/
manager.py
2156 lines (1965 loc) · 104 KB
/
manager.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# Copyright 2013 IBM Corp.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""Handles database requests from other nova services."""
import collections
import contextlib
import copy
import eventlet
import functools
import sys
from keystoneauth1 import exceptions as ks_exc
from oslo_config import cfg
from oslo_db import exception as db_exc
from oslo_limit import exception as limit_exceptions
from oslo_log import log as logging
import oslo_messaging as messaging
from oslo_serialization import jsonutils
from oslo_utils import excutils
from oslo_utils import timeutils
from oslo_utils import versionutils
from nova.accelerator import cyborg
from nova import availability_zones
from nova.compute import instance_actions
from nova.compute import rpcapi as compute_rpcapi
from nova.compute import task_states
from nova.compute import utils as compute_utils
from nova.compute.utils import wrap_instance_event
from nova.compute import vm_states
from nova.conductor.tasks import cross_cell_migrate
from nova.conductor.tasks import live_migrate
from nova.conductor.tasks import migrate
from nova import context as nova_context
from nova import exception
from nova.i18n import _
from nova.image import glance
from nova.limit import placement as placement_limits
from nova import manager
from nova.network import neutron
from nova import notifications
from nova import objects
from nova.objects import base as nova_object
from nova.objects import fields
from nova import profiler
from nova import rpc
from nova.scheduler.client import query
from nova.scheduler.client import report
from nova.scheduler import utils as scheduler_utils
from nova import servicegroup
from nova import utils
from nova.volume import cinder
LOG = logging.getLogger(__name__)
CONF = cfg.CONF
def targets_cell(fn):
"""Wrap a method and automatically target the instance's cell.
This decorates a method with signature func(self, context, instance, ...)
and automatically targets the context with the instance's cell
mapping. It does this by looking up the InstanceMapping.
"""
@functools.wraps(fn)
def wrapper(self, context, *args, **kwargs):
instance = kwargs.get('instance') or args[0]
try:
im = objects.InstanceMapping.get_by_instance_uuid(
context, instance.uuid)
except exception.InstanceMappingNotFound:
LOG.error('InstanceMapping not found, unable to target cell',
instance=instance)
except db_exc.CantStartEngineError:
# Check to see if we can ignore API DB connection failures
# because we might already be in the cell conductor.
with excutils.save_and_reraise_exception() as err_ctxt:
if CONF.api_database.connection is None:
err_ctxt.reraise = False
else:
LOG.debug('Targeting cell %(cell)s for conductor method %(meth)s',
{'cell': im.cell_mapping.identity,
'meth': fn.__name__})
# NOTE(danms): Target our context to the cell for the rest of
# this request, so that none of the subsequent code needs to
# care about it.
nova_context.set_target_cell(context, im.cell_mapping)
return fn(self, context, *args, **kwargs)
return wrapper
class ConductorManager(manager.Manager):
"""Mission: Conduct things.
The methods in the base API for nova-conductor are various proxy operations
performed on behalf of the nova-compute service running on compute nodes.
Compute nodes are not allowed to directly access the database, so this set
of methods allows them to get specific work done without locally accessing
the database.
The nova-conductor service also exposes an API in the 'compute_task'
namespace. See the ComputeTaskManager class for details.
"""
target = messaging.Target(version='3.0')
def __init__(self, *args, **kwargs):
super(ConductorManager, self).__init__(service_name='conductor',
*args, **kwargs)
self.compute_task_mgr = ComputeTaskManager()
self.additional_endpoints.append(self.compute_task_mgr)
# NOTE(hanlind): This can be removed in version 4.0 of the RPC API
def provider_fw_rule_get_all(self, context):
# NOTE(hanlind): Simulate an empty db result for compat reasons.
return []
def _object_dispatch(self, target, method, args, kwargs):
"""Dispatch a call to an object method.
This ensures that object methods get called and any exception
that is raised gets wrapped in an ExpectedException for forwarding
back to the caller (without spamming the conductor logs).
"""
try:
# NOTE(danms): Keep the getattr inside the try block since
# a missing method is really a client problem
return getattr(target, method)(*args, **kwargs)
except Exception:
raise messaging.ExpectedException()
def object_class_action_versions(self, context, objname, objmethod,
object_versions, args, kwargs):
objclass = nova_object.NovaObject.obj_class_from_name(
objname, object_versions[objname])
args = tuple([context] + list(args))
result = self._object_dispatch(objclass, objmethod, args, kwargs)
# NOTE(danms): The RPC layer will convert to primitives for us,
# but in this case, we need to honor the version the client is
# asking for, so we do it before returning here.
# NOTE(hanlind): Do not convert older than requested objects,
# see bug #1596119.
if isinstance(result, nova_object.NovaObject):
target_version = object_versions[objname]
requested_version = versionutils.convert_version_to_tuple(
target_version)
actual_version = versionutils.convert_version_to_tuple(
result.VERSION)
do_backport = requested_version < actual_version
other_major_version = requested_version[0] != actual_version[0]
if do_backport or other_major_version:
result = result.obj_to_primitive(
target_version=target_version,
version_manifest=object_versions)
return result
def object_action(self, context, objinst, objmethod, args, kwargs):
"""Perform an action on an object."""
oldobj = objinst.obj_clone()
result = self._object_dispatch(objinst, objmethod, args, kwargs)
updates = dict()
# NOTE(danms): Diff the object with the one passed to us and
# generate a list of changes to forward back
for name, field in objinst.fields.items():
if not objinst.obj_attr_is_set(name):
# Avoid demand-loading anything
continue
if (not oldobj.obj_attr_is_set(name) or
getattr(oldobj, name) != getattr(objinst, name)):
updates[name] = field.to_primitive(objinst, name,
getattr(objinst, name))
# This is safe since a field named this would conflict with the
# method anyway
updates['obj_what_changed'] = objinst.obj_what_changed()
return updates, result
def object_backport_versions(self, context, objinst, object_versions):
target = object_versions[objinst.obj_name()]
LOG.debug('Backporting %(obj)s to %(ver)s with versions %(manifest)s',
{'obj': objinst.obj_name(),
'ver': target,
'manifest': ','.join(
['%s=%s' % (name, ver)
for name, ver in object_versions.items()])})
return objinst.obj_to_primitive(target_version=target,
version_manifest=object_versions)
def reset(self):
objects.Service.clear_min_version_cache()
@contextlib.contextmanager
def try_target_cell(context, cell):
"""If cell is not None call func with context.target_cell.
This is a method to help during the transition period. Currently
various mappings may not exist if a deployment has not migrated to
cellsv2. If there is no mapping call the func as normal, otherwise
call it in a target_cell context.
"""
if cell:
with nova_context.target_cell(context, cell) as cell_context:
yield cell_context
else:
yield context
@contextlib.contextmanager
def obj_target_cell(obj, cell):
"""Run with object's context set to a specific cell"""
with try_target_cell(obj._context, cell) as target:
with obj.obj_alternate_context(target):
yield target
@profiler.trace_cls("rpc")
class ComputeTaskManager:
"""Namespace for compute methods.
This class presents an rpc API for nova-conductor under the 'compute_task'
namespace. The methods here are compute operations that are invoked
by the API service. These methods see the operation to completion, which
may involve coordinating activities on multiple compute nodes.
"""
target = messaging.Target(namespace='compute_task', version='1.23')
def __init__(self):
self.compute_rpcapi = compute_rpcapi.ComputeAPI()
self.volume_api = cinder.API()
self.image_api = glance.API()
self.network_api = neutron.API()
self.servicegroup_api = servicegroup.API()
self.query_client = query.SchedulerQueryClient()
self.notifier = rpc.get_notifier('compute')
# Help us to record host in EventReporter
self.host = CONF.host
try:
# Test our placement client during initialization
self.report_client
except (ks_exc.EndpointNotFound,
ks_exc.DiscoveryFailure,
ks_exc.RequestTimeout,
ks_exc.GatewayTimeout,
ks_exc.ConnectFailure) as e:
# Non-fatal, likely transient (although not definitely);
# continue startup but log the warning so that when things
# fail later, it will be clear why we can not do certain
# things.
LOG.warning('Unable to initialize placement client (%s); '
'Continuing with startup, but some operations '
'will not be possible.', e)
except (ks_exc.MissingAuthPlugin,
ks_exc.Unauthorized) as e:
# This is almost definitely fatal mis-configuration. The
# Unauthorized error might be transient, but it is
# probably reasonable to consider it fatal.
LOG.error('Fatal error initializing placement client; '
'config is incorrect or incomplete: %s', e)
raise
except Exception as e:
# Unknown/unexpected errors here are fatal
LOG.error('Fatal error initializing placement client: %s', e)
raise
@property
def report_client(self):
return report.report_client_singleton()
def reset(self):
LOG.info('Reloading compute RPC API')
compute_rpcapi.LAST_VERSION = None
self.compute_rpcapi = compute_rpcapi.ComputeAPI()
# TODO(tdurakov): remove `live` parameter here on compute task api RPC
# version bump to 2.x
# TODO(danms): remove the `reservations` parameter here on compute task api
# RPC version bump to 2.x
@messaging.expected_exceptions(
exception.NoValidHost,
exception.ComputeServiceUnavailable,
exception.ComputeHostNotFound,
exception.InvalidHypervisorType,
exception.InvalidCPUInfo,
exception.UnableToMigrateToSelf,
exception.DestinationHypervisorTooOld,
exception.InvalidLocalStorage,
exception.InvalidSharedStorage,
exception.HypervisorUnavailable,
exception.InstanceInvalidState,
exception.MigrationPreCheckError,
exception.UnsupportedPolicyException)
@targets_cell
@wrap_instance_event(prefix='conductor')
def migrate_server(self, context, instance, scheduler_hint, live, rebuild,
flavor, block_migration, disk_over_commit, reservations=None,
clean_shutdown=True, request_spec=None, host_list=None):
if instance and not isinstance(instance, nova_object.NovaObject):
# NOTE(danms): Until v2 of the RPC API, we need to tolerate
# old-world instance objects here
attrs = ['metadata', 'system_metadata', 'info_cache',
'security_groups']
instance = objects.Instance._from_db_object(
context, objects.Instance(), instance,
expected_attrs=attrs)
# NOTE: Remove this when we drop support for v1 of the RPC API
if flavor and not isinstance(flavor, objects.Flavor):
# Code downstream may expect extra_specs to be populated since it
# is receiving an object, so lookup the flavor to ensure this.
flavor = objects.Flavor.get_by_id(context, flavor['id'])
if live and not rebuild and not flavor:
self._live_migrate(context, instance, scheduler_hint,
block_migration, disk_over_commit, request_spec)
elif not live and not rebuild and flavor:
instance_uuid = instance.uuid
with compute_utils.EventReporter(context, 'cold_migrate',
self.host, instance_uuid):
self._cold_migrate(context, instance, flavor,
scheduler_hint['filter_properties'],
clean_shutdown, request_spec,
host_list)
else:
raise NotImplementedError()
@staticmethod
def _get_request_spec_for_cold_migrate(context, instance, flavor,
filter_properties, request_spec):
# NOTE(sbauza): If a reschedule occurs when prep_resize(), then
# it only provides filter_properties legacy dict back to the
# conductor with no RequestSpec part of the payload for <Stein
# computes.
# TODO(mriedem): We can remove this compat code for no request spec
# coming to conductor in ComputeTaskAPI RPC API version 2.0
if not request_spec:
image_meta = utils.get_image_from_system_metadata(
instance.system_metadata)
# Make sure we hydrate a new RequestSpec object with the new flavor
# and not the nested one from the instance
request_spec = objects.RequestSpec.from_components(
context, instance.uuid, image_meta,
flavor, instance.numa_topology, instance.pci_requests,
filter_properties, None, instance.availability_zone,
project_id=instance.project_id, user_id=instance.user_id)
elif not isinstance(request_spec, objects.RequestSpec):
# Prior to compute RPC API 5.1 conductor would pass a legacy dict
# version of the request spec to compute and Stein compute
# could be sending that back to conductor on reschedule, so if we
# got a dict convert it to an object.
# TODO(mriedem): We can drop this compat code when we only support
# compute RPC API >=6.0.
request_spec = objects.RequestSpec.from_primitives(
context, request_spec, filter_properties)
# We don't have to set the new flavor on the request spec because
# if we got here it was due to a reschedule from the compute and
# the request spec would already have the new flavor in it from the
# else block below.
else:
# NOTE(sbauza): Resizes means new flavor, so we need to update the
# original RequestSpec object for make sure the scheduler verifies
# the right one and not the original flavor
request_spec.flavor = flavor
return request_spec
def _cold_migrate(self, context, instance, flavor, filter_properties,
clean_shutdown, request_spec, host_list):
request_spec = self._get_request_spec_for_cold_migrate(
context, instance, flavor, filter_properties, request_spec)
task = self._build_cold_migrate_task(context, instance, flavor,
request_spec, clean_shutdown, host_list)
try:
task.execute()
except exception.NoValidHost as ex:
vm_state = instance.vm_state
if not vm_state:
vm_state = vm_states.ACTIVE
updates = {'vm_state': vm_state, 'task_state': None}
self._set_vm_state_and_notify(context, instance.uuid,
'migrate_server',
updates, ex, request_spec)
# if the flavor IDs match, it's migrate; otherwise resize
if flavor.id == instance.instance_type_id:
msg = _("No valid host found for cold migrate")
else:
msg = _("No valid host found for resize")
raise exception.NoValidHost(reason=msg)
except exception.UnsupportedPolicyException as ex:
with excutils.save_and_reraise_exception():
vm_state = instance.vm_state
if not vm_state:
vm_state = vm_states.ACTIVE
updates = {'vm_state': vm_state, 'task_state': None}
self._set_vm_state_and_notify(context, instance.uuid,
'migrate_server',
updates, ex, request_spec)
except Exception as ex:
with excutils.save_and_reraise_exception():
# Refresh the instance so we don't overwrite vm_state changes
# set after we executed the task.
try:
instance.refresh()
# Passing vm_state is kind of silly but it's expected in
# set_vm_state_and_notify.
updates = {'vm_state': instance.vm_state,
'task_state': None}
self._set_vm_state_and_notify(context, instance.uuid,
'migrate_server',
updates, ex, request_spec)
except exception.InstanceNotFound:
# We can't send the notification because the instance is
# gone so just log it.
LOG.info('During %s the instance was deleted.',
'resize' if instance.instance_type_id != flavor.id
else 'cold migrate', instance=instance)
# NOTE(sbauza): Make sure we persist the new flavor in case we had
# a successful scheduler call if and only if nothing bad happened
if request_spec.obj_what_changed():
request_spec.save()
def _set_vm_state_and_notify(self, context, instance_uuid, method, updates,
ex, request_spec):
scheduler_utils.set_vm_state_and_notify(
context, instance_uuid, 'compute_task', method, updates,
ex, request_spec)
def _cleanup_allocated_networks(
self, context, instance, requested_networks):
try:
# If we were told not to allocate networks let's save ourselves
# the trouble of calling the network API.
if not (requested_networks and requested_networks.no_allocate):
self.network_api.deallocate_for_instance(
context, instance, requested_networks=requested_networks)
except Exception:
LOG.exception('Failed to deallocate networks', instance=instance)
return
instance.system_metadata['network_allocated'] = 'False'
try:
instance.save()
except exception.InstanceNotFound:
# NOTE: It's possible that we're cleaning up the networks
# because the instance was deleted. If that's the case then this
# exception will be raised by instance.save()
pass
@targets_cell
@wrap_instance_event(prefix='conductor')
def live_migrate_instance(self, context, instance, scheduler_hint,
block_migration, disk_over_commit, request_spec):
self._live_migrate(context, instance, scheduler_hint,
block_migration, disk_over_commit, request_spec)
def _live_migrate(self, context, instance, scheduler_hint,
block_migration, disk_over_commit, request_spec):
destination = scheduler_hint.get("host")
def _set_vm_state(context, instance, ex, vm_state=None,
task_state=None):
request_spec = {'instance_properties': {
'uuid': instance.uuid, },
}
scheduler_utils.set_vm_state_and_notify(context,
instance.uuid,
'compute_task', 'migrate_server',
dict(vm_state=vm_state,
task_state=task_state,
expected_task_state=task_states.MIGRATING,),
ex, request_spec)
migration = objects.Migration(context=context.elevated())
migration.dest_compute = destination
migration.status = 'accepted'
migration.instance_uuid = instance.uuid
migration.source_compute = instance.host
migration.migration_type = fields.MigrationType.LIVE_MIGRATION
if instance.obj_attr_is_set('flavor'):
migration.old_instance_type_id = instance.flavor.id
migration.new_instance_type_id = instance.flavor.id
else:
migration.old_instance_type_id = instance.instance_type_id
migration.new_instance_type_id = instance.instance_type_id
migration.create()
task = self._build_live_migrate_task(context, instance, destination,
block_migration, disk_over_commit,
migration, request_spec)
try:
task.execute()
except (exception.NoValidHost,
exception.ComputeHostNotFound,
exception.ComputeServiceUnavailable,
exception.InvalidHypervisorType,
exception.InvalidCPUInfo,
exception.UnableToMigrateToSelf,
exception.DestinationHypervisorTooOld,
exception.InvalidLocalStorage,
exception.InvalidSharedStorage,
exception.HypervisorUnavailable,
exception.InstanceInvalidState,
exception.MigrationPreCheckError,
exception.MigrationSchedulerRPCError) as ex:
with excutils.save_and_reraise_exception():
_set_vm_state(context, instance, ex, instance.vm_state)
migration.status = 'error'
migration.save()
except Exception as ex:
LOG.error('Migration of instance %(instance_id)s to host'
' %(dest)s unexpectedly failed.',
{'instance_id': instance.uuid, 'dest': destination},
exc_info=True)
# Reset the task state to None to indicate completion of
# the operation as it is done in case of known exceptions.
_set_vm_state(context, instance, ex, vm_states.ERROR,
task_state=None)
migration.status = 'error'
migration.save()
raise exception.MigrationError(reason=str(ex))
def _build_live_migrate_task(self, context, instance, destination,
block_migration, disk_over_commit, migration,
request_spec=None):
return live_migrate.LiveMigrationTask(context, instance,
destination, block_migration,
disk_over_commit, migration,
self.compute_rpcapi,
self.servicegroup_api,
self.query_client,
self.report_client,
request_spec)
def _build_cold_migrate_task(self, context, instance, flavor, request_spec,
clean_shutdown, host_list):
return migrate.MigrationTask(context, instance, flavor,
request_spec, clean_shutdown,
self.compute_rpcapi,
self.query_client, self.report_client,
host_list, self.network_api)
def _destroy_build_request(self, context, instance):
# The BuildRequest needs to be stored until the instance is mapped to
# an instance table. At that point it will never be used again and
# should be deleted.
build_request = objects.BuildRequest.get_by_instance_uuid(
context, instance.uuid)
# TODO(alaski): Sync API updates of the build_request to the
# instance before it is destroyed. Right now only locked_by can
# be updated before this is destroyed.
build_request.destroy()
def _populate_instance_mapping(self, context, instance, host):
try:
inst_mapping = objects.InstanceMapping.get_by_instance_uuid(
context, instance.uuid)
except exception.InstanceMappingNotFound:
# NOTE(alaski): If nova-api is up to date this exception should
# never be hit. But during an upgrade it's possible that an old
# nova-api didn't create an instance_mapping during this boot
# request.
LOG.debug('Instance was not mapped to a cell, likely due '
'to an older nova-api service running.',
instance=instance)
return None
else:
try:
host_mapping = objects.HostMapping.get_by_host(context,
host.service_host)
except exception.HostMappingNotFound:
# NOTE(alaski): For now this exception means that a
# deployment has not migrated to cellsv2 and we should
# remove the instance_mapping that has been created.
# Eventually this will indicate a failure to properly map a
# host to a cell and we may want to reschedule.
inst_mapping.destroy()
return None
else:
inst_mapping.cell_mapping = host_mapping.cell_mapping
inst_mapping.save()
return inst_mapping
def _validate_existing_attachment_ids(self, context, instance, bdms):
"""Ensure any attachment ids referenced by the bdms exist.
New attachments will only be created if the attachment ids referenced
by the bdms no longer exist. This can happen when an instance is
rescheduled after a failure to spawn as cleanup code on the previous
host will delete attachments before rescheduling.
"""
for bdm in bdms:
if bdm.is_volume and bdm.attachment_id:
try:
self.volume_api.attachment_get(context, bdm.attachment_id)
except exception.VolumeAttachmentNotFound:
attachment = self.volume_api.attachment_create(
context, bdm.volume_id, instance.uuid)
bdm.attachment_id = attachment['id']
bdm.save()
def _cleanup_when_reschedule_fails(
self, context, instance, exception, legacy_request_spec,
requested_networks):
"""Set the instance state and clean up.
It is only used in case build_instance fails while rescheduling the
instance
"""
updates = {'vm_state': vm_states.ERROR,
'task_state': None}
self._set_vm_state_and_notify(
context, instance.uuid, 'build_instances', updates, exception,
legacy_request_spec)
self._cleanup_allocated_networks(
context, instance, requested_networks)
arq_uuids = None
# arqs have not bound to port/instance yet
if requested_networks:
arq_uuids = [req.arq_uuid
for req in requested_networks if req.arq_uuid]
compute_utils.delete_arqs_if_needed(context, instance, arq_uuids)
# NOTE(danms): This is never cell-targeted because it is only used for
# n-cpu reschedules which go to the cell conductor and thus are always
# cell-specific.
def build_instances(self, context, instances, image, filter_properties,
admin_password, injected_files, requested_networks,
security_groups, block_device_mapping=None, legacy_bdm=True,
request_spec=None, host_lists=None):
# TODO(ndipanov): Remove block_device_mapping and legacy_bdm in version
# 2.0 of the RPC API.
# TODO(danms): Remove this in version 2.0 of the RPC API
if (requested_networks and
not isinstance(requested_networks,
objects.NetworkRequestList)):
requested_networks = objects.NetworkRequestList.from_tuples(
requested_networks)
# TODO(melwitt): Remove this in version 2.0 of the RPC API
flavor = filter_properties.get('instance_type')
if flavor and not isinstance(flavor, objects.Flavor):
# Code downstream may expect extra_specs to be populated since it
# is receiving an object, so lookup the flavor to ensure this.
flavor = objects.Flavor.get_by_id(context, flavor['id'])
filter_properties = dict(filter_properties, instance_type=flavor)
# Older computes will not send a request_spec during reschedules so we
# need to check and build our own if one is not provided.
if request_spec is None:
legacy_request_spec = scheduler_utils.build_request_spec(
image, instances)
else:
# TODO(mriedem): This is annoying but to populate the local
# request spec below using the filter_properties, we have to pass
# in a primitive version of the request spec. Yes it's inefficient
# and we can remove it once the populate_retry and
# populate_filter_properties utility methods are converted to
# work on a RequestSpec object rather than filter_properties.
# NOTE(gibi): we have to keep a reference to the original
# RequestSpec object passed to this function as we lose information
# during the below legacy conversion
legacy_request_spec = request_spec.to_legacy_request_spec_dict()
# 'host_lists' will be None during a reschedule from a pre-Queens
# compute. In all other cases, it will be a list of lists, though the
# lists may be empty if there are no more hosts left in a rescheduling
# situation.
is_reschedule = host_lists is not None
try:
# check retry policy. Rather ugly use of instances[0]...
# but if we've exceeded max retries... then we really only
# have a single instance.
# TODO(sbauza): Provide directly the RequestSpec object
# when populate_retry() accepts it
scheduler_utils.populate_retry(
filter_properties, instances[0].uuid)
instance_uuids = [instance.uuid for instance in instances]
spec_obj = objects.RequestSpec.from_primitives(
context, legacy_request_spec, filter_properties)
LOG.debug("Rescheduling: %s", is_reschedule)
if is_reschedule:
# Make sure that we have a host, as we may have exhausted all
# our alternates
if not host_lists[0]:
# We have an empty list of hosts, so this instance has
# failed to build.
msg = ("Exhausted all hosts available for retrying build "
"failures for instance %(instance_uuid)s." %
{"instance_uuid": instances[0].uuid})
raise exception.MaxRetriesExceeded(reason=msg)
else:
# This is not a reschedule, so we need to call the scheduler to
# get appropriate hosts for the request.
# NOTE(gibi): We only call the scheduler if we are rescheduling
# from a really old compute. In that case we do not support
# externally-defined resource requests, like port QoS. So no
# requested_resources are set on the RequestSpec here.
host_lists = self._schedule_instances(context, spec_obj,
instance_uuids, return_alternates=True)
except Exception as exc:
# NOTE(mriedem): If we're rescheduling from a failed build on a
# compute, "retry" will be set and num_attempts will be >1 because
# populate_retry above will increment it. If the server build was
# forced onto a host/node or [scheduler]/max_attempts=1, "retry"
# won't be in filter_properties and we won't get here because
# nova-compute will just abort the build since reschedules are
# disabled in those cases.
num_attempts = filter_properties.get(
'retry', {}).get('num_attempts', 1)
for instance in instances:
# If num_attempts > 1, we're in a reschedule and probably
# either hit NoValidHost or MaxRetriesExceeded. Either way,
# the build request should already be gone and we probably
# can't reach the API DB from the cell conductor.
if num_attempts <= 1:
try:
# If the BuildRequest stays around then instance
# show/lists will pull from it rather than the errored
# instance.
self._destroy_build_request(context, instance)
except exception.BuildRequestNotFound:
pass
self._cleanup_when_reschedule_fails(
context, instance, exc, legacy_request_spec,
requested_networks)
return
elevated = context.elevated()
for (instance, host_list) in zip(instances, host_lists):
host = host_list.pop(0)
if is_reschedule:
# If this runs in the superconductor, the first instance will
# already have its resources claimed in placement. If this is a
# retry, though, this is running in the cell conductor, and we
# need to claim first to ensure that the alternate host still
# has its resources available. Note that there are schedulers
# that don't support Placement, so must assume that the host is
# still available.
host_available = False
while host and not host_available:
if host.allocation_request:
alloc_req = jsonutils.loads(host.allocation_request)
else:
alloc_req = None
if alloc_req:
try:
host_available = scheduler_utils.claim_resources(
elevated, self.report_client, spec_obj,
instance.uuid, alloc_req,
host.allocation_request_version)
if request_spec and host_available:
# NOTE(gibi): redo the request group - resource
# provider mapping as the above claim call
# moves the allocation of the instance to
# another host
scheduler_utils.fill_provider_mapping(
request_spec, host)
except Exception as exc:
self._cleanup_when_reschedule_fails(
context, instance, exc, legacy_request_spec,
requested_networks)
return
else:
# Some deployments use different schedulers that do not
# use Placement, so they will not have an
# allocation_request to claim with. For those cases,
# there is no concept of claiming, so just assume that
# the host is valid.
host_available = True
if not host_available:
# Insufficient resources remain on that host, so
# discard it and try the next.
host = host_list.pop(0) if host_list else None
if not host_available:
# No more available hosts for retrying the build.
msg = ("Exhausted all hosts available for retrying build "
"failures for instance %(instance_uuid)s." %
{"instance_uuid": instance.uuid})
exc = exception.MaxRetriesExceeded(reason=msg)
self._cleanup_when_reschedule_fails(
context, instance, exc, legacy_request_spec,
requested_networks)
return
# The availability_zone field was added in v1.1 of the Selection
# object so make sure to handle the case where it is missing.
if 'availability_zone' in host:
instance.availability_zone = host.availability_zone
else:
try:
instance.availability_zone = (
availability_zones.get_host_availability_zone(context,
host.service_host))
except Exception as exc:
# Put the instance into ERROR state, set task_state to
# None, inject a fault, etc.
self._cleanup_when_reschedule_fails(
context, instance, exc, legacy_request_spec,
requested_networks)
continue
try:
# NOTE(danms): This saves the az change above, refreshes our
# instance, and tells us if it has been deleted underneath us
instance.save()
except (exception.InstanceNotFound,
exception.InstanceInfoCacheNotFound):
LOG.debug('Instance deleted during build', instance=instance)
continue
local_filter_props = copy.deepcopy(filter_properties)
scheduler_utils.populate_filter_properties(local_filter_props,
host)
# Populate the request_spec with the local_filter_props information
# like retries and limits. Note that at this point the request_spec
# could have come from a compute via reschedule and it would
# already have some things set, like scheduler_hints.
local_reqspec = objects.RequestSpec.from_primitives(
context, legacy_request_spec, local_filter_props)
# NOTE(gibi): at this point the request spec already got converted
# to a legacy dict and then back to an object so we lost the non
# legacy part of the spec. Re-populate the requested_resources
# field based on the original request spec object passed to this
# function.
if request_spec:
local_reqspec.requested_resources = (
request_spec.requested_resources)
# The block_device_mapping passed from the api doesn't contain
# instance specific information
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
# This is populated in scheduler_utils.populate_retry
num_attempts = local_filter_props.get('retry',
{}).get('num_attempts', 1)
if num_attempts <= 1:
# If this is a reschedule the instance is already mapped to
# this cell and the BuildRequest is already deleted so ignore
# the logic below.
inst_mapping = self._populate_instance_mapping(context,
instance,
host)
try:
self._destroy_build_request(context, instance)
except exception.BuildRequestNotFound:
# This indicates an instance delete has been requested in
# the API. Stop the build, cleanup the instance_mapping and
# potentially the block_device_mappings
# TODO(alaski): Handle block_device_mapping cleanup
if inst_mapping:
inst_mapping.destroy()
return
else:
# NOTE(lyarwood): If this is a reschedule then recreate any
# attachments that were previously removed when cleaning up
# after failures to spawn etc.
self._validate_existing_attachment_ids(context, instance, bdms)
alts = [(alt.service_host, alt.nodename) for alt in host_list]
LOG.debug("Selected host: %s; Selected node: %s; Alternates: %s",
host.service_host, host.nodename, alts, instance=instance)
try:
accel_uuids = self._create_and_bind_arq_for_instance(
context, instance, host.nodename, local_reqspec,
requested_networks)
except Exception as exc:
LOG.exception('Failed to reschedule. Reason: %s', exc)
self._cleanup_when_reschedule_fails(
context, instance, exc, legacy_request_spec,
requested_networks)
continue
self.compute_rpcapi.build_and_run_instance(context,
instance=instance, host=host.service_host, image=image,
request_spec=local_reqspec,
filter_properties=local_filter_props,
admin_password=admin_password,
injected_files=injected_files,
requested_networks=requested_networks,
security_groups=security_groups,
block_device_mapping=bdms, node=host.nodename,
limits=host.limits, host_list=host_list,
accel_uuids=accel_uuids)
def _create_and_bind_arq_for_instance(
self, context, instance, hostname,
request_spec, requested_networks=None):
try:
resource_provider_mapping = (
request_spec.get_request_group_mapping())
# Using nodename instead of hostname. See:
# http://lists.openstack.org/pipermail/openstack-discuss/2019-November/011044.html # noqa
cyclient = cyborg.get_client(context)
bindings = {}
port_bindings = {}
# Create ARQs comes from extra specs.
bindings = self._create_and_bind_arqs(
cyclient, instance.uuid, instance.flavor.extra_specs,
hostname, resource_provider_mapping)
if requested_networks:
# Create ARQs comes from port device profile
port_bindings = self._create_arqs_for_ports(
cyclient, instance.uuid, requested_networks,
hostname, resource_provider_mapping)
# Initiate Cyborg binding asynchronously
bindings.update(port_bindings)
if bindings:
cyclient.bind_arqs(bindings)
return list(bindings.keys())
except exception.AcceleratorRequestBindingFailed as exc:
# If anything failed here we need to cleanup and bail out.
cyclient = cyborg.get_client(context)
cyclient.delete_arqs_by_uuid(exc.arqs)
raise
def _schedule_instances(self, context, request_spec,
instance_uuids=None, return_alternates=False):
scheduler_utils.setup_instance_group(context, request_spec)
with timeutils.StopWatch() as timer:
host_lists = self.query_client.select_destinations(
context, request_spec, instance_uuids, return_objects=True,
return_alternates=return_alternates)
LOG.debug('Took %0.2f seconds to select destinations for %s '
'instance(s).', timer.elapsed(), len(instance_uuids))
return host_lists
@staticmethod
def _restrict_request_spec_to_cell(context, instance, request_spec):
"""Sets RequestSpec.requested_destination.cell for the move operation
Move operations, e.g. evacuate and unshelve, must be restricted to the
cell in which the instance already exists, so this method is used to
target the RequestSpec, which is sent to the scheduler via the
_schedule_instances method, to the instance's current cell.
:param context: nova auth RequestContext
"""
instance_mapping = \
objects.InstanceMapping.get_by_instance_uuid(
context, instance.uuid)
LOG.debug('Requesting cell %(cell)s during scheduling',
{'cell': instance_mapping.cell_mapping.identity},
instance=instance)
if ('requested_destination' in request_spec and
request_spec.requested_destination):
request_spec.requested_destination.cell = (
instance_mapping.cell_mapping)
else:
request_spec.requested_destination = (
objects.Destination(
cell=instance_mapping.cell_mapping))
# TODO(mriedem): Make request_spec required in ComputeTaskAPI RPC v2.0.
@targets_cell
def unshelve_instance(self, context, instance, request_spec=None):
sys_meta = instance.system_metadata
def safe_image_show(ctx, image_id):
if image_id:
return self.image_api.get(ctx, image_id, show_deleted=False)
else:
raise exception.ImageNotFound(image_id='')
if instance.vm_state == vm_states.SHELVED:
instance.task_state = task_states.POWERING_ON
instance.save(expected_task_state=task_states.UNSHELVING)
self.compute_rpcapi.start_instance(context, instance)
elif instance.vm_state == vm_states.SHELVED_OFFLOADED:
image = None
image_id = sys_meta.get('shelved_image_id')
# No need to check for image if image_id is None as
# "shelved_image_id" key is not set for volume backed
# instance during the shelve process
if image_id:
with compute_utils.EventReporter(
context, 'get_image_info', self.host, instance.uuid):
try:
image = safe_image_show(context, image_id)
except exception.ImageNotFound as error:
instance.vm_state = vm_states.ERROR