Fix Longhorn default StorageClass to actually use 2 replicas

apps/longhorn/values.yaml

@@ -14,8 +14,17 @@ longhorn:
     createDefaultDiskLabeledNodes: true
     # There are two independent storage workers, one on Gimli and one on
     # Titus. Three replicas could never schedule; two keeps one complete copy
-    # on each host and remains readable if either worker fails.
+    # on each host and remains readable if either worker fails. This setting
+    # only affects volumes created through Longhorn's own UI—Kubernetes PVCs
+    # actually read numberOfReplicas from the StorageClass below instead, so
+    # both must be kept in sync.
     defaultReplicaCount: 2
+  # The default StorageClass Longhorn's chart creates reads numberOfReplicas
+  # from here, not from defaultSettings above. Every PVC that doesn't name a
+  # storageClassName goes through this class, so this is the value that
+  # actually governs replica count for real workloads.
+  persistence:
+    defaultClassReplicaCount: 2
     # Longhorn's own headroom settings, kept conservative: stop scheduling
     # new replicas onto a disk once free space drops below these
     # thresholds, so the cluster degrades gracefully instead of actually