Pause the operator only when you need a short-lived manual window.

Pausing tells the operator to stop normal reconciliation for a specific cluster while you inspect or repair it. Use it for deliberate tactical work, not as a substitute for recovery workflows or as a long-term steady state.

Use this control when

• you need the operator to stop applying changes during a short repair window
• you are debugging a live cluster and do not want reconciliation to race your inspection
• you plan to resume normal management after the manual intervention finishes
• the cluster is not already in a dedicated safe-mode or recovery workflow

Decision matrix

What pausing changes

What pausing changes.

Surface	What happens while paused	What does not change
Managed resources	The operator stops reconciling StatefulSets, Services, ConfigMaps, Secrets, and similar managed objects.	Kubernetes native controllers such as StatefulSet behavior still exist and may react independently.
Self-healing	The operator does not actively repair drift or push the workload back to the desired spec.	A broken workload does not become safe just because reconciliation stopped.
Deletion	Finalizer and deletion paths can still run if the custom resource is deleted.	Pausing is not a hard freeze of every lifecycle path.

Pause the cluster

Apply

Pause reconciliation for one cluster

bash

kubectl patch openbaocluster <name> -n <namespace> --type merge -p '{
"spec": {
  "paused": true
}
}'

Use this only for the cluster you are actively working on. A paused cluster can drift further away from its desired state while the operator is quiet.

Verify that the pause is active

Verify

Check the paused flag

bash

kubectl get openbaocluster <name> -n <namespace> -o jsonpath='{.spec.paused}{"\n"}'

If updates appear to do nothing, check this field first before assuming the controller is broken.

Resume reconciliation

Apply

Resume normal reconciliation

bash

kubectl patch openbaocluster <name> -n <namespace> --type merge -p '{
"spec": {
  "paused": false
}
}'

You can also remove the field entirely. The important outcome is that the operator can reconcile the cluster again and converge any pending drift.

Pause is not safe mode

If the operator entered break-glass or safe mode because of quorum loss or another critical failure, simply flipping spec.paused back to false is not the recovery step. Use the dedicated safe-mode or recovery workflow instead.

Choose the next step

Run planned maintenanceUse the broader maintenance workflow when the change involves drains, restarts, scaling, or maintenance mode.Open safe mode recoverySwitch to the break-glass path when the cluster is no longer in an ordinary troubleshooting state.Troubleshoot the clusterGo back to the symptom-driven troubleshooting route when you still need to diagnose the failure before changing anything else.