compute: enhance compute evacuate instance to support target state

Related to the bp/allowing-target-state-for-evacuate. This change
is extending compute API to accept a new argument targetState.

The targetState argument when set will force state of an evacuated
instance to the destination host.

Signed-off-by: Sahid Orentino Ferdjaoui <[email protected]>
Change-Id: I9660d42937ad62d647afc6be965f166cc5631392

Author: sahid

nova/compute/api.py

@@ -3797,7 +3797,8 @@ def _reset_image_metadata():
                 orig_sys_metadata=orig_sys_metadata, bdms=bdms,
                 preserve_ephemeral=preserve_ephemeral, host=host,
                 request_spec=request_spec,
-                reimage_boot_volume=reimage_boot_volume)
+                reimage_boot_volume=reimage_boot_volume,
+                target_state=None)
 
     def _check_volume_status(self, context, bdms):
         """Check whether the status of the volume is "in-use".
@@ -5617,7 +5618,7 @@ def live_migrate_abort(self, context, instance, migration_id,
     @check_instance_state(vm_state=[vm_states.ACTIVE, vm_states.STOPPED,
                                     vm_states.ERROR], task_state=None)
     def evacuate(self, context, instance, host, on_shared_storage,
-                 admin_password=None, force=None):
+                 admin_password=None, force=None, target_state=None):
         """Running evacuate to target host.
 
         Checking vm compute host state, if the host not in expected_state,
@@ -5628,6 +5629,7 @@ def evacuate(self, context, instance, host, on_shared_storage,
         :param on_shared_storage: True if instance files on shared storage
         :param admin_password: password to set on rebuilt instance
         :param force: Force the evacuation to the specific host target
+        :param target_state: Set a target state for the evacuated instance
 
         """
         LOG.debug('vm evacuation scheduled', instance=instance)
@@ -5691,7 +5693,7 @@ def evacuate(self, context, instance, host, on_shared_storage,
                        on_shared_storage=on_shared_storage,
                        host=host,
                        request_spec=request_spec,
-                       )
+                       target_state=target_state)
 
     def get_migrations(self, context, filters):
         """Get all migrations for the given filters."""

nova/compute/manager.py

@@ -618,7 +618,7 @@ def update_compute_provider_status(self, context, rp_uuid, enabled):
 class ComputeManager(manager.Manager):
     """Manages the running instances from creation to destruction."""
 
-    target = messaging.Target(version='6.1')
+    target = messaging.Target(version='6.2')
 
     def __init__(self, compute_driver=None, *args, **kwargs):
         """Load configuration options and connect to the hypervisor."""
@@ -3674,7 +3674,7 @@ def rebuild_instance(self, context, instance, orig_image_ref, image_ref,
                          bdms, recreate, on_shared_storage,
                          preserve_ephemeral, migration,
                          scheduled_node, limits, request_spec, accel_uuids,
-                         reimage_boot_volume):
+                         reimage_boot_volume, target_state):
         """Destroy and re-make this instance.
 
         A 'rebuild' effectively purges all existing data from the system and
@@ -3709,6 +3709,7 @@ def rebuild_instance(self, context, instance, orig_image_ref, image_ref,
         :param reimage_boot_volume: Boolean to specify whether the user has
                                     explicitly requested to rebuild a boot
                                     volume
+        :param target_state: Set a target state for the evacuated instance.
 
         """
         # recreate=True means the instance is being evacuated from a failed
@@ -3773,7 +3774,8 @@ def rebuild_instance(self, context, instance, orig_image_ref, image_ref,
                     image_meta, injected_files, new_pass, orig_sys_metadata,
                     bdms, evacuate, on_shared_storage, preserve_ephemeral,
                     migration, request_spec, allocs, rebuild_claim,
-                    scheduled_node, limits, accel_uuids, reimage_boot_volume)
+                    scheduled_node, limits, accel_uuids, reimage_boot_volume,
+                    target_state)
             except (exception.ComputeResourcesUnavailable,
                     exception.RescheduledException) as e:
                 if isinstance(e, exception.ComputeResourcesUnavailable):
@@ -3833,7 +3835,7 @@ def _do_rebuild_instance_with_claim(
             injected_files, new_pass, orig_sys_metadata, bdms, evacuate,
             on_shared_storage, preserve_ephemeral, migration, request_spec,
             allocations, rebuild_claim, scheduled_node, limits, accel_uuids,
-            reimage_boot_volume):
+            reimage_boot_volume, target_state):
         """Helper to avoid deep nesting in the top-level method."""
 
         provider_mapping = None
@@ -3857,7 +3859,8 @@ def _do_rebuild_instance_with_claim(
                 context, instance, orig_image_ref, image_meta, injected_files,
                 new_pass, orig_sys_metadata, bdms, evacuate, on_shared_storage,
                 preserve_ephemeral, migration, request_spec, allocations,
-                provider_mapping, accel_uuids, reimage_boot_volume)
+                provider_mapping, accel_uuids, reimage_boot_volume,
+                target_state)
 
     @staticmethod
     def _get_image_name(image_meta):
@@ -3871,10 +3874,18 @@ def _do_rebuild_instance(
             injected_files, new_pass, orig_sys_metadata, bdms, evacuate,
             on_shared_storage, preserve_ephemeral, migration, request_spec,
             allocations, request_group_resource_providers_mapping,
-            accel_uuids, reimage_boot_volume):
+            accel_uuids, reimage_boot_volume, target_state):
         orig_vm_state = instance.vm_state
 
         if evacuate:
+            if target_state and orig_vm_state != vm_states.ERROR:
+                # This will ensure that at destination the instance will have
+                # the desired state.
+                if target_state not in vm_states.ALLOW_TARGET_STATES:
+                    raise exception.InstanceEvacuateNotSupportedTargetState(
+                        target_state=target_state)
+                orig_vm_state = target_state
+
             if request_spec:
                 # NOTE(gibi): Do a late check of server group policy as
                 # parallel scheduling could violate such policy. This will
@@ -11347,7 +11358,7 @@ def rebuild_instance(self, context, instance, orig_image_ref, image_ref,
             bdms, recreate, on_shared_storage,
             preserve_ephemeral, migration,
             scheduled_node, limits, request_spec,
-            accel_uuids, False)
+            accel_uuids, False, None)
 
     # 5.13 support for optional accel_uuids argument
     def shelve_instance(self, context, instance, image_id,

nova/compute/rpcapi.py

@@ -403,6 +403,7 @@ class ComputeAPI(object):
         * ...  - Rename the instance_type argument of resize_instance() to
                  flavor
         * 6.1 - Add reimage_boot_volume parameter to rebuild_instance()
+        * 6.2 - Add target_state parameter to rebuild_instance()
     '''
 
     VERSION_ALIASES = {
@@ -424,6 +425,7 @@ class ComputeAPI(object):
         'xena': '6.0',
         'yoga': '6.0',
         'zed': '6.1',
+        'antilope': '6.2',
     }
 
     @property
@@ -1083,7 +1085,7 @@ def rebuild_instance(
             image_ref, orig_image_ref, orig_sys_metadata, bdms,
             recreate, on_shared_storage, host, node,
             preserve_ephemeral, migration, limits, request_spec, accel_uuids,
-            reimage_boot_volume):
+            reimage_boot_volume, target_state):
 
         # NOTE(edleafe): compute nodes can only use the dict form of limits.
         if isinstance(limits, objects.SchedulerLimits):
@@ -1096,11 +1098,19 @@ def rebuild_instance(
             'limits': limits,
             'request_spec': request_spec,
             'accel_uuids': accel_uuids,
-            'reimage_boot_volume': reimage_boot_volume
+            'reimage_boot_volume': reimage_boot_volume,
+            'target_state': target_state,
         }
-
-        version = '6.1'
+        version = '6.2'
         client = self.router.client(ctxt)
+        if not client.can_send_version(version):
+            if msg_args['target_state']:
+                raise exception.UnsupportedRPCVersion(
+                    api="rebuild_instance",
+                    required="6.2")
+            else:
+                del msg_args['target_state']
+            version = '6.1'
         if not client.can_send_version(version):
             if msg_args['reimage_boot_volume']:
                 raise exception.NovaException(

nova/compute/vm_states.py

@@ -76,3 +76,6 @@
 
 # states we allow resources to be freed in
 ALLOW_RESOURCE_REMOVAL = [DELETED, SHELVED_OFFLOADED]
+
+# states we allow for evacuate instance
+ALLOW_TARGET_STATES = [STOPPED]

nova/conductor/api.py

@@ -144,7 +144,8 @@ def rebuild_instance(self, context, instance, orig_image_ref, image_ref,
                          injected_files, new_pass, orig_sys_metadata,
                          bdms, recreate=False, on_shared_storage=False,
                          preserve_ephemeral=False, host=None,
-                         request_spec=None, reimage_boot_volume=False):
+                         request_spec=None, reimage_boot_volume=False,
+                         target_state=None):
         self.conductor_compute_rpcapi.rebuild_instance(context,
                 instance=instance,
                 new_pass=new_pass,
@@ -158,7 +159,8 @@ def rebuild_instance(self, context, instance, orig_image_ref, image_ref,
                 preserve_ephemeral=preserve_ephemeral,
                 host=host,
                 request_spec=request_spec,
-                reimage_boot_volume=reimage_boot_volume)
+                reimage_boot_volume=reimage_boot_volume,
+                target_state=target_state)
 
     def cache_images(self, context, aggregate, image_ids):
         """Request images be pre-cached on hosts within an aggregate.

nova/conductor/manager.py

@@ -235,7 +235,7 @@ class ComputeTaskManager:
     may involve coordinating activities on multiple compute nodes.
     """
 
-    target = messaging.Target(namespace='compute_task', version='1.24')
+    target = messaging.Target(namespace='compute_task', version='1.25')
 
     def __init__(self):
         self.compute_rpcapi = compute_rpcapi.ComputeAPI()
@@ -1152,7 +1152,8 @@ def rebuild_instance(self, context, instance, orig_image_ref, image_ref,
                          injected_files, new_pass, orig_sys_metadata,
                          bdms, recreate, on_shared_storage,
                          preserve_ephemeral=False, host=None,
-                         request_spec=None, reimage_boot_volume=False):
+                         request_spec=None, reimage_boot_volume=False,
+                         target_state=None):
         # recreate=True means the instance is being evacuated from a failed
         # host to a new destination host. The 'recreate' variable name is
         # confusing, so rename it to evacuate here at the top, which is simpler
@@ -1356,7 +1357,8 @@ def rebuild_instance(self, context, instance, orig_image_ref, image_ref,
                 limits=limits,
                 request_spec=request_spec,
                 accel_uuids=accel_uuids,
-                reimage_boot_volume=reimage_boot_volume)
+                reimage_boot_volume=reimage_boot_volume,
+                target_state=target_state)
 
     def _validate_image_traits_for_rebuild(self, context, instance, image_ref):
         """Validates that the traits specified in the image can be satisfied

nova/conductor/rpcapi.py

@@ -287,6 +287,7 @@ class ComputeTaskAPI(object):
     1.22 - Added confirm_snapshot_based_resize()
     1.23 - Added revert_snapshot_based_resize()
     1.24 - Add reimage_boot_volume parameter to rebuild_instance()
+    1.25 - Add target_state parameter to rebuild_instance()
     """
 
     def __init__(self):
@@ -428,8 +429,8 @@ def rebuild_instance(self, ctxt, instance, new_pass, injected_files,
             image_ref, orig_image_ref, orig_sys_metadata, bdms,
             recreate=False, on_shared_storage=False, host=None,
             preserve_ephemeral=False, request_spec=None,
-            reimage_boot_volume=False):
-        version = '1.24'
+            reimage_boot_volume=False, target_state=None):
+        version = '1.25'
         kw = {'instance': instance,
               'new_pass': new_pass,
               'injected_files': injected_files,
@@ -442,8 +443,16 @@ def rebuild_instance(self, ctxt, instance, new_pass, injected_files,
               'preserve_ephemeral': preserve_ephemeral,
               'host': host,
               'request_spec': request_spec,
-              'reimage_boot_volume': reimage_boot_volume
+              'reimage_boot_volume': reimage_boot_volume,
+              'target_state': target_state,
               }
+        if not self.client.can_send_version(version):
+            if kw['target_state']:
+                raise exception.UnsupportedRPCVersion(
+                    api="rebuild_instance", required="1.25")
+            else:
+                del kw['target_state']
+            version = '1.24'
         if not self.client.can_send_version(version):
             if kw['reimage_boot_volume']:
                 raise exception.NovaException(

nova/exception.py

@@ -1451,6 +1451,11 @@ class InstanceEvacuateNotSupported(Invalid):
     msg_fmt = _('Instance evacuate is not supported.')
 
 
+class InstanceEvacuateNotSupportedTargetState(Invalid):
+    msg_fmt = _("Target state '%(target_state)s' for instance evacuate "
+                "is not supported.")
+
+
 class DBNotAllowed(NovaException):
     msg_fmt = _('%(binary)s attempted direct database access which is '
                 'not allowed by policy')
@@ -1479,6 +1484,11 @@ class UnsupportedRescueImage(Invalid):
     msg_fmt = _("Requested rescue image '%(image)s' is not supported")
 
 
+class UnsupportedRPCVersion(Invalid):
+    msg_fmt = _("Unsupported RPC version for %(api)s. "
+                "Required >= %(required)s")
+
+
 class Base64Exception(NovaException):
     msg_fmt = _("Invalid Base 64 data for file %(path)s")
 

nova/objects/service.py

@@ -31,7 +31,7 @@
 
 
 # NOTE(danms): This is the global service version counter
-SERVICE_VERSION = 65
+SERVICE_VERSION = 66
 
 
 # NOTE(danms): This is our SERVICE_VERSION history. The idea is that any
@@ -228,6 +228,9 @@
     # Version 65: Compute RPC v6.1:
     # Added stable local node identity
     {'compute_rpc': '6.1'},
+    # Version 66: Compute RPC v6.2:
+    # Add target_state parameter to rebuild_instance()
+    {'compute_rpc': '6.2'},
 )
 
 # This is the version after which we can rely on having a persistent

nova/tests/functional/api_sample_tests/test_evacuate.py

@@ -80,7 +80,7 @@ def test_server_evacuate(self, rebuild_mock):
                 orig_sys_metadata=mock.ANY, bdms=mock.ANY, recreate=mock.ANY,
                 on_shared_storage=False, preserve_ephemeral=mock.ANY,
                 host='testHost', request_spec=mock.ANY,
-                reimage_boot_volume=False)
+                reimage_boot_volume=False, target_state=None)
 
     @mock.patch('nova.conductor.manager.ComputeTaskManager.rebuild_instance')
     def test_server_evacuate_find_host(self, rebuild_mock):
@@ -97,7 +97,7 @@ def test_server_evacuate_find_host(self, rebuild_mock):
                 orig_sys_metadata=mock.ANY, bdms=mock.ANY, recreate=mock.ANY,
                 on_shared_storage=False, preserve_ephemeral=mock.ANY,
                 host=None, request_spec=mock.ANY,
-                reimage_boot_volume=False)
+                reimage_boot_volume=False, target_state=None)
 
 
 class EvacuateJsonTestV214(EvacuateJsonTest):
@@ -119,7 +119,7 @@ def test_server_evacuate(self, rebuild_mock):
                 orig_sys_metadata=mock.ANY, bdms=mock.ANY, recreate=mock.ANY,
                 on_shared_storage=None, preserve_ephemeral=mock.ANY,
                 host='testHost', request_spec=mock.ANY,
-                reimage_boot_volume=False)
+                reimage_boot_volume=False, target_state=None)
 
     @mock.patch('nova.conductor.manager.ComputeTaskManager.rebuild_instance')
     def test_server_evacuate_find_host(self, rebuild_mock):
@@ -135,7 +135,7 @@ def test_server_evacuate_find_host(self, rebuild_mock):
                 orig_sys_metadata=mock.ANY, bdms=mock.ANY, recreate=mock.ANY,
                 on_shared_storage=None, preserve_ephemeral=mock.ANY,
                 host=None, request_spec=mock.ANY,
-                reimage_boot_volume=False)
+                reimage_boot_volume=False, target_state=None)
 
 
 class EvacuateJsonTestV229(EvacuateJsonTestV214):
@@ -163,7 +163,7 @@ def test_server_evacuate(self, compute_node_get_all_by_host, rebuild_mock):
                 orig_sys_metadata=mock.ANY, bdms=mock.ANY, recreate=mock.ANY,
                 on_shared_storage=None, preserve_ephemeral=mock.ANY,
                 host=None, request_spec=mock.ANY,
-                reimage_boot_volume=False)
+                reimage_boot_volume=False, target_state=None)
 
     @mock.patch('nova.conductor.manager.ComputeTaskManager.rebuild_instance')
     @mock.patch('nova.objects.ComputeNodeList.get_all_by_host')
@@ -184,7 +184,7 @@ def test_server_evacuate_with_force(self, compute_node_get_all_by_host,
                 orig_sys_metadata=mock.ANY, bdms=mock.ANY, recreate=mock.ANY,
                 on_shared_storage=None, preserve_ephemeral=mock.ANY,
                 host='testHost', request_spec=mock.ANY,
-                reimage_boot_volume=False)
+                reimage_boot_volume=False, target_state=None)
 
 
 class EvacuateJsonTestV268(EvacuateJsonTestV229):
@@ -211,7 +211,7 @@ def test_server_evacuate(self, compute_node_get_all_by_host, rebuild_mock):
                 orig_sys_metadata=mock.ANY, bdms=mock.ANY, recreate=mock.ANY,
                 on_shared_storage=None, preserve_ephemeral=mock.ANY,
                 host=None, request_spec=mock.ANY,
-                reimage_boot_volume=False)
+                reimage_boot_volume=False, target_state=None)
 
     def test_server_evacuate_with_force(self):
         # doesn't apply to v2.68+, which removed the ability to force migrate