


C. Pod State Snapshots


D. Deployment Pod Template Annotations

# Command used:
# for deploy in cloud-network-config-controller multus-admission-controller network-node-identity ovnkube-control-plane; do
#   echo "=== $deploy ==="
#   oc get deploy -n clusters-brcox-sm-dev-hc $deploy -o jsonpath='{.spec.template.metadata.annotations}' | jq .
# done


E. HostedControlPlane CR (annotations excerpt)

# oc get hcp -n clusters-brcox-sm-dev-hc brcox-sm-dev-hc -o jsonpath='{.metadata.annotations}' | jq .


F. Troubleshooting Notes

CNO Reconcile Delay

CNO uses a 3-minute ResyncPeriod. After setting the restart-date annotation on the HostedCluster, expect up to 3 minutes before CNO picks up the change and triggers a rollout.

Server-Side Apply Idempotency

CNO uses deterministic manifests with server-side apply. If the restart-date value is unchanged between reconciles, the pod template annotation stays identical, so Kubernetes does not trigger a new rollout. This is the expected idempotent behavior tested in Scenario 4.

CNCC Only on Cloud Platforms

The cloud-network-config-controller is only deployed on AWS, Azure, GCP, and OpenStack with OVN networking. On KubeVirt or None platform, there is no CNCC pod to restart, but SetRestartDateAnnotation still processes all rendered objects — there are simply no CNCC manifests to annotate.