diff --git a/deploy/helm/kubestellar-console/values.yaml b/deploy/helm/kubestellar-console/values.yaml
index 3eec9a76ef..44d9af9d1e 100644
--- a/deploy/helm/kubestellar-console/values.yaml
+++ b/deploy/helm/kubestellar-console/values.yaml
@@ -186,13 +186,17 @@ resources:
 # an early "Progress deadline exceeded" rollback before Helm times out.
 deployment:
   progressDeadlineSeconds: 900
-  # Deployment rollout strategy. Use Recreate for single-replica deployments
-  # backed by a ReadWriteOnce PVC (e.g. kc-live on block storage) to avoid
-  # the RollingUpdate deadlock where the new pod cannot mount the RWO volume
-  # while the old pod still holds it.
+  # Deployment rollout strategy. Recreate is required for single-replica
+  # deployments backed by a ReadWriteOnce PVC (e.g. kc-live on block storage,
+  # such as the vllm-d cluster's ibmc-vpc-block-10iops-tier storage class) to
+  # avoid the RollingUpdate deadlock where the new pod cannot mount the RWO
+  # volume while the old pod still holds it, which manifests as the
+  # Deployment getting stuck with "Pending termination" and the CSI driver
+  # repeatedly failing to mount with "an operation with the given Volume ID
+  # ... already exists".
   # Values: RollingUpdate | Recreate
   strategy:
-    type: RollingUpdate
+    type: Recreate
     rollingUpdate:
       maxSurge: 1
       maxUnavailable: 0