Fix sfam 2793

simplyblock_core/controllers/lvol_controller.py

@@ -930,14 +930,6 @@ def _create_bdev_stack(lvol, snode, is_primary=True):
         type = bdev['type']
         name = bdev['name']
         params = bdev['params']
-        # Idempotency probe per stack bdev. A by-name bdev_get_bdevs resolves
-        # names, aliases and uuids server-side and returns [] / an -ENODEV
-        # error (→ None) when absent — equivalent to the previous full-dump
-        # membership test, but O(1) instead of serializing every bdev on the
-        # node into the response (the dump grows with lvol count and was the
-        # single largest cost of mass creates).
-        if rpc_client.get_bdevs(name):
-            continue
 
         ret = None
         if type == "bmap_init":
@@ -1043,12 +1035,50 @@ def _fail_after_bdev(lvol, rpc_client, msg):
     the SPDK clone-blob in place, which then blocks the parent snapshot delete
     with "vbdev_lvol_destroy: ... has N clones". Logs but does not raise on
     rollback failure so the caller still sees the original error."""
-    try:
-        _remove_bdev_stack(lvol.bdev_stack[::-1], rpc_client)
-        lvol.status = LVol.STATUS_IN_DELETION
-        lvol.write_to_db(DBController().kv_store)
-    except Exception:
-        logger.exception("rollback of bdev stack failed for %s", lvol.get_id())
+    db_controller = DBController()
+    host_node = db_controller.get_storage_node_by_id(lvol.node_id)
+    secondary_ids = [host_node.secondary_node_id]
+    if host_node.tertiary_node_id:
+        secondary_ids.append(host_node.tertiary_node_id)
+
+    all_nodes = [host_node]
+    for sid in secondary_ids:
+        try:
+            all_nodes.append(db_controller.get_storage_node_by_id(sid))
+        except KeyError:
+            logger.debug(
+                "Skipping missing secondary/tertiary node '%s' during rollback for lvol '%s'",
+                sid,
+                lvol.get_id(),
+            )
+
+    if lvol.blobid:
+        try:
+            from simplyblock_core.storage_node_ops import find_leader_with_failover
+            primary_node, non_leaders = find_leader_with_failover(all_nodes, lvol.lvs_name)
+            if primary_node is None:
+                msg = "No leader available, marking it for deletion and let lvol monitor handle deletion"
+            else:
+                existing_on_primary = primary_node.rpc_client().get_bdevs(f"{lvol.lvs_name}/{lvol.lvol_bdev}")
+                if not existing_on_primary:
+                    for sn_id in lvol.nodes:
+                        if sn_id != lvol.node_id:
+                            rpc_client = db_controller.get_storage_node_by_id(sn_id).rpc_client()
+                            existing_on_secondary = rpc_client.get_bdevs(f"{lvol.lvs_name}/{lvol.lvol_bdev}")
+                            if existing_on_secondary:
+                                # register on primary
+                                rpc_client.bdev_lvol_register(
+                                    existing_on_secondary[0], lvol.lvs_name, lvol.lvol_uuid, lvol.blobid, lvol.lvol_priority_class)
+
+            lvol.status = LVol.STATUS_IN_DELETION
+            lvol.deletion_status = ""
+            lvol.write_to_db(DBController().kv_store)
+        except Exception:
+            logger.exception("rollback of bdev stack failed for %s", lvol.get_id())
+    else:
+        for node in all_nodes:
+            _remove_lvol_subsys_from_node(lvol, node.rpc_client())
+        lvol.remove_from_db(db_controller.kv_store)
     return False, msg
 
 

simplyblock_core/env_var

@@ -1,5 +1,5 @@
 SIMPLY_BLOCK_COMMAND_NAME=sbcli-dev
 SIMPLY_BLOCK_VERSION=19.2.34
 
-SIMPLY_BLOCK_DOCKER_IMAGE=public.ecr.aws/simply-block/simplyblock:main
+SIMPLY_BLOCK_DOCKER_IMAGE=public.ecr.aws/simply-block/simplyblock:main-sfam-2793
 SIMPLY_BLOCK_SPDK_ULTRA_IMAGE=public.ecr.aws/simply-block/ultra:main-latest

simplyblock_core/storage_node_ops.py

@@ -6988,8 +6988,10 @@ def check_non_leader_for_operation(node_id, lvs_name, operation_type="create",
                  StorageNode.RESTART_PHASE_POST_UNBLOCK):
         return "queue"
 
-    # 3. Fabric is connected — check RPC responsiveness
-    if _is_node_rpc_responsive(node, lvs_name):
+    # 3. Fabric is connected — check the SPDK process
+    nodeapi = SNodeClient(node)
+    is_up, _ = nodeapi.spdk_process_is_up(node.rpc_port, node.cluster_id)
+    if is_up:
         return "proceed"
 
     # 4. RPC failing but fabric connected