diff --git a/e2e/keywords/k8s.resource b/e2e/keywords/k8s.resource
index 0b7c18a7a..a68157591 100644
--- a/e2e/keywords/k8s.resource
+++ b/e2e/keywords/k8s.resource
@@ -31,6 +31,9 @@ Delete volume of ${workload_kind} ${workload_id} replica node
 
 Add deleted node back
     reboot_node_by_name    ${deleted_node}
+    ${host_provider}=    Get Environment Variable    HOST_PROVIDER
+    ${disk_path}=    Set Variable If    "${host_provider}" == "harvester"    /dev/vdc    /dev/xvdh
+    add_disk    block-disk    ${deleted_node}    block    ${disk_path}
 
 Force drain volume of ${workload_kind} ${workload_id} volume node
     ${workload_name} =   generate_name_with_suffix    ${workload_kind}    ${workload_id}
diff --git a/e2e/libs/node/node.py b/e2e/libs/node/node.py
index a45f15554..fd2238a5e 100644
--- a/e2e/libs/node/node.py
+++ b/e2e/libs/node/node.py
@@ -62,10 +62,19 @@ def wait_for_disk_update(self, node_name, disk_num):
         assert len(node.disks) == disk_num, f"Waiting for node {node_name} disk updated to {disk_num} failed: {disks}"
 
     def add_disk(self, node_name, disk):
-        node = get_longhorn_client().by_id_node(node_name)
-        disks = node.disks
-        disks.update(disk)
-        self.update_disks(node_name, disks)
+        added = False
+        for i in range(self.retry_count):
+            try:
+                node = get_longhorn_client().by_id_node(node_name)
+                disks = node.disks
+                disks.update(disk)
+                self.update_disks(node_name, disks)
+                added = True
+                break
+            except Exception as e:
+                logging(f"Adding disk {disk} to node {node_name} error: {e}")
+            time.sleep(self.retry_interval)
+        assert added, f"Adding disk {disk} to node {node_name} failed"
 
     def reset_disks(self, node_name):
         node = get_longhorn_client().by_id_node(node_name)