Add a workaround to skip hypervisor version check on LM

When turned on, this will disable the version-checking of hypervisors
during live-migration.  This can be useful for operators in certain
scenarios when upgrading.  E.g. if you want to relocate all instances
off a compute node due to an emergency hardware issue, and you only have
another old compute node ready at the time.

Note, though: libvirt will do its own internal compatibility checks, and
might still reject live migration if the destination is incompatible.

Closes-Bug: #1982853

Change-Id: Iec387dcbc49ddb91ebf5cfd188224eaf6021c0e1
Signed-off-by: Kashyap Chamarthy <kchamart@redhat.com>
(cherry picked from commit 00ed8a232bc22f48011e95a0b47750520a5b4d47)

4 files changed, 53 insertions, 2 deletions

diff --git a/nova/conductor/tasks/live_migrate.py b/nova/conductor/tasks/live_migrate.py
index 1acae88b26..f8819b0dc8 100644
--- a/nova/conductor/tasks/live_migrate.py
+++ b/nova/conductor/tasks/live_migrate.py
@@ -347,8 +347,9 @@ class LiveMigrationTask(base.TaskBase):
 
         source_version = source_info.hypervisor_version
         destination_version = destination_info.hypervisor_version
-        if source_version > destination_version:
-            raise exception.DestinationHypervisorTooOld()
+        if not CONF.workarounds.skip_hypervisor_version_check_on_lm:
+            if source_version > destination_version:
+                raise exception.DestinationHypervisorTooOld()
         return source_info, destination_info
 
     def _call_livem_checks_on_host(self, destination, provider_mapping):
diff --git a/nova/conf/workarounds.py b/nova/conf/workarounds.py
index 6c52eae8e5..2ec53282cd 100644
--- a/nova/conf/workarounds.py
+++ b/nova/conf/workarounds.py
@@ -410,6 +410,13 @@ with the destination host. When using QEMU >= 2.9 and libvirt >=
 4.4.0, libvirt will do the correct thing with respect to checking CPU
 compatibility on the destination host during live migration.
 """),
+    cfg.BoolOpt(
+        'skip_hypervisor_version_check_on_lm',
+        default=False,
+        help="""
+When this is enabled, it will skip version-checking of hypervisors
+during live migration.
+"""),
 ]
 
 
diff --git a/nova/tests/unit/conductor/tasks/test_live_migrate.py b/nova/tests/unit/conductor/tasks/test_live_migrate.py
index cb40c076c8..dd4ee7c3fe 100644
--- a/nova/tests/unit/conductor/tasks/test_live_migrate.py
+++ b/nova/tests/unit/conductor/tasks/test_live_migrate.py
@@ -345,6 +345,36 @@ class LiveMigrationTaskTestCase(test.NoDBTestCase):
                           mock.call(self.destination)],
                          mock_get_info.call_args_list)
 
+    @mock.patch.object(live_migrate.LiveMigrationTask, '_get_compute_info')
+    def test_skip_hypervisor_version_check_on_lm_raise_ex(self, mock_get_info):
+        host1 = {'hypervisor_type': 'a', 'hypervisor_version': 7}
+        host2 = {'hypervisor_type': 'a', 'hypervisor_version': 6}
+        self.flags(group='workarounds',
+                   skip_hypervisor_version_check_on_lm=False)
+        mock_get_info.side_effect = [objects.ComputeNode(**host1),
+                                     objects.ComputeNode(**host2)]
+        self.assertRaises(exception.DestinationHypervisorTooOld,
+                          self.task._check_compatible_with_source_hypervisor,
+                          self.destination)
+        self.assertEqual([mock.call(self.instance_host),
+                          mock.call(self.destination)],
+                         mock_get_info.call_args_list)
+
+    @mock.patch.object(live_migrate.LiveMigrationTask, '_get_compute_info')
+    def test_skip_hypervisor_version_check_on_lm_do_not_raise_ex(
+        self, mock_get_info
+    ):
+        host1 = {'hypervisor_type': 'a', 'hypervisor_version': 7}
+        host2 = {'hypervisor_type': 'a', 'hypervisor_version': 6}
+        self.flags(group='workarounds',
+                   skip_hypervisor_version_check_on_lm=True)
+        mock_get_info.side_effect = [objects.ComputeNode(**host1),
+                                     objects.ComputeNode(**host2)]
+        self.task._check_compatible_with_source_hypervisor(self.destination)
+        self.assertEqual([mock.call(self.instance_host),
+                          mock.call(self.destination)],
+                         mock_get_info.call_args_list)
+
     @mock.patch.object(compute_rpcapi.ComputeAPI,
                        'check_can_live_migrate_destination')
     def test_check_requested_destination(self, mock_check):
diff --git a/releasenotes/notes/skip-hypervisor-version-check-on-lm-a87f2dcb4f8bf0f2.yaml b/releasenotes/notes/skip-hypervisor-version-check-on-lm-a87f2dcb4f8bf0f2.yaml
new file mode 100644
index 0000000000..00fe6a24c7
--- /dev/null
+++ b/releasenotes/notes/skip-hypervisor-version-check-on-lm-a87f2dcb4f8bf0f2.yaml
@@ -0,0 +1,13 @@
+---
+feature:
+  - |
+    Adds a workaround that allows one to disable hypervisor
+    version-check on live migration.  This workaround option can be
+    useful in certain scenarios when upgrading.  E.g. if you want to
+    relocate all instances off a compute node due to an emergency
+    hardware issue, and you only have another old compute node ready at
+    the time.
+
+    To enable this, use the config attribute
+    ``[workarounds]skip_hypervisor_version_check_on_lm=True`` in
+    ``nova.conf``.  The option defaults to ``False``.