The race is that post_live_migration_at_destination sets the instance task_state to None which effectively puts it back into ACTIVE status before the migration status marked 'completed' on the source node in _post_live_migration, hence the race. I seem to remember a similar race in another live migration functional test where gibi added a waiter for a notification before checking the migration status.
The race is that post_live_ migration_ at_destination sets the instance task_state to None which effectively puts it back into ACTIVE status before the migration status marked 'completed' on the source node in _post_live_ migration, hence the race. I seem to remember a similar race in another live migration functional test where gibi added a waiter for a notification before checking the migration status.