def _execute(self):
image = self.request_spec.get("image")
self.quotas = objects.Quotas.from_reservations(self.context, self.reservations, instance=self.instance)
scheduler_utils.setup_instance_group(self.context, self.request_spec, self.filter_properties)
scheduler_utils.populate_retry(self.filter_properties, self.instance.uuid)
hosts = self.scheduler_client.select_destinations(self.context, self.request_spec, self.filter_properties)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(self.filter_properties, host_state)
# context is not serializable
self.filter_properties.pop("context", None)
(host, node) = (host_state["host"], host_state["nodename"])
self.compute_rpcapi.prep_resize(
self.context,
image,
self.instance,
self.flavor,
host,
self.reservations,
request_spec=self.request_spec,
filter_properties=self.filter_properties,
node=node,
clean_shutdown=self.clean_shutdown,
)
def _execute(self):
self.quotas = objects.Quotas.from_reservations(self.context,
self.reservations,
instance=self.instance)
# TODO(sbauza): Remove that once prep_resize() accepts a RequestSpec
# object in the signature and all the scheduler.utils methods too
legacy_spec = self.request_spec.to_legacy_request_spec_dict()
legacy_props = self.request_spec.to_legacy_filter_properties_dict()
scheduler_utils.setup_instance_group(self.context, legacy_spec,
legacy_props)
scheduler_utils.populate_retry(legacy_props,
self.instance.uuid)
# NOTE(sbauza): Force_hosts/nodes needs to be reset
# if we want to make sure that the next destination
# is not forced to be the original host
self.request_spec.reset_forced_destinations()
# NOTE(danms): Right now we only support migrate to the same
# cell as the current instance, so request that the scheduler
# limit thusly.
instance_mapping = objects.InstanceMapping.get_by_instance_uuid(
self.context, self.instance.uuid)
LOG.debug('Requesting cell %(cell)s while migrating',
{'cell': instance_mapping.cell_mapping.identity},
instance=self.instance)
if ('requested_destination' in self.request_spec and
self.request_spec.requested_destination):
self.request_spec.requested_destination.cell = (
instance_mapping.cell_mapping)
else:
self.request_spec.requested_destination = objects.Destination(
cell=instance_mapping.cell_mapping)
hosts = self.scheduler_client.select_destinations(
self.context, self.request_spec)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(legacy_props,
host_state)
# context is not serializable
legacy_props.pop('context', None)
(host, node) = (host_state['host'], host_state['nodename'])
self.instance.availability_zone = (
availability_zones.get_host_availability_zone(
self.context, host))
# FIXME(sbauza): Serialize/Unserialize the legacy dict because of
# oslo.messaging #1529084 to transform datetime values into strings.
# tl;dr: datetimes in dicts are not accepted as correct values by the
# rpc fake driver.
legacy_spec = jsonutils.loads(jsonutils.dumps(legacy_spec))
self.compute_rpcapi.prep_resize(
self.context, self.instance, legacy_spec['image'],
self.flavor, host, self.reservations,
request_spec=legacy_spec, filter_properties=legacy_props,
node=node, clean_shutdown=self.clean_shutdown)
def _execute(self):
image = self.request_spec.get('image')
self.quotas = objects.Quotas.from_reservations(self.context,
self.reservations,
instance=self.instance)
scheduler_utils.setup_instance_group(self.context, self.request_spec,
self.filter_properties)
scheduler_utils.populate_retry(self.filter_properties,
self.instance.uuid)
# TODO(sbauza): Hydrate here the object until we modify the
# scheduler.utils methods to directly use the RequestSpec object
spec_obj = objects.RequestSpec.from_primitives(
self.context, self.request_spec, self.filter_properties)
hosts = self.scheduler_client.select_destinations(
self.context, spec_obj)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(self.filter_properties,
host_state)
# context is not serializable
self.filter_properties.pop('context', None)
(host, node) = (host_state['host'], host_state['nodename'])
self.compute_rpcapi.prep_resize(
self.context, image, self.instance, self.flavor, host,
self.reservations, request_spec=self.request_spec,
filter_properties=self.filter_properties, node=node,
clean_shutdown=self.clean_shutdown)
def unshelve_instance(self, context, instance):
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", instance.uuid):
try:
image = safe_image_show(context, image_id)
except exception.ImageNotFound:
instance.vm_state = vm_states.ERROR
instance.save()
reason = _("Unshelve attempted but the image %s " "cannot be found.") % image_id
LOG.error(reason, instance=instance)
raise exception.UnshelveException(instance_id=instance.uuid, reason=reason)
try:
with compute_utils.EventReporter(context, "schedule_instances", instance.uuid):
filter_properties = {}
scheduler_utils.populate_retry(filter_properties, instance.uuid)
request_spec = scheduler_utils.build_request_spec(context, image, [instance])
hosts = self._schedule_instances(context, request_spec, filter_properties)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(filter_properties, host_state)
(host, node) = (host_state["host"], host_state["nodename"])
self.compute_rpcapi.unshelve_instance(
context, instance, host, image=image, filter_properties=filter_properties, node=node
)
except (exception.NoValidHost, exception.UnsupportedPolicyException):
instance.task_state = None
instance.save()
LOG.warning(_LW("No valid host found for unshelve instance"), instance=instance)
return
except Exception:
with excutils.save_and_reraise_exception():
instance.task_state = None
instance.save()
LOG.error(_LE("Unshelve attempted but an error " "has occurred"), instance=instance)
else:
LOG.error(_LE("Unshelve attempted but vm_state not SHELVED or " "SHELVED_OFFLOADED"), instance=instance)
instance.vm_state = vm_states.ERROR
instance.save()
return
def _cold_migrate(self, context, instance, flavor, filter_properties,
reservations):
image_ref = instance.image_ref
image = compute_utils.get_image_metadata(
context, self.image_api, image_ref, instance)
request_spec = scheduler_utils.build_request_spec(
context, image, [instance], instance_type=flavor)
quotas = objects.Quotas.from_reservations(context,
reservations,
instance=instance)
scheduler_utils.setup_instance_group(context, request_spec,
filter_properties)
try:
scheduler_utils.populate_retry(filter_properties, instance['uuid'])
hosts = self.scheduler_client.select_destinations(
context, request_spec, filter_properties)
host_state = hosts[0]
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, 'migrate_server',
updates, ex, request_spec)
quotas.rollback()
# 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)
try:
scheduler_utils.populate_filter_properties(filter_properties,
host_state)
# context is not serializable
filter_properties.pop('context', None)
# TODO(timello): originally, instance_type in request_spec
# on compute.api.resize does not have 'extra_specs', so we
# remove it for now to keep tests backward compatibility.
request_spec['instance_type'].pop('extra_specs', None)
(host, node) = (host_state['host'], host_state['nodename'])
self.compute_rpcapi.prep_resize(
context, image, instance,
flavor, host,
reservations, request_spec=request_spec,
filter_properties=filter_properties, node=node)
except Exception as ex:
with excutils.save_and_reraise_exception():
updates = {'vm_state': instance['vm_state'],
'task_state': None}
self._set_vm_state_and_notify(context, 'migrate_server',
updates, ex, request_spec)
quotas.rollback()
def _cold_migrate(self, context, instance, flavor, filter_properties, reservations, clean_shutdown):
image_ref = instance.image_ref
image = compute_utils.get_image_metadata(context, self.image_api, image_ref, instance)
request_spec = scheduler_utils.build_request_spec(context, image, [instance], instance_type=flavor)
quotas = objects.Quotas.from_reservations(context, reservations, instance=instance)
try:
scheduler_utils.setup_instance_group(context, request_spec, filter_properties)
scheduler_utils.populate_retry(filter_properties, instance["uuid"])
hosts = self.scheduler_client.select_destinations(context, request_spec, filter_properties)
host_state = hosts[0]
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)
quotas.rollback()
# 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)
quotas.rollback()
try:
scheduler_utils.populate_filter_properties(filter_properties, host_state)
# context is not serializable
filter_properties.pop("context", None)
(host, node) = (host_state["host"], host_state["nodename"])
self.compute_rpcapi.prep_resize(
context,
image,
instance,
flavor,
host,
reservations,
request_spec=request_spec,
filter_properties=filter_properties,
node=node,
clean_shutdown=clean_shutdown,
)
except Exception as ex:
with excutils.save_and_reraise_exception():
updates = {"vm_state": instance.vm_state, "task_state": None}
self._set_vm_state_and_notify(context, instance.uuid, "migrate_server", updates, ex, request_spec)
quotas.rollback()
def build_instances(self, context, instances, image, filter_properties,
admin_password, injected_files, requested_networks,
security_groups, block_device_mapping=None, legacy_bdm=True):
# TODO(ndipanov): Remove block_device_mapping and legacy_bdm in version
# 2.0 of the RPC API.
request_spec = scheduler_utils.build_request_spec(context, image,
instances)
scheduler_utils.setup_instance_group(context, request_spec,
filter_properties)
# 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(
objects=[objects.NetworkRequest.from_tuple(t)
for t in requested_networks])
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.
scheduler_utils.populate_retry(filter_properties,
instances[0].uuid)
hosts = self.scheduler_client.select_destinations(context,
request_spec, filter_properties)
except Exception as exc:
for instance in instances:
scheduler_driver.handle_schedule_error(context, exc,
instance.uuid, request_spec)
return
for (instance, host) in itertools.izip(instances, hosts):
try:
instance.refresh()
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)
# The block_device_mapping passed from the api doesn't contain
# instance specific information
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
self.compute_rpcapi.build_and_run_instance(context,
instance=instance, host=host['host'], image=image,
request_spec=request_spec,
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'])
def _cold_migrate(self, context, instance, flavor, filter_properties, reservations):
image_ref = instance.image_ref
image = compute_utils.get_image_metadata(context, self.image_api, image_ref, instance)
request_spec = scheduler_utils.build_request_spec(context, image, [instance], instance_type=flavor)
quotas = quotas_obj.Quotas.from_reservations(context, reservations, instance=instance)
try:
scheduler_utils.populate_retry(filter_properties, instance["uuid"])
hosts = self.scheduler_rpcapi.select_destinations(context, request_spec, filter_properties)
host_state = hosts[0]
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, "migrate_server", updates, ex, request_spec)
quotas.rollback()
msg = _("No valid host found for cold migrate")
raise exception.NoValidHost(reason=msg)
try:
scheduler_utils.populate_filter_properties(filter_properties, host_state)
# context is not serializable
filter_properties.pop("context", None)
# TODO(timello): originally, instance_type in request_spec
# on compute.api.resize does not have 'extra_specs', so we
# remove it for now to keep tests backward compatibility.
request_spec["instance_type"].pop("extra_specs")
(host, node) = (host_state["host"], host_state["nodename"])
self.compute_rpcapi.prep_resize(
context,
image,
instance,
flavor,
host,
reservations,
request_spec=request_spec,
filter_properties=filter_properties,
node=node,
)
except Exception as ex:
with excutils.save_and_reraise_exception():
updates = {"vm_state": instance["vm_state"], "task_state": None}
self._set_vm_state_and_notify(context, "migrate_server", updates, ex, request_spec)
quotas.rollback()
def _execute(self):
image = self.request_spec.image
self.quotas = objects.Quotas.from_reservations(self.context,
self.reservations,
instance=self.instance)
# TODO(sbauza): Remove that once prep_resize() accepts a RequestSpec
# object in the signature and all the scheduler.utils methods too
legacy_spec = self.request_spec.to_legacy_request_spec_dict()
legacy_props = self.request_spec.to_legacy_filter_properties_dict()
scheduler_utils.setup_instance_group(self.context, legacy_spec,
legacy_props)
scheduler_utils.populate_retry(legacy_props,
self.instance.uuid)
# TODO(sbauza): Remove that RequestSpec rehydratation once
# scheduler.utils methods use directly the NovaObject.
self.request_spec = objects.RequestSpec.from_components(
self.context, self.instance.uuid, image,
self.flavor, self.instance.numa_topology,
self.instance.pci_requests, legacy_props, None,
self.instance.availability_zone)
# NOTE(sbauza): Force_hosts/nodes needs to be reset
# if we want to make sure that the next destination
# is not forced to be the original host
self.request_spec.reset_forced_destinations()
hosts = self.scheduler_client.select_destinations(
self.context, self.request_spec)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(legacy_props,
host_state)
# context is not serializable
legacy_props.pop('context', None)
(host, node) = (host_state['host'], host_state['nodename'])
# FIXME(sbauza): Serialize/Unserialize the legacy dict because of
# oslo.messaging #1529084 to transform datetime values into strings.
# tl;dr: datetimes in dicts are not accepted as correct values by the
# rpc fake driver.
legacy_spec = jsonutils.loads(jsonutils.dumps(legacy_spec))
self.compute_rpcapi.prep_resize(
self.context, self.instance, legacy_spec['image'],
self.flavor, host, self.reservations,
request_spec=legacy_spec, filter_properties=legacy_props,
node=node, clean_shutdown=self.clean_shutdown)
def _execute(self):
image = self.request_spec.get('image')
self.quotas = objects.Quotas.from_reservations(self.context,
self.reservations,
instance=self.instance)
scheduler_utils.setup_instance_group(self.context, self.request_spec,
self.filter_properties)
scheduler_utils.populate_retry(self.filter_properties,
self.instance.uuid)
# TODO(sbauza): Hydrate here the object until we modify the
# scheduler.utils methods to directly use the RequestSpec object
spec_obj = objects.RequestSpec.from_primitives(
self.context, self.request_spec, self.filter_properties)
hosts = self.scheduler_client.select_destinations(
self.context, spec_obj)
LOG.info('TTTType hosts:%s' % hosts)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(self.filter_properties,
host_state)
# context is not serializable
self.filter_properties.pop('context', None)
#(host, node) = (host_state['host'], host_state['nodename'])
LOG.info('TTTType host_state:%s' % host_state)
# add by jiahua
#(host, node) = (self.filter_properties['destination_host_name'],
# self.filter_properties['destination_host_name'])
# specify host use specify host, else use localhost
#node = self.filter_properties['destination_host_name'] if self.filter_properties['destination_host_name'] \
# else host_state['nodename']
#host = self.filter_properties['destination_host_name'] if self.filter_properties['destination_host_name'] \
# else host_state['host']
node = self.filter_properties['destination_host_name'] if self.filter_properties['destination_host_name'] \
else self.filter_properties['instance_host_name']
host = self.filter_properties['destination_host_name'] if self.filter_properties['destination_host_name'] \
else self.filter_properties['instance_host_name']
LOG.info('TTTType node:%s' % node)
LOG.info('TTTType host:%s' % host)
self.compute_rpcapi.prep_resize(
self.context, image, self.instance, self.flavor, host,
self.reservations, request_spec=self.request_spec,
filter_properties=self.filter_properties, node=node,
clean_shutdown=self.clean_shutdown)
def _execute(self):
# TODO(sbauza): Remove that once prep_resize() accepts a RequestSpec
# object in the signature and all the scheduler.utils methods too
legacy_spec = self.request_spec.to_legacy_request_spec_dict()
legacy_props = self.request_spec.to_legacy_filter_properties_dict()
scheduler_utils.setup_instance_group(self.context, self.request_spec)
# If a target host is set in a requested destination,
# 'populate_retry' need not be executed.
if not ('requested_destination' in self.request_spec
and self.request_spec.requested_destination
and 'host' in self.request_spec.requested_destination):
scheduler_utils.populate_retry(legacy_props, self.instance.uuid)
# NOTE(sbauza): Force_hosts/nodes needs to be reset
# if we want to make sure that the next destination
# is not forced to be the original host
self.request_spec.reset_forced_destinations()
# NOTE(danms): Right now we only support migrate to the same
# cell as the current instance, so request that the scheduler
# limit thusly.
instance_mapping = objects.InstanceMapping.get_by_instance_uuid(
self.context, self.instance.uuid)
LOG.debug('Requesting cell %(cell)s while migrating',
{'cell': instance_mapping.cell_mapping.identity},
instance=self.instance)
if ('requested_destination' in self.request_spec
and self.request_spec.requested_destination):
self.request_spec.requested_destination.cell = (
instance_mapping.cell_mapping)
# NOTE(takashin): In the case that the target host is specified,
# if the migration is failed, it is not necessary to retry
# the cold migration to the same host. So make sure that
# reschedule will not occur.
if 'host' in self.request_spec.requested_destination:
legacy_props.pop('retry', None)
self.request_spec.retry = None
else:
self.request_spec.requested_destination = objects.Destination(
cell=instance_mapping.cell_mapping)
# Once _preallocate_migration() is done, the source node allocation is
# moved from the instance consumer to the migration record consumer,
# and the instance consumer doesn't have any allocations. If this is
# the first time through here (not a reschedule), select_destinations
# below will allocate resources on the selected destination node for
# the instance consumer. If we're rescheduling, host_list is not None
# and we'll call claim_resources for the instance and the selected
# alternate. If we exhaust our alternates and raise MaxRetriesExceeded,
# the rollback() method should revert the allocation swaparoo and move
# the source node allocation from the migration record back to the
# instance record.
migration = self._preallocate_migration()
self.request_spec.ensure_project_id(self.instance)
# On an initial call to migrate, 'self.host_list' will be None, so we
# have to call the scheduler to get a list of acceptable hosts to
# migrate to. That list will consist of a selected host, along with
# zero or more alternates. On a reschedule, though, the alternates will
# be passed to this object and stored in 'self.host_list', so we can
# pop the first alternate from the list to use for the destination, and
# pass the remaining alternates to the compute.
if self.host_list is None:
selection_lists = self.scheduler_client.select_destinations(
self.context,
self.request_spec, [self.instance.uuid],
return_objects=True,
return_alternates=True)
# Since there is only ever one instance to migrate per call, we
# just need the first returned element.
selection_list = selection_lists[0]
# The selected host is the first item in the list, with the
# alternates being the remainder of the list.
selection, self.host_list = selection_list[0], selection_list[1:]
else:
# This is a reschedule that will use the supplied alternate hosts
# in the host_list as destinations. Since the resources on these
# alternates may have been consumed and might not be able to
# support the migrated instance, we need to first claim the
# resources to verify the host still has sufficient availabile
# resources.
elevated = self.context.elevated()
host_available = False
while self.host_list and not host_available:
selection = self.host_list.pop(0)
if selection.allocation_request:
alloc_req = jsonutils.loads(selection.allocation_request)
else:
alloc_req = None
if alloc_req:
# If this call succeeds, the resources on the destination
# host will be claimed by the instance.
host_available = scheduler_utils.claim_resources(
elevated, self.reportclient, self.request_spec,
self.instance.uuid, alloc_req,
selection.allocation_request_version)
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
# There are no more available hosts. Raise a MaxRetriesExceeded
# exception in that case.
if not host_available:
reason = ("Exhausted all hosts available for retrying build "
"failures for instance %(instance_uuid)s." % {
"instance_uuid": self.instance.uuid
})
raise exception.MaxRetriesExceeded(reason=reason)
scheduler_utils.populate_filter_properties(legacy_props, selection)
# context is not serializable
legacy_props.pop('context', None)
(host, node) = (selection.service_host, selection.nodename)
self.instance.availability_zone = (
availability_zones.get_host_availability_zone(self.context, host))
# FIXME(sbauza): Serialize/Unserialize the legacy dict because of
# oslo.messaging #1529084 to transform datetime values into strings.
# tl;dr: datetimes in dicts are not accepted as correct values by the
# rpc fake driver.
legacy_spec = jsonutils.loads(jsonutils.dumps(legacy_spec))
LOG.debug(
"Calling prep_resize with selected host: %s; "
"Selected node: %s; Alternates: %s",
host,
node,
self.host_list,
instance=self.instance)
# RPC cast to the destination host to start the migration process.
self.compute_rpcapi.prep_resize(self.context,
self.instance,
legacy_spec['image'],
self.flavor,
host,
migration,
request_spec=legacy_spec,
filter_properties=legacy_props,
node=node,
clean_shutdown=self.clean_shutdown,
host_list=self.host_list)
def build_instances(self, context, instances, image, filter_properties,
admin_password, injected_files, requested_networks,
security_groups, block_device_mapping=None, legacy_bdm=True):
# 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)
request_spec = {}
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.
request_spec = scheduler_utils.build_request_spec(
context, image, instances)
scheduler_utils.populate_retry(
filter_properties, instances[0].uuid)
hosts = self._schedule_instances(
context, request_spec, filter_properties)
except Exception as exc:
updates = {'vm_state': vm_states.ERROR, 'task_state': None}
for instance in instances:
self._set_vm_state_and_notify(
context, instance.uuid, 'build_instances', updates,
exc, request_spec)
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_allocated_networks(
context, instance, requested_networks)
return
for (instance, host) in six.moves.zip(instances, hosts):
try:
instance.refresh()
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)
# 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
self.compute_rpcapi.build_and_run_instance(context,
instance=instance, host=host['host'], image=image,
request_spec=request_spec,
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'])
def schedule_run_instance(self, context, request_spec, admin_password,
injected_files, requested_networks,
is_first_time, filter_properties,
legacy_bdm_in_spec):
"""Provisions instances that needs to be scheduled
Applies filters and weighters on request properties to get a list of
compute hosts and calls them to spawn instance(s).
"""
payload = dict(request_spec=request_spec)
self.notifier.info(context, 'scheduler.run_instance.start', payload)
instance_uuids = request_spec.get('instance_uuids')
LOG.info(
_("Attempting to build %(num_instances)d instance(s) "
"uuids: %(instance_uuids)s"), {
'num_instances': len(instance_uuids),
'instance_uuids': instance_uuids
})
LOG.debug("Request Spec: %s" % request_spec)
# check retry policy. Rather ugly use of instance_uuids[0]...
# but if we've exceeded max retries... then we really only
# have a single instance.
scheduler_utils.populate_retry(filter_properties, instance_uuids[0])
weighed_hosts = self._schedule(context, request_spec,
filter_properties)
# NOTE: Pop instance_uuids as individual creates do not need the
# set of uuids. Do not pop before here as the upper exception
# handler fo NoValidHost needs the uuid to set error state
instance_uuids = request_spec.pop('instance_uuids')
# NOTE(comstud): Make sure we do not pass this through. It
# contains an instance of RpcContext that cannot be serialized.
filter_properties.pop('context', None)
for num, instance_uuid in enumerate(instance_uuids):
request_spec['instance_properties']['launch_index'] = num
try:
try:
weighed_host = weighed_hosts.pop(0)
LOG.info(
_("Choosing host %(weighed_host)s "
"for instance %(instance_uuid)s"), {
'weighed_host': weighed_host,
'instance_uuid': instance_uuid
})
except IndexError:
raise exception.NoValidHost(reason="")
self._provision_resource(context,
weighed_host,
request_spec,
filter_properties,
requested_networks,
injected_files,
admin_password,
is_first_time,
instance_uuid=instance_uuid,
legacy_bdm_in_spec=legacy_bdm_in_spec)
except Exception as ex:
# NOTE(vish): we don't reraise the exception here to make sure
# that all instances in the request get set to
# error properly
driver.handle_schedule_error(context, ex, instance_uuid,
request_spec)
# scrub retry host list in case we're scheduling multiple
# instances:
retry = filter_properties.get('retry', {})
retry['hosts'] = []
self.notifier.info(context, 'scheduler.run_instance.end', payload)
def _execute(self):
self.quotas = objects.Quotas.from_reservations(self.context,
self.reservations,
instance=self.instance)
# TODO(sbauza): Remove that once prep_resize() accepts a RequestSpec
# object in the signature and all the scheduler.utils methods too
legacy_spec = self.request_spec.to_legacy_request_spec_dict()
legacy_props = self.request_spec.to_legacy_filter_properties_dict()
scheduler_utils.setup_instance_group(self.context, legacy_spec,
legacy_props)
scheduler_utils.populate_retry(legacy_props, self.instance.uuid)
# NOTE(sbauza): Force_hosts/nodes needs to be reset
# if we want to make sure that the next destination
# is not forced to be the original host
self.request_spec.reset_forced_destinations()
# NOTE(danms): Right now we only support migrate to the same
# cell as the current instance, so request that the scheduler
# limit thusly.
instance_mapping = objects.InstanceMapping.get_by_instance_uuid(
self.context, self.instance.uuid)
LOG.debug('Requesting cell %(cell)s while migrating',
{'cell': instance_mapping.cell_mapping.identity},
instance=self.instance)
if ('requested_destination' in self.request_spec
and self.request_spec.requested_destination):
self.request_spec.requested_destination.cell = (
instance_mapping.cell_mapping)
else:
self.request_spec.requested_destination = objects.Destination(
cell=instance_mapping.cell_mapping)
hosts = self.scheduler_client.select_destinations(
self.context, self.request_spec, [self.instance.uuid])
host_state = hosts[0]
scheduler_utils.populate_filter_properties(legacy_props, host_state)
# context is not serializable
legacy_props.pop('context', None)
(host, node) = (host_state['host'], host_state['nodename'])
self.instance.availability_zone = (
availability_zones.get_host_availability_zone(self.context, host))
# FIXME(sbauza): Serialize/Unserialize the legacy dict because of
# oslo.messaging #1529084 to transform datetime values into strings.
# tl;dr: datetimes in dicts are not accepted as correct values by the
# rpc fake driver.
legacy_spec = jsonutils.loads(jsonutils.dumps(legacy_spec))
self.compute_rpcapi.prep_resize(self.context,
self.instance,
legacy_spec['image'],
self.flavor,
host,
self.reservations,
request_spec=legacy_spec,
filter_properties=legacy_props,
node=node,
clean_shutdown=self.clean_shutdown)
def _execute(self):
# NOTE(sbauza): Force_hosts/nodes needs to be reset if we want to make
# sure that the next destination is not forced to be the original host.
# This needs to be done before the populate_retry call otherwise
# retries will be disabled if the server was created with a forced
# host/node.
self.request_spec.reset_forced_destinations()
# TODO(sbauza): Remove once all the scheduler.utils methods accept a
# RequestSpec object in the signature.
legacy_props = self.request_spec.to_legacy_filter_properties_dict()
scheduler_utils.setup_instance_group(self.context, self.request_spec)
# If a target host is set in a requested destination,
# 'populate_retry' need not be executed.
if not ('requested_destination' in self.request_spec
and self.request_spec.requested_destination
and 'host' in self.request_spec.requested_destination):
scheduler_utils.populate_retry(legacy_props, self.instance.uuid)
port_res_req = self.network_api.get_requested_resource_for_instance(
self.context, self.instance.uuid)
# NOTE(gibi): When cyborg or other module wants to handle similar
# non-nova resources then here we have to collect all the external
# resource requests in a single list and add them to the RequestSpec.
self.request_spec.requested_resources = port_res_req
self._restrict_request_spec_to_cell(legacy_props)
# Once _preallocate_migration() is done, the source node allocation is
# moved from the instance consumer to the migration record consumer,
# and the instance consumer doesn't have any allocations. If this is
# the first time through here (not a reschedule), select_destinations
# below will allocate resources on the selected destination node for
# the instance consumer. If we're rescheduling, host_list is not None
# and we'll call claim_resources for the instance and the selected
# alternate. If we exhaust our alternates and raise MaxRetriesExceeded,
# the rollback() method should revert the allocation swaparoo and move
# the source node allocation from the migration record back to the
# instance record.
migration = self._preallocate_migration()
self.request_spec.ensure_project_and_user_id(self.instance)
self.request_spec.ensure_network_metadata(self.instance)
compute_utils.heal_reqspec_is_bfv(self.context, self.request_spec,
self.instance)
# On an initial call to migrate, 'self.host_list' will be None, so we
# have to call the scheduler to get a list of acceptable hosts to
# migrate to. That list will consist of a selected host, along with
# zero or more alternates. On a reschedule, though, the alternates will
# be passed to this object and stored in 'self.host_list', so we can
# pop the first alternate from the list to use for the destination, and
# pass the remaining alternates to the compute.
if self.host_list is None:
selection = self._schedule()
else:
# This is a reschedule that will use the supplied alternate hosts
# in the host_list as destinations.
selection = self._reschedule()
scheduler_utils.populate_filter_properties(legacy_props, selection)
# context is not serializable
legacy_props.pop('context', None)
(host, node) = (selection.service_host, selection.nodename)
self.instance.availability_zone = (
availability_zones.get_host_availability_zone(self.context, host))
LOG.debug(
"Calling prep_resize with selected host: %s; "
"Selected node: %s; Alternates: %s",
host,
node,
self.host_list,
instance=self.instance)
# RPC cast to the destination host to start the migration process.
self.compute_rpcapi.prep_resize(
# NOTE(mriedem): Using request_spec.image here is potentially
# dangerous if it is not kept up to date (i.e. rebuild/unshelve);
# seems like the sane thing to do would be to pass the current
# instance.image_meta since that is what MoveClaim will use for
# any NUMA topology claims on the destination host...
self.context,
self.instance,
self.request_spec.image,
self.flavor,
host,
migration,
request_spec=self.request_spec,
filter_properties=legacy_props,
node=node,
clean_shutdown=self.clean_shutdown,
host_list=self.host_list)
def _cold_migrate(self, context, instance, flavor, filter_properties,
reservations, clean_shutdown):
image_ref = instance.image_ref
image = compute_utils.get_image_metadata(context, self.image_api,
image_ref, instance)
request_spec = scheduler_utils.build_request_spec(context,
image, [instance],
instance_type=flavor)
quotas = objects.Quotas.from_reservations(context,
reservations,
instance=instance)
try:
scheduler_utils.setup_instance_group(context, request_spec,
filter_properties)
scheduler_utils.populate_retry(filter_properties, instance['uuid'])
hosts = self.scheduler_client.select_destinations(
context, request_spec, filter_properties)
host_state = hosts[0]
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)
quotas.rollback()
# 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)
quotas.rollback()
try:
scheduler_utils.populate_filter_properties(filter_properties,
host_state)
# context is not serializable
filter_properties.pop('context', None)
(host, node) = (host_state['host'], host_state['nodename'])
self.compute_rpcapi.prep_resize(
context,
image,
instance,
flavor,
host,
reservations,
request_spec=request_spec,
filter_properties=filter_properties,
node=node,
clean_shutdown=clean_shutdown)
except Exception as ex:
with excutils.save_and_reraise_exception():
updates = {'vm_state': instance.vm_state, 'task_state': None}
self._set_vm_state_and_notify(context, instance.uuid,
'migrate_server', updates, ex,
request_spec)
quotas.rollback()
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',
instance.uuid):
try:
image = safe_image_show(context, image_id)
except exception.ImageNotFound:
instance.vm_state = vm_states.ERROR
instance.save()
reason = _('Unshelve attempted but the image %s '
'cannot be found.') % image_id
LOG.error(reason, instance=instance)
raise exception.UnshelveException(
instance_id=instance.uuid, reason=reason)
try:
with compute_utils.EventReporter(context, 'schedule_instances',
instance.uuid):
if not request_spec:
# NOTE(sbauza): We were unable to find an original
# RequestSpec object - probably because the instance is
# old. We need to mock that the old way
filter_properties = {}
request_spec = scheduler_utils.build_request_spec(
context, image, [instance])
else:
# NOTE(sbauza): Force_hosts/nodes needs to be reset
# if we want to make sure that the next destination
# is not forced to be the original host
request_spec.reset_forced_destinations()
# TODO(sbauza): Provide directly the RequestSpec object
# when _schedule_instances(),
# populate_filter_properties and populate_retry()
# accept it
filter_properties = request_spec.\
to_legacy_filter_properties_dict()
request_spec = request_spec.\
to_legacy_request_spec_dict()
scheduler_utils.populate_retry(filter_properties,
instance.uuid)
hosts = self._schedule_instances(context, request_spec,
filter_properties)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(
filter_properties, host_state)
(host, node) = (host_state['host'], host_state['nodename'])
self.compute_rpcapi.unshelve_instance(
context,
instance,
host,
image=image,
filter_properties=filter_properties,
node=node)
except (exception.NoValidHost,
exception.UnsupportedPolicyException):
instance.task_state = None
instance.save()
LOG.warning(_LW("No valid host found for unshelve instance"),
instance=instance)
return
except Exception:
with excutils.save_and_reraise_exception():
instance.task_state = None
instance.save()
LOG.error(_LE("Unshelve attempted but an error "
"has occurred"),
instance=instance)
else:
LOG.error(_LE('Unshelve attempted but vm_state not SHELVED or '
'SHELVED_OFFLOADED'),
instance=instance)
instance.vm_state = vm_states.ERROR
instance.save()
return
def build_instances(self, context, instances, image, filter_properties,
admin_password, injected_files, requested_networks,
security_groups, block_device_mapping=None, legacy_bdm=True):
# TODO(ndipanov): Remove block_device_mapping and legacy_bdm in version
# 2.0 of the RPC API.
request_spec = scheduler_utils.build_request_spec(context, image,
instances)
# 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(
objects=[objects.NetworkRequest.from_tuple(t)
for t in 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)
try:
scheduler_utils.setup_instance_group(context, request_spec,
filter_properties)
# check retry policy. Rather ugly use of instances[0]...
# but if we've exceeded max retries... then we really only
# have a single instance.
scheduler_utils.populate_retry(filter_properties,
instances[0].uuid)
hosts = self.scheduler_client.select_destinations(context,
request_spec, filter_properties)
except Exception as exc:
updates = {'vm_state': vm_states.ERROR, 'task_state': None}
for instance in instances:
self._set_vm_state_and_notify(
context, instance.uuid, 'build_instances', updates,
exc, request_spec)
return
for (instance, host) in itertools.izip(instances, hosts):
try:
instance.refresh()
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)
# The block_device_mapping passed from the api doesn't contain
# instance specific information
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
self.compute_rpcapi.build_and_run_instance(context,
instance=instance, host=host['host'], image=image,
request_spec=request_spec,
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'])
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', instance.uuid):
try:
image = safe_image_show(context, image_id)
except exception.ImageNotFound:
instance.vm_state = vm_states.ERROR
instance.save()
reason = _('Unshelve attempted but the image %s '
'cannot be found.') % image_id
LOG.error(reason, instance=instance)
raise exception.UnshelveException(
instance_id=instance.uuid, reason=reason)
try:
with compute_utils.EventReporter(context, 'schedule_instances',
instance.uuid):
if not request_spec:
# NOTE(sbauza): We were unable to find an original
# RequestSpec object - probably because the instance is
# old. We need to mock that the old way
filter_properties = {}
request_spec = scheduler_utils.build_request_spec(
context, image, [instance])
else:
# NOTE(sbauza): Force_hosts/nodes needs to be reset
# if we want to make sure that the next destination
# is not forced to be the original host
request_spec.reset_forced_destinations()
# TODO(sbauza): Provide directly the RequestSpec object
# when _schedule_instances(),
# populate_filter_properties and populate_retry()
# accept it
filter_properties = request_spec.\
to_legacy_filter_properties_dict()
request_spec = request_spec.\
to_legacy_request_spec_dict()
scheduler_utils.populate_retry(filter_properties,
instance.uuid)
hosts = self._schedule_instances(
context, request_spec, filter_properties)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(
filter_properties, host_state)
(host, node) = (host_state['host'], host_state['nodename'])
self.compute_rpcapi.unshelve_instance(
context, instance, host, image=image,
filter_properties=filter_properties, node=node)
except (exception.NoValidHost,
exception.UnsupportedPolicyException):
instance.task_state = None
instance.save()
LOG.warning(_LW("No valid host found for unshelve instance"),
instance=instance)
return
except Exception:
with excutils.save_and_reraise_exception():
instance.task_state = None
instance.save()
LOG.error(_LE("Unshelve attempted but an error "
"has occurred"), instance=instance)
else:
LOG.error(_LE('Unshelve attempted but vm_state not SHELVED or '
'SHELVED_OFFLOADED'), instance=instance)
instance.vm_state = vm_states.ERROR
instance.save()
return
def build_instances(self, context, instances, image, filter_properties,
admin_password, injected_files, requested_networks,
security_groups, block_device_mapping=None, legacy_bdm=True):
# 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)
request_spec = {}
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.
request_spec = scheduler_utils.build_request_spec(
context, image, instances)
scheduler_utils.populate_retry(
filter_properties, instances[0].uuid)
hosts = self._schedule_instances(
context, request_spec, filter_properties)
except Exception as exc:
updates = {'vm_state': vm_states.ERROR, 'task_state': None}
for instance in instances:
self._set_vm_state_and_notify(
context, instance.uuid, 'build_instances', updates,
exc, request_spec)
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_allocated_networks(
context, instance, requested_networks)
return
for (instance, host) in six.moves.zip(instances, hosts):
try:
instance.refresh()
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)
# 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
self.compute_rpcapi.build_and_run_instance(context,
instance=instance, host=host['host'], image=image,
request_spec=request_spec,
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'])
def unshelve_instance(self, context, instance):
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)
snapshot_id = sys_meta.get('shelved_image_id')
if snapshot_id:
self._delete_image(context, snapshot_id)
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',
instance.uuid):
try:
image = safe_image_show(context, image_id)
except exception.ImageNotFound:
instance.vm_state = vm_states.ERROR
instance.save()
reason = _('Unshelve attempted but the image %s '
'cannot be found.') % image_id
LOG.error(reason, instance=instance)
raise exception.UnshelveException(
instance_id=instance.uuid, reason=reason)
try:
with compute_utils.EventReporter(context, 'schedule_instances',
instance.uuid):
filter_properties = {}
scheduler_utils.populate_retry(filter_properties,
instance.uuid)
hosts = self._schedule_instances(context, image,
filter_properties,
instance)
host_state = hosts[0]
scheduler_utils.populate_filter_properties(
filter_properties, host_state)
(host, node) = (host_state['host'], host_state['nodename'])
self.compute_rpcapi.unshelve_instance(
context,
instance,
host,
image=image,
filter_properties=filter_properties,
node=node)
except (exception.NoValidHost,
exception.UnsupportedPolicyException):
instance.task_state = None
instance.save()
LOG.warning(_LW("No valid host found for unshelve instance"),
instance=instance)
return
except Exception:
with excutils.save_and_reraise_exception():
instance.task_state = None
instance.save()
LOG.error(_LE("Unshelve attempted but an error "
"has occurred"),
instance=instance)
else:
LOG.error(_LE('Unshelve attempted but vm_state not SHELVED or '
'SHELVED_OFFLOADED'),
instance=instance)
instance.vm_state = vm_states.ERROR
instance.save()
return
for key in ['shelved_at', 'shelved_image_id', 'shelved_host']:
if key in sys_meta:
del (sys_meta[key])
instance.system_metadata = sys_meta
instance.save()
def build_instances(self,
context,
instances,
image,
filter_properties,
admin_password,
injected_files,
requested_networks,
security_groups,
block_device_mapping=None,
legacy_bdm=True):
# TODO(ndipanov): Remove block_device_mapping and legacy_bdm in version
# 2.0 of the RPC API.
request_spec = scheduler_utils.build_request_spec(
context, image, instances)
# 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(objects=[
objects.NetworkRequest.from_tuple(t)
for t in 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)
try:
scheduler_utils.setup_instance_group(context, request_spec,
filter_properties)
# check retry policy. Rather ugly use of instances[0]...
# but if we've exceeded max retries... then we really only
# have a single instance.
scheduler_utils.populate_retry(filter_properties,
instances[0].uuid)
hosts = self.scheduler_client.select_destinations(
context, request_spec, filter_properties)
except Exception as exc:
updates = {'vm_state': vm_states.ERROR, 'task_state': None}
for instance in instances:
self._set_vm_state_and_notify(context, instance.uuid,
'build_instances', updates, exc,
request_spec)
return
for (instance, host) in itertools.izip(instances, hosts):
try:
instance.refresh()
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)
# The block_device_mapping passed from the api doesn't contain
# instance specific information
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
self.compute_rpcapi.build_and_run_instance(
context,
instance=instance,
host=host['host'],
image=image,
request_spec=request_spec,
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'])
def _execute(self):
# TODO(sbauza): Remove once all the scheduler.utils methods accept a
# RequestSpec object in the signature.
legacy_props = self.request_spec.to_legacy_filter_properties_dict()
scheduler_utils.setup_instance_group(self.context, self.request_spec)
# If a target host is set in a requested destination,
# 'populate_retry' need not be executed.
if not ('requested_destination' in self.request_spec and
self.request_spec.requested_destination and
'host' in self.request_spec.requested_destination):
scheduler_utils.populate_retry(legacy_props,
self.instance.uuid)
# NOTE(sbauza): Force_hosts/nodes needs to be reset
# if we want to make sure that the next destination
# is not forced to be the original host
self.request_spec.reset_forced_destinations()
# TODO(gibi): We need to make sure that the requested_resources field
# is re calculated based on neutron ports.
self._restrict_request_spec_to_cell(legacy_props)
# Once _preallocate_migration() is done, the source node allocation is
# moved from the instance consumer to the migration record consumer,
# and the instance consumer doesn't have any allocations. If this is
# the first time through here (not a reschedule), select_destinations
# below will allocate resources on the selected destination node for
# the instance consumer. If we're rescheduling, host_list is not None
# and we'll call claim_resources for the instance and the selected
# alternate. If we exhaust our alternates and raise MaxRetriesExceeded,
# the rollback() method should revert the allocation swaparoo and move
# the source node allocation from the migration record back to the
# instance record.
migration = self._preallocate_migration()
self.request_spec.ensure_project_and_user_id(self.instance)
self.request_spec.ensure_network_metadata(self.instance)
compute_utils.heal_reqspec_is_bfv(
self.context, self.request_spec, self.instance)
# On an initial call to migrate, 'self.host_list' will be None, so we
# have to call the scheduler to get a list of acceptable hosts to
# migrate to. That list will consist of a selected host, along with
# zero or more alternates. On a reschedule, though, the alternates will
# be passed to this object and stored in 'self.host_list', so we can
# pop the first alternate from the list to use for the destination, and
# pass the remaining alternates to the compute.
if self.host_list is None:
selection_lists = self.query_client.select_destinations(
self.context, self.request_spec, [self.instance.uuid],
return_objects=True, return_alternates=True)
# Since there is only ever one instance to migrate per call, we
# just need the first returned element.
selection_list = selection_lists[0]
# The selected host is the first item in the list, with the
# alternates being the remainder of the list.
selection, self.host_list = selection_list[0], selection_list[1:]
else:
# This is a reschedule that will use the supplied alternate hosts
# in the host_list as destinations. Since the resources on these
# alternates may have been consumed and might not be able to
# support the migrated instance, we need to first claim the
# resources to verify the host still has sufficient availabile
# resources.
elevated = self.context.elevated()
host_available = False
while self.host_list and not host_available:
selection = self.host_list.pop(0)
if selection.allocation_request:
alloc_req = jsonutils.loads(selection.allocation_request)
else:
alloc_req = None
if alloc_req:
# If this call succeeds, the resources on the destination
# host will be claimed by the instance.
host_available = scheduler_utils.claim_resources(
elevated, self.reportclient, self.request_spec,
self.instance.uuid, alloc_req,
selection.allocation_request_version)
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
# There are no more available hosts. Raise a MaxRetriesExceeded
# exception in that case.
if not host_available:
reason = ("Exhausted all hosts available for retrying build "
"failures for instance %(instance_uuid)s." %
{"instance_uuid": self.instance.uuid})
raise exception.MaxRetriesExceeded(reason=reason)
scheduler_utils.populate_filter_properties(legacy_props, selection)
# context is not serializable
legacy_props.pop('context', None)
(host, node) = (selection.service_host, selection.nodename)
self.instance.availability_zone = (
availability_zones.get_host_availability_zone(
self.context, host))
LOG.debug("Calling prep_resize with selected host: %s; "
"Selected node: %s; Alternates: %s", host, node,
self.host_list, instance=self.instance)
# RPC cast to the destination host to start the migration process.
self.compute_rpcapi.prep_resize(
# NOTE(mriedem): Using request_spec.image here is potentially
# dangerous if it is not kept up to date (i.e. rebuild/unshelve);
# seems like the sane thing to do would be to pass the current
# instance.image_meta since that is what MoveClaim will use for
# any NUMA topology claims on the destination host...
self.context, self.instance, self.request_spec.image,
self.flavor, host, migration,
request_spec=self.request_spec, filter_properties=legacy_props,
node=node, clean_shutdown=self.clean_shutdown,
host_list=self.host_list)
def _cold_migrate(self, context, instance, flavor, filter_properties,
reservations):
image_ref = instance.image_ref
image = compute_utils.get_image_metadata(context, self.image_api,
image_ref, instance)
request_spec = scheduler_utils.build_request_spec(context,
image, [instance],
instance_type=flavor)
quotas = objects.Quotas.from_reservations(context,
reservations,
instance=instance)
try:
scheduler_utils.populate_retry(filter_properties, instance['uuid'])
hosts = self.scheduler_client.select_destinations(
context, request_spec, filter_properties)
host_state = hosts[0]
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, 'migrate_server', updates,
ex, request_spec)
quotas.rollback()
# 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)
try:
scheduler_utils.populate_filter_properties(filter_properties,
host_state)
# context is not serializable
filter_properties.pop('context', None)
# TODO(timello): originally, instance_type in request_spec
# on compute.api.resize does not have 'extra_specs', so we
# remove it for now to keep tests backward compatibility.
request_spec['instance_type'].pop('extra_specs')
(host, node) = (host_state['host'], host_state['nodename'])
self.compute_rpcapi.prep_resize(
context,
image,
instance,
flavor,
host,
reservations,
request_spec=request_spec,
filter_properties=filter_properties,
node=node)
except Exception as ex:
with excutils.save_and_reraise_exception():
updates = {
'vm_state': instance['vm_state'],
'task_state': None
}
self._set_vm_state_and_notify(context, 'migrate_server',
updates, ex, request_spec)
quotas.rollback()
def _execute(self):
# TODO(sbauza): Remove that once prep_resize() accepts a RequestSpec
# object in the signature and all the scheduler.utils methods too
legacy_spec = self.request_spec.to_legacy_request_spec_dict()
legacy_props = self.request_spec.to_legacy_filter_properties_dict()
scheduler_utils.setup_instance_group(self.context, self.request_spec)
# If a target host is set in a requested destination,
# 'populate_retry' need not be executed.
if not ('requested_destination' in self.request_spec and
self.request_spec.requested_destination and
'host' in self.request_spec.requested_destination):
scheduler_utils.populate_retry(legacy_props,
self.instance.uuid)
# NOTE(sbauza): Force_hosts/nodes needs to be reset
# if we want to make sure that the next destination
# is not forced to be the original host
self.request_spec.reset_forced_destinations()
# NOTE(danms): Right now we only support migrate to the same
# cell as the current instance, so request that the scheduler
# limit thusly.
instance_mapping = objects.InstanceMapping.get_by_instance_uuid(
self.context, self.instance.uuid)
LOG.debug('Requesting cell %(cell)s while migrating',
{'cell': instance_mapping.cell_mapping.identity},
instance=self.instance)
if ('requested_destination' in self.request_spec and
self.request_spec.requested_destination):
self.request_spec.requested_destination.cell = (
instance_mapping.cell_mapping)
# NOTE(takashin): In the case that the target host is specified,
# if the migration is failed, it is not necessary to retry
# the cold migration to the same host. So make sure that
# reschedule will not occur.
if 'host' in self.request_spec.requested_destination:
legacy_props.pop('retry', None)
self.request_spec.retry = None
else:
self.request_spec.requested_destination = objects.Destination(
cell=instance_mapping.cell_mapping)
# Once _preallocate_migration() is done, the source node allocation is
# moved from the instance consumer to the migration record consumer,
# and the instance consumer doesn't have any allocations. If this is
# the first time through here (not a reschedule), select_destinations
# below will allocate resources on the selected destination node for
# the instance consumer. If we're rescheduling, host_list is not None
# and we'll call claim_resources for the instance and the selected
# alternate. If we exhaust our alternates and raise MaxRetriesExceeded,
# the rollback() method should revert the allocation swaparoo and move
# the source node allocation from the migration record back to the
# instance record.
migration = self._preallocate_migration()
self.request_spec.ensure_project_and_user_id(self.instance)
# On an initial call to migrate, 'self.host_list' will be None, so we
# have to call the scheduler to get a list of acceptable hosts to
# migrate to. That list will consist of a selected host, along with
# zero or more alternates. On a reschedule, though, the alternates will
# be passed to this object and stored in 'self.host_list', so we can
# pop the first alternate from the list to use for the destination, and
# pass the remaining alternates to the compute.
if self.host_list is None:
selection_lists = self.scheduler_client.select_destinations(
self.context, self.request_spec, [self.instance.uuid],
return_objects=True, return_alternates=True)
# Since there is only ever one instance to migrate per call, we
# just need the first returned element.
selection_list = selection_lists[0]
# The selected host is the first item in the list, with the
# alternates being the remainder of the list.
selection, self.host_list = selection_list[0], selection_list[1:]
else:
# This is a reschedule that will use the supplied alternate hosts
# in the host_list as destinations. Since the resources on these
# alternates may have been consumed and might not be able to
# support the migrated instance, we need to first claim the
# resources to verify the host still has sufficient availabile
# resources.
elevated = self.context.elevated()
host_available = False
while self.host_list and not host_available:
selection = self.host_list.pop(0)
if selection.allocation_request:
alloc_req = jsonutils.loads(selection.allocation_request)
else:
alloc_req = None
if alloc_req:
# If this call succeeds, the resources on the destination
# host will be claimed by the instance.
host_available = scheduler_utils.claim_resources(
elevated, self.reportclient, self.request_spec,
self.instance.uuid, alloc_req,
selection.allocation_request_version)
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
# There are no more available hosts. Raise a MaxRetriesExceeded
# exception in that case.
if not host_available:
reason = ("Exhausted all hosts available for retrying build "
"failures for instance %(instance_uuid)s." %
{"instance_uuid": self.instance.uuid})
raise exception.MaxRetriesExceeded(reason=reason)
scheduler_utils.populate_filter_properties(legacy_props, selection)
# context is not serializable
legacy_props.pop('context', None)
(host, node) = (selection.service_host, selection.nodename)
self.instance.availability_zone = (
availability_zones.get_host_availability_zone(
self.context, host))
# FIXME(sbauza): Serialize/Unserialize the legacy dict because of
# oslo.messaging #1529084 to transform datetime values into strings.
# tl;dr: datetimes in dicts are not accepted as correct values by the
# rpc fake driver.
legacy_spec = jsonutils.loads(jsonutils.dumps(legacy_spec))
LOG.debug("Calling prep_resize with selected host: %s; "
"Selected node: %s; Alternates: %s", host, node,
self.host_list, instance=self.instance)
# RPC cast to the destination host to start the migration process.
self.compute_rpcapi.prep_resize(
self.context, self.instance, legacy_spec['image'],
self.flavor, host, migration,
request_spec=legacy_spec, filter_properties=legacy_props,
node=node, clean_shutdown=self.clean_shutdown,
host_list=self.host_list)
def build_instances(self,
context,
instances,
image,
filter_properties,
admin_password,
injected_files,
requested_networks,
security_groups,
block_device_mapping=None,
legacy_bdm=True):
# TODO(ndipanov): Remove block_device_mapping and legacy_bdm in version
# 2.0 of the RPC API.
request_spec = scheduler_utils.build_request_spec(
context, image, instances)
# 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(objects=[
objects.NetworkRequest.from_tuple(t)
for t in requested_networks
])
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.
scheduler_utils.populate_retry(filter_properties,
instances[0].uuid)
hosts = self.scheduler_client.select_destinations(
context, request_spec, filter_properties)
except Exception as exc:
for instance in instances:
scheduler_driver.handle_schedule_error(context, exc,
instance.uuid,
request_spec)
return
for (instance, host) in itertools.izip(instances, hosts):
try:
instance.refresh()
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)
# The block_device_mapping passed from the api doesn't contain
# instance specific information
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
self.compute_rpcapi.build_and_run_instance(
context,
instance=instance,
host=host['host'],
image=image,
request_spec=request_spec,
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'])
def schedule_run_instance(self, context, request_spec,
admin_password, injected_files,
requested_networks, is_first_time,
filter_properties, legacy_bdm_in_spec):
"""Provisions instances that needs to be scheduled
Applies filters and weighters on request properties to get a list of
compute hosts and calls them to spawn instance(s).
"""
payload = dict(request_spec=request_spec)
self.notifier.info(context, 'scheduler.run_instance.start', payload)
instance_uuids = request_spec.get('instance_uuids')
LOG.info(_("Attempting to build %(num_instances)d instance(s) "
"uuids: %(instance_uuids)s"),
{'num_instances': len(instance_uuids),
'instance_uuids': instance_uuids})
LOG.debug("Request Spec: %s" % request_spec)
# check retry policy. Rather ugly use of instance_uuids[0]...
# but if we've exceeded max retries... then we really only
# have a single instance.
scheduler_utils.populate_retry(filter_properties,
instance_uuids[0])
weighed_hosts = self._schedule(context, request_spec,
filter_properties)
# NOTE: Pop instance_uuids as individual creates do not need the
# set of uuids. Do not pop before here as the upper exception
# handler fo NoValidHost needs the uuid to set error state
instance_uuids = request_spec.pop('instance_uuids')
# NOTE(comstud): Make sure we do not pass this through. It
# contains an instance of RpcContext that cannot be serialized.
filter_properties.pop('context', None)
for num, instance_uuid in enumerate(instance_uuids):
request_spec['instance_properties']['launch_index'] = num
try:
try:
weighed_host = weighed_hosts.pop(0)
LOG.info(_("Choosing host %(weighed_host)s "
"for instance %(instance_uuid)s"),
{'weighed_host': weighed_host,
'instance_uuid': instance_uuid})
except IndexError:
raise exception.NoValidHost(reason="")
self._provision_resource(context, weighed_host,
request_spec,
filter_properties,
requested_networks,
injected_files, admin_password,
is_first_time,
instance_uuid=instance_uuid,
legacy_bdm_in_spec=legacy_bdm_in_spec)
except Exception as ex:
# NOTE(vish): we don't reraise the exception here to make sure
# that all instances in the request get set to
# error properly
driver.handle_schedule_error(context, ex, instance_uuid,
request_spec)
# scrub retry host list in case we're scheduling multiple
# instances:
retry = filter_properties.get('retry', {})
retry['hosts'] = []
self.notifier.info(context, 'scheduler.run_instance.end', payload)
def build_instances(self,
context,
instances,
image,
filter_properties,
admin_password,
injected_files,
requested_networks,
security_groups,
block_device_mapping=None,
legacy_bdm=True):
# TODO(ndipanov): Remove block_device_mapping and legacy_bdm in version
# 2.0 of the RPC API.
request_spec = scheduler_utils.build_request_spec(
context, image, instances)
# NOTE(sbauza): filter_properties['hints'] can be None
hints = filter_properties.get('scheduler_hints', {}) or {}
group_hint = hints.get('group')
group_hosts = filter_properties.get('group_hosts')
group_info = scheduler_utils.setup_instance_group(
context, group_hint, group_hosts)
if isinstance(group_info, tuple):
filter_properties['group_updated'] = True
(filter_properties['group_hosts'],
filter_properties['group_policies']) = group_info
# 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(objects=[
objects.NetworkRequest.from_tuple(t)
for t in requested_networks
])
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.
# (luzhq) 验证重试策略
# 更新filter_properties中的重试属性，若当前为重试部署则同时检测当前
# 的重试次数是否超过最大重试次数，需要注意的是：这里使用instances[0]
# 表示如重试的话只会有一个instance重试
scheduler_utils.populate_retry(filter_properties,
instances[0].uuid)
hosts = self.scheduler_client.select_destinations(
context, request_spec, filter_properties)
except Exception as exc:
for instance in instances:
scheduler_driver.handle_schedule_error(context, exc,
instance.uuid,
request_spec)
return
for (instance, host) in itertools.izip(instances, hosts):
try:
instance.refresh()
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)
# The block_device_mapping passed from the api doesn't contain
# instance specific information
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
# (luzhq) self.compute_rpcapi = compute_rpcapi.ComputeAPI()
self.compute_rpcapi.build_and_run_instance(
context,
instance=instance,
host=host['host'],
image=image,
request_spec=request_spec,
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'])