Legacy OpenShift installs can stall at the last SDN-to-OVN migration phase

Red Hat is investigating a failure mode in OpenShift SDN-to-OVN migrations where the node rollout completes but the network transition never clears its final cleanup phase. The Customer Portal solution, updated Sept. 20 and still marked “solution in progress,” says Network.config/cluster remains with NetworkTypeMigrationOriginalCNIPurged=False while an ovs DaemonSet remains in the openshift-sdn namespace.

Where the migration stalls

The reported stall occurs after MachineConfigPools have rolled out the new OVN configuration. In other words, seeing updated pools does not prove that the migration controller has removed the original OpenShift SDN assets. The authoritative completion signal is the network configuration’s migration condition, not the node rollout alone.

Red Hat scopes the documented environment to OpenShift 4.x clusters originally installed on version 4.6 or earlier. That origin detail is easy to lose after years of upgrades, so platform teams should recover it from cluster records before scheduling the network change.

Operational blast radius

A cluster in this state has progressed far enough to require caution but has not reached the declared end state. The lingering ovs DaemonSet and false purge condition indicate that cleanup is incomplete. Operators should avoid treating those objects as ordinary leftovers or deleting them by hand: Red Hat has not published the recovery procedure in the public portion of the in-progress solution.

Preflight checks

Before migration, record the cluster’s original installation version, current network type, migration conditions, MachineConfigPool state and the contents of openshift-sdn. Confirm that the support team can access the full Red Hat solution or open a case if the old-install condition applies.

During the change, use explicit gates:

1. Wait for each MachineConfigPool to finish updating.
2. Re-read Network.config/cluster and inspect every migration condition.
3. Check whether the openshift-sdn namespace still contains the ovs DaemonSet.
4. Verify workload connectivity before moving from observation to cleanup.

Recovery sequence

If NetworkTypeMigrationOriginalCNIPurged remains false, freeze further cleanup and collect the network resource, operator conditions, relevant controller logs and the surviving DaemonSet definition. Preserve that evidence before restarting components or deleting objects.

Because Red Hat labels the solution in progress and hides the resolution behind subscription access, the supported next step is to follow the current portal procedure or escalate through support. The safe sequence is diagnose, preserve state, obtain the supported remediation, then verify that the purge condition turns true. Manual deletion first would erase a symptom without proving that the migration controller has reconciled the cluster.