k8s-rwo-deployment-recreate-strategy

A single-replica Kubernetes deployment using a ReadWriteOnce PVC deadlocks during rolling updates because the new pod cannot mount the volume while the old pod holds it. The old pod often refuses to terminate due to maxUnavailable constraints.

This error occurs when patching a live Deployment that still exposes rollingUpdate fields with a Recreate strategy. You must follow a safe delete-and-recreate workflow to remove the live resource and apply the new configuration cleanly.

Set the deployment strategy to Recreate directly in your Helm values to avoid RollingUpdate conflicts during upgrades. This ensures the old pod terminates before the new pod starts, preventing PVC multi-attach deadlocks.

Using the Recreate strategy for single-replica deployments introduces downtime because the existing pod must terminate before the new pod initializes. This trade-off is necessary to release ReadWriteOnce PVCs and prevent update deadlocks.

You need a delete-and-recreate workflow when a live Kubernetes deployment exposes rollingUpdate fields that conflict with a Recreate strategy, preventing direct patching. This path safely removes the old resource to apply the new strategy configuration.