diff --git a/internal/controller/nnf_access_controller.go b/internal/controller/nnf_access_controller.go
index b7b335b43..b2c4240a0 100644
--- a/internal/controller/nnf_access_controller.go
+++ b/internal/controller/nnf_access_controller.go
@@ -625,8 +625,24 @@ func (r *NnfAccessReconciler) mapClientLocalStorage(ctx context.Context, access
 				// If no ClientReference exists, then the mounts are for the Rabbit nodes. Use references
 				// to the NnfNodeStorage resource so the client mounter can access the swordfish objects
 				if access.Spec.ClientReference == (corev1.ObjectReference{}) {
+					// For rabbit mounts, use unique index mount directories that consist of
+					// <namespace>-<index>. These unique directories guard against potential data
+					// loss when doing copy out data movement operations. Having the namespace
+					// (rabbit name) included in the mount path ensures that these individual
+					// compute mount points are always unique.
+					//
+					// Ex: /mnt/nnf/7b5dda61-9d91-4b50-a0d3-f863d0aac25b-0/rabbit-node-1-0
+					//     /mnt/nnf/7b5dda61-9d91-4b50-a0d3-f863d0aac25b-0/rabbit-node-2-0
+					// If you did not include the namespace, then the paths would be:
+					//     /mnt/nnf/7b5dda61-9d91-4b50-a0d3-f863d0aac25b-0/0
+					//     /mnt/nnf/7b5dda61-9d91-4b50-a0d3-f863d0aac25b-0/0
+					// When data movement copies the MountPathPrefix to global lustre, then the
+					// contents of these directories are merged together and potential data loss can
+					// occur if the contents do not have unique filenames.
+					indexMountDir := fmt.Sprintf("%s-%s", nnfNodeStorage.Namespace, strconv.Itoa(i))
+					mountInfo.MountPath = filepath.Join(access.Spec.MountPathPrefix, indexMountDir)
+
 					mountInfo.Device.Type = dwsv1alpha2.ClientMountDeviceTypeReference
-					mountInfo.MountPath = filepath.Join(access.Spec.MountPathPrefix, strconv.Itoa(i))
 				} else {
 					mountInfo.MountPath = access.Spec.MountPath
 					mountInfo.Device.Type = dwsv1alpha2.ClientMountDeviceTypeLVM