From dc51a4271b6b328fdd655d48559f7220efcf794b Mon Sep 17 00:00:00 2001
From: Julien Le Jeune <julien.le-jeune@ovhcloud.com>
Date: Fri, 12 Sep 2025 10:15:32 +0200
Subject: [PATCH] nova-conductor puts instance in error state

Nova-conductor puts instance in error if an unknown exception is raised
in the _build_live_migrate_task during the live-migration. [1]
The exception comes from _call_livem_checks_on_host and we can see raise
exception.MigrationPreCheckError if we face to
messaging.MessagingTimeout exception for example. [2]
The function check_can_live_migrate_destination does a check also on source
host with check_can_live_migrate_source [3] and this check can also
return exceptions like MessagingTimeout and this one is not caught properly
because it's a remote "Remote error: MessagingTimeout" due to dest host try to
contact source host and this source host not reply.

[1] https://github.com/openstack/nova/blob/master/nova/conductor/manager.py#L523
[2] https://github.com/openstack/nova/blob/master/nova/conductor/tasks/live_migrate.py#L381
[3] https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L9090

Closes-Bug: #2044235
Change-Id: Ie1f96fee743c235ab35113a9ad1549a67b975839
Signed-off-by: Julien Le Jeune <julien.le-jeune@ovhcloud.com>
---
 nova/compute/manager.py                          | 16 ++++++++++++++--
 .../functional/regressions/test_bug_2044235.py   |  3 +--
 nova/tests/unit/compute/test_compute_mgr.py      |  4 ++--
 ...ce-after-rpc-issue-during-live-migration.yaml |  9 +++++++++
 4 files changed, 26 insertions(+), 6 deletions(-)
 create mode 100644 releasenotes/notes/bug-2044235-reset-instance-after-rpc-issue-during-live-migration.yaml

diff --git a/nova/compute/manager.py b/nova/compute/manager.py
index 7b7da4fd56..022128612e 100644
--- a/nova/compute/manager.py
+++ b/nova/compute/manager.py
@@ -9087,8 +9087,20 @@ class ComputeManager(manager.Manager):
         try:
             allocs = self.reportclient.get_allocations_for_consumer(
                 ctxt, instance.uuid)
-            migrate_data = self.compute_rpcapi.check_can_live_migrate_source(
-                ctxt, instance, dest_check_data)
+            try:
+                migrate_data = (
+                    self.compute_rpcapi.check_can_live_migrate_source(
+                        ctxt, instance, dest_check_data)
+                )
+            except Exception as ex:
+                msg = _("Error while check_can_live_migrate_source from "
+                        "%(src)s to host %(dest)s: %(ex_type)s %(ex)s") % {
+                            'src': instance.host,
+                            'dest': CONF.host,
+                            'ex_type': type(ex).__name__,
+                            'ex': ex
+                        }
+                raise exception.MigrationPreCheckError(msg)
             if ('src_supports_numa_live_migration' in migrate_data and
                     migrate_data.src_supports_numa_live_migration):
                 migrate_data = self._live_migration_claim(
diff --git a/nova/tests/functional/regressions/test_bug_2044235.py b/nova/tests/functional/regressions/test_bug_2044235.py
index c81847f58f..71f99c8f67 100644
--- a/nova/tests/functional/regressions/test_bug_2044235.py
+++ b/nova/tests/functional/regressions/test_bug_2044235.py
@@ -57,5 +57,4 @@ class TestMessagingTimeoutDuringLiveMigrationCheck(
                 self._live_migrate, server, "failed"
             )
 
-        # bug lp-2044235 - instance is in ERROR but it should not
-        self._wait_for_state_change(server, "ERROR")
+        self._wait_for_state_change(server, "ACTIVE")
diff --git a/nova/tests/unit/compute/test_compute_mgr.py b/nova/tests/unit/compute/test_compute_mgr.py
index e0c492fa15..fe7d236d5e 100644
--- a/nova/tests/unit/compute/test_compute_mgr.py
+++ b/nova/tests/unit/compute/test_compute_mgr.py
@@ -5233,7 +5233,7 @@ class ComputeManagerUnitTestCase(test.NoDBTestCase,
             'nova.network.neutron.API.has_port_binding_extension',
             lambda *args: True))
         self.assertRaises(
-            test.TestingException,
+            exception.MigrationPreCheckError,
             self._test_check_can_live_migrate_destination,
             do_raise=True)
 
@@ -5414,7 +5414,7 @@ class ComputeManagerUnitTestCase(test.NoDBTestCase,
                                    'check_can_live_migrate_source',
                                    side_effect=messaging.MessagingTimeout):
                 self.assertRaises(
-                    messaging.MessagingTimeout,
+                    exception.MigrationPreCheckError,
                     self.compute.check_can_live_migrate_destination,
                     self.context, instance, None, None, None, None)
         _do_test()
diff --git a/releasenotes/notes/bug-2044235-reset-instance-after-rpc-issue-during-live-migration.yaml b/releasenotes/notes/bug-2044235-reset-instance-after-rpc-issue-during-live-migration.yaml
new file mode 100644
index 0000000000..7f8743afaf
--- /dev/null
+++ b/releasenotes/notes/bug-2044235-reset-instance-after-rpc-issue-during-live-migration.yaml
@@ -0,0 +1,9 @@
+---
+fixes:
+  - |
+    Fixed the issue
+    `bug 2044235 <https://bugs.launchpad.net/nova/+bug/2044235>`__ where Nova
+    Conductor puts an instance into an error state if any errors occur during
+    execution of the 'check_can_live_migrate_source()' method in an RPC call.
+    Now, any error is caught and a MigrationPreCheckError exception is re-raised
+    to reset the instance state.