OpenStack Compute (Nova)
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 

6540 lines
295 KiB

# Copyright 2010 United States Government as represented by the
# Administrator of the National Aeronautics and Space Administration.
# Copyright 2011 Justin Santa Barbara
# All Rights Reserved.
#
# 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 all processes relating to instances (guest vms).
The :py:class:`ComputeManager` class is a :py:class:`nova.manager.Manager` that
handles RPC calls relating to creating instances. It is responsible for
building a disk image, launching it via the underlying virtualization driver,
responding to calls to check its state, attaching persistent storage, and
terminating it.
"""
import base64
import contextlib
import functools
import socket
import sys
import time
import traceback
import uuid
from cinderclient import exceptions as cinder_exception
import eventlet.event
from eventlet import greenthread
import eventlet.semaphore
import eventlet.timeout
from keystoneclient import exceptions as keystone_exception
from oslo_config import cfg
from oslo_log import log as logging
import oslo_messaging as messaging
from oslo_serialization import jsonutils
from oslo_service import loopingcall
from oslo_service import periodic_task
from oslo_utils import excutils
from oslo_utils import strutils
from oslo_utils import timeutils
import six
from six.moves import range
from nova import block_device
from nova.cells import rpcapi as cells_rpcapi
from nova.cloudpipe import pipelib
from nova import compute
from nova.compute import build_results
from nova.compute import power_state
from nova.compute import resource_tracker
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 import vm_states
from nova import conductor
from nova import consoleauth
import nova.context
from nova import exception
from nova import hooks
from nova.i18n import _
from nova.i18n import _LE
from nova.i18n import _LI
from nova.i18n import _LW
from nova import image
from nova.image import glance
from nova import manager
from nova import network
from nova.network import base_api as base_net_api
from nova.network import model as network_model
from nova.network.security_group import openstack_driver
from nova import objects
from nova.objects import base as obj_base
from nova import paths
from nova import rpc
from nova import safe_utils
from nova.scheduler import client as scheduler_client
from nova import utils
from nova.virt import block_device as driver_block_device
from nova.virt import configdrive
from nova.virt import driver
from nova.virt import event as virtevent
from nova.virt import storage_users
from nova.virt import virtapi
from nova import volume
from nova.volume import encryptors
compute_opts = [
cfg.StrOpt('console_host',
default=socket.gethostname(),
help='Console proxy host to use to connect '
'to instances on this host.'),
cfg.StrOpt('default_access_ip_network_name',
help='Name of network to use to set access IPs for instances'),
cfg.BoolOpt('defer_iptables_apply',
default=False,
help='Whether to batch up the application of IPTables rules'
' during a host restart and apply all at the end of the'
' init phase'),
cfg.StrOpt('instances_path',
default=paths.state_path_def('instances'),
help='Where instances are stored on disk'),
cfg.BoolOpt('instance_usage_audit',
default=False,
help="Generate periodic compute.instance.exists"
" notifications"),
cfg.IntOpt('live_migration_retry_count',
default=30,
help="Number of 1 second retries needed in live_migration"),
cfg.BoolOpt('resume_guests_state_on_host_boot',
default=False,
help='Whether to start guests that were running before the '
'host rebooted'),
cfg.IntOpt('network_allocate_retries',
default=0,
help="Number of times to retry network allocation on failures"),
cfg.IntOpt('max_concurrent_builds',
default=10,
help='Maximum number of instance builds to run concurrently'),
cfg.IntOpt('max_concurrent_live_migrations',
default=1,
help='Maximum number of live migrations to run concurrently. '
'This limit is enforced to avoid outbound live migrations '
'overwhelming the host/network and causing failures. It '
'is not recommended that you change this unless you are '
'very sure that doing so is safe and stable in your '
'environment.'),
cfg.IntOpt('block_device_allocate_retries',
default=60,
help='Number of times to retry block device'
' allocation on failures')
]
interval_opts = [
cfg.IntOpt('bandwidth_poll_interval',
default=600,
help='Interval to pull network bandwidth usage info. Not '
'supported on all hypervisors. Set to -1 to disable. '
'Setting this to 0 will run at the default rate.'),
cfg.IntOpt('sync_power_state_interval',
default=600,
help='Interval to sync power states between the database and '
'the hypervisor. Set to -1 to disable. '
'Setting this to 0 will run at the default rate.'),
cfg.IntOpt("heal_instance_info_cache_interval",
default=60,
help="Number of seconds between instance network information "
"cache updates"),
cfg.IntOpt('reclaim_instance_interval',
default=0,
help='Interval in seconds for reclaiming deleted instances'),
cfg.IntOpt('volume_usage_poll_interval',
default=0,
help='Interval in seconds for gathering volume usages'),
cfg.IntOpt('shelved_poll_interval',
default=3600,
help='Interval in seconds for polling shelved instances to '
'offload. Set to -1 to disable.'
'Setting this to 0 will run at the default rate.'),
cfg.IntOpt('shelved_offload_time',
default=0,
help='Time in seconds before a shelved instance is eligible '
'for removing from a host. -1 never offload, 0 offload '
'immediately when shelved'),
cfg.IntOpt('instance_delete_interval',
default=300,
help='Interval in seconds for retrying failed instance file '
'deletes. Set to -1 to disable. '
'Setting this to 0 will run at the default rate.'),
cfg.IntOpt('block_device_allocate_retries_interval',
default=3,
help='Waiting time interval (seconds) between block'
' device allocation retries on failures'),
cfg.IntOpt('scheduler_instance_sync_interval',
default=120,
help='Waiting time interval (seconds) between sending the '
'scheduler a list of current instance UUIDs to verify '
'that its view of instances is in sync with nova. If the '
'CONF option `scheduler_tracks_instance_changes` is '
'False, changing this option will have no effect.'),
cfg.IntOpt('update_resources_interval',
default=0,
help='Interval in seconds for updating compute resources. A '
'number less than 0 means to disable the task completely. '
'Leaving this at the default of 0 will cause this to run '
'at the default periodic interval. Setting it to any '
'positive value will cause it to run at approximately '
'that number of seconds.'),
]
timeout_opts = [
cfg.IntOpt("reboot_timeout",
default=0,
help="Automatically hard reboot an instance if it has been "
"stuck in a rebooting state longer than N seconds. "
"Set to 0 to disable."),
cfg.IntOpt("instance_build_timeout",
default=0,
help="Amount of time in seconds an instance can be in BUILD "
"before going into ERROR status. "
"Set to 0 to disable."),
cfg.IntOpt("rescue_timeout",
default=0,
help="Automatically unrescue an instance after N seconds. "
"Set to 0 to disable."),
cfg.IntOpt("resize_confirm_window",
default=0,
help="Automatically confirm resizes after N seconds. "
"Set to 0 to disable."),
cfg.IntOpt("shutdown_timeout",
default=60,
help="Total amount of time to wait in seconds for an instance "
"to perform a clean shutdown."),
]
running_deleted_opts = [
cfg.StrOpt("running_deleted_instance_action",
default="reap",
choices=('noop', 'log', 'shutdown', 'reap'),
help="Action to take if a running deleted instance is detected."
"Set to 'noop' to take no action."),
cfg.IntOpt("running_deleted_instance_poll_interval",
default=1800,
help="Number of seconds to wait between runs of the cleanup "
"task."),
cfg.IntOpt("running_deleted_instance_timeout",
default=0,
help="Number of seconds after being deleted when a running "
"instance should be considered eligible for cleanup."),
]
instance_cleaning_opts = [
cfg.IntOpt('maximum_instance_delete_attempts',
default=5,
help='The number of times to attempt to reap an instance\'s '
'files.'),
]
CONF = cfg.CONF
CONF.register_opts(compute_opts)
CONF.register_opts(interval_opts)
CONF.register_opts(timeout_opts)
CONF.register_opts(running_deleted_opts)
CONF.register_opts(instance_cleaning_opts)
CONF.import_opt('console_topic', 'nova.console.rpcapi')
CONF.import_opt('host', 'nova.netconf')
CONF.import_opt('enabled', 'nova.vnc', group='vnc')
CONF.import_opt('enabled', 'nova.spice', group='spice')
CONF.import_opt('enable', 'nova.cells.opts', group='cells')
CONF.import_opt('image_cache_manager_interval', 'nova.virt.imagecache')
CONF.import_opt('enabled', 'nova.rdp', group='rdp')
CONF.import_opt('html5_proxy_base_url', 'nova.rdp', group='rdp')
CONF.import_opt('enabled', 'nova.mks', group='mks')
CONF.import_opt('mksproxy_base_url', 'nova.mks', group='mks')
CONF.import_opt('enabled', 'nova.console.serial', group='serial_console')
CONF.import_opt('base_url', 'nova.console.serial', group='serial_console')
CONF.import_opt('destroy_after_evacuate', 'nova.utils', group='workarounds')
CONF.import_opt('scheduler_tracks_instance_changes',
'nova.scheduler.host_manager')
LOG = logging.getLogger(__name__)
get_notifier = functools.partial(rpc.get_notifier, service='compute')
wrap_exception = functools.partial(exception.wrap_exception,
get_notifier=get_notifier)
@utils.expects_func_args('migration')
def errors_out_migration(function):
"""Decorator to error out migration on failure."""
@functools.wraps(function)
def decorated_function(self, context, *args, **kwargs):
try:
return function(self, context, *args, **kwargs)
except Exception as ex:
with excutils.save_and_reraise_exception():
wrapped_func = utils.get_wrapped_function(function)
keyed_args = safe_utils.getcallargs(wrapped_func, context,
*args, **kwargs)
migration = keyed_args['migration']
# NOTE(rajesht): If InstanceNotFound error is thrown from
# decorated function, migration status should be set to
# 'error', without checking current migration status.
if not isinstance(ex, exception.InstanceNotFound):
status = migration.status
if status not in ['migrating', 'post-migrating']:
return
migration.status = 'error'
try:
with migration.obj_as_admin():
migration.save()
except Exception:
LOG.debug('Error setting migration status '
'for instance %s.',
migration.instance_uuid, exc_info=True)
return decorated_function
@utils.expects_func_args('instance')
def reverts_task_state(function):
"""Decorator to revert task_state on failure."""
@functools.wraps(function)
def decorated_function(self, context, *args, **kwargs):
try:
return function(self, context, *args, **kwargs)
except exception.UnexpectedTaskStateError as e:
# Note(maoy): unexpected task state means the current
# task is preempted. Do not clear task state in this
# case.
with excutils.save_and_reraise_exception():
LOG.info(_LI("Task possibly preempted: %s"),
e.format_message())
except Exception:
with excutils.save_and_reraise_exception():
wrapped_func = utils.get_wrapped_function(function)
keyed_args = safe_utils.getcallargs(wrapped_func, context,
*args, **kwargs)
# NOTE(mriedem): 'instance' must be in keyed_args because we
# have utils.expects_func_args('instance') decorating this
# method.
instance = keyed_args['instance']
try:
self._instance_update(context, instance, task_state=None)
except exception.InstanceNotFound:
# We might delete an instance that failed to build shortly
# after it errored out this is an expected case and we
# should not trace on it.
pass
except Exception as e:
msg = _LW("Failed to revert task state for instance. "
"Error: %s")
LOG.warning(msg, e, instance=instance)
return decorated_function
@utils.expects_func_args('instance')
def wrap_instance_fault(function):
"""Wraps a method to catch exceptions related to instances.
This decorator wraps a method to catch any exceptions having to do with
an instance that may get thrown. It then logs an instance fault in the db.
"""
@functools.wraps(function)
def decorated_function(self, context, *args, **kwargs):
try:
return function(self, context, *args, **kwargs)
except exception.InstanceNotFound:
raise
except Exception as e:
# NOTE(gtt): If argument 'instance' is in args rather than kwargs,
# we will get a KeyError exception which will cover up the real
# exception. So, we update kwargs with the values from args first.
# then, we can get 'instance' from kwargs easily.
kwargs.update(dict(zip(function.func_code.co_varnames[2:], args)))
with excutils.save_and_reraise_exception():
compute_utils.add_instance_fault_from_exc(context,
kwargs['instance'], e, sys.exc_info())
return decorated_function
@utils.expects_func_args('instance')
def wrap_instance_event(function):
"""Wraps a method to log the event taken on the instance, and result.
This decorator wraps a method to log the start and result of an event, as
part of an action taken on an instance.
"""
@functools.wraps(function)
def decorated_function(self, context, *args, **kwargs):
wrapped_func = utils.get_wrapped_function(function)
keyed_args = safe_utils.getcallargs(wrapped_func, context, *args,
**kwargs)
instance_uuid = keyed_args['instance']['uuid']
event_name = 'compute_{0}'.format(function.func_name)
with compute_utils.EventReporter(context, event_name, instance_uuid):
return function(self, context, *args, **kwargs)
return decorated_function
@utils.expects_func_args('image_id', 'instance')
def delete_image_on_error(function):
"""Used for snapshot related method to ensure the image created in
compute.api is deleted when an error occurs.
"""
@functools.wraps(function)
def decorated_function(self, context, image_id, instance,
*args, **kwargs):
try:
return function(self, context, image_id, instance,
*args, **kwargs)
except Exception:
with excutils.save_and_reraise_exception():
LOG.debug("Cleaning up image %s", image_id,
exc_info=True, instance=instance)
try:
self.image_api.delete(context, image_id)
except Exception:
LOG.exception(_LE("Error while trying to clean up "
"image %s"), image_id,
instance=instance)
return decorated_function
# TODO(danms): Remove me after Icehouse
# NOTE(mikal): if the method being decorated has more than one decorator, then
# put this one first. Otherwise the various exception handling decorators do
# not function correctly.
def object_compat(function):
"""Wraps a method that expects a new-world instance
This provides compatibility for callers passing old-style dict
instances.
"""
@functools.wraps(function)
def decorated_function(self, context, *args, **kwargs):
def _load_instance(instance_or_dict):
if isinstance(instance_or_dict, dict):
# try to get metadata and system_metadata for most cases but
# only attempt to load those if the db instance already has
# those fields joined
metas = [meta for meta in ('metadata', 'system_metadata')
if meta in instance_or_dict]
instance = objects.Instance._from_db_object(
context, objects.Instance(), instance_or_dict,
expected_attrs=metas)
instance._context = context
return instance
return instance_or_dict
try:
kwargs['instance'] = _load_instance(kwargs['instance'])
except KeyError:
args = (_load_instance(args[0]),) + args[1:]
migration = kwargs.get('migration')
if isinstance(migration, dict):
migration = objects.Migration._from_db_object(
context.elevated(), objects.Migration(),
migration)
kwargs['migration'] = migration
return function(self, context, *args, **kwargs)
return decorated_function
class InstanceEvents(object):
def __init__(self):
self._events = {}
@staticmethod
def _lock_name(instance):
return '%s-%s' % (instance.uuid, 'events')
def prepare_for_instance_event(self, instance, event_name):
"""Prepare to receive an event for an instance.
This will register an event for the given instance that we will
wait on later. This should be called before initiating whatever
action will trigger the event. The resulting eventlet.event.Event
object should be wait()'d on to ensure completion.
:param instance: the instance for which the event will be generated
:param event_name: the name of the event we're expecting
:returns: an event object that should be wait()'d on
"""
if self._events is None:
# NOTE(danms): We really should have a more specific error
# here, but this is what we use for our default error case
raise exception.NovaException('In shutdown, no new events '
'can be scheduled')
@utils.synchronized(self._lock_name(instance))
def _create_or_get_event():
instance_events = self._events.setdefault(instance.uuid, {})
return instance_events.setdefault(event_name,
eventlet.event.Event())
LOG.debug('Preparing to wait for external event %(event)s',
{'event': event_name}, instance=instance)
return _create_or_get_event()
def pop_instance_event(self, instance, event):
"""Remove a pending event from the wait list.
This will remove a pending event from the wait list so that it
can be used to signal the waiters to wake up.
:param instance: the instance for which the event was generated
:param event: the nova.objects.external_event.InstanceExternalEvent
that describes the event
:returns: the eventlet.event.Event object on which the waiters
are blocked
"""
no_events_sentinel = object()
no_matching_event_sentinel = object()
@utils.synchronized(self._lock_name(instance))
def _pop_event():
if not self._events:
LOG.debug('Unexpected attempt to pop events during shutdown',
instance=instance)
return no_events_sentinel
events = self._events.get(instance.uuid)
if not events:
return no_events_sentinel
_event = events.pop(event.key, None)
if not events:
del self._events[instance.uuid]
if _event is None:
return no_matching_event_sentinel
return _event
result = _pop_event()
if result is no_events_sentinel:
LOG.debug('No waiting events found dispatching %(event)s',
{'event': event.key},
instance=instance)
return None
elif result is no_matching_event_sentinel:
LOG.debug('No event matching %(event)s in %(events)s',
{'event': event.key,
'events': self._events.get(instance.uuid, {}).keys()},
instance=instance)
return None
else:
return result
def clear_events_for_instance(self, instance):
"""Remove all pending events for an instance.
This will remove all events currently pending for an instance
and return them (indexed by event name).
:param instance: the instance for which events should be purged
:returns: a dictionary of {event_name: eventlet.event.Event}
"""
@utils.synchronized(self._lock_name(instance))
def _clear_events():
if self._events is None:
LOG.debug('Unexpected attempt to clear events during shutdown',
instance=instance)
return dict()
return self._events.pop(instance.uuid, {})
return _clear_events()
def cancel_all_events(self):
our_events = self._events
# NOTE(danms): Block new events
self._events = None
for instance_uuid, events in our_events.items():
for event_name, eventlet_event in events.items():
LOG.debug('Canceling in-flight event %(event)s for '
'instance %(instance_uuid)s',
{'event': event_name,
'instance_uuid': instance_uuid})
name, tag = event_name.rsplit('-', 1)
event = objects.InstanceExternalEvent(
instance_uuid=instance_uuid,
name=name, status='failed',
tag=tag, data={})
eventlet_event.send(event)
class ComputeVirtAPI(virtapi.VirtAPI):
def __init__(self, compute):
super(ComputeVirtAPI, self).__init__()
self._compute = compute
def provider_fw_rule_get_all(self, context):
return self._compute.conductor_api.provider_fw_rule_get_all(context)
def _default_error_callback(self, event_name, instance):
raise exception.NovaException(_('Instance event failed'))
@contextlib.contextmanager
def wait_for_instance_event(self, instance, event_names, deadline=300,
error_callback=None):
"""Plan to wait for some events, run some code, then wait.
This context manager will first create plans to wait for the
provided event_names, yield, and then wait for all the scheduled
events to complete.
Note that this uses an eventlet.timeout.Timeout to bound the
operation, so callers should be prepared to catch that
failure and handle that situation appropriately.
If the event is not received by the specified timeout deadline,
eventlet.timeout.Timeout is raised.
If the event is received but did not have a 'completed'
status, a NovaException is raised. If an error_callback is
provided, instead of raising an exception as detailed above
for the failure case, the callback will be called with the
event_name and instance, and can return True to continue
waiting for the rest of the events, False to stop processing,
or raise an exception which will bubble up to the waiter.
:param instance: The instance for which an event is expected
:param event_names: A list of event names. Each element can be a
string event name or tuple of strings to
indicate (name, tag).
:param deadline: Maximum number of seconds we should wait for all
of the specified events to arrive.
:param error_callback: A function to be called if an event arrives
"""
if error_callback is None:
error_callback = self._default_error_callback
events = {}
for event_name in event_names:
if isinstance(event_name, tuple):
name, tag = event_name
event_name = objects.InstanceExternalEvent.make_key(
name, tag)
try:
events[event_name] = (
self._compute.instance_events.prepare_for_instance_event(
instance, event_name))
except exception.NovaException:
error_callback(event_name, instance)
# NOTE(danms): Don't wait for any of the events. They
# should all be canceled and fired immediately below,
# but don't stick around if not.
deadline = 0
yield
with eventlet.timeout.Timeout(deadline):
for event_name, event in events.items():
actual_event = event.wait()
if actual_event.status == 'completed':
continue
decision = error_callback(event_name, instance)
if decision is False:
break
class ComputeManager(manager.Manager):
"""Manages the running instances from creation to destruction."""
target = messaging.Target(version='4.4')
# How long to wait in seconds before re-issuing a shutdown
# signal to an instance during power off. The overall
# time to wait is set by CONF.shutdown_timeout.
SHUTDOWN_RETRY_INTERVAL = 10
def __init__(self, compute_driver=None, *args, **kwargs):
"""Load configuration options and connect to the hypervisor."""
self.virtapi = ComputeVirtAPI(self)
self.network_api = network.API()
self.volume_api = volume.API()
self.image_api = image.API()
self._last_host_check = 0
self._last_bw_usage_poll = 0
self._bw_usage_supported = True
self._last_bw_usage_cell_update = 0
self.compute_api = compute.API()
self.compute_rpcapi = compute_rpcapi.ComputeAPI()
self.conductor_api = conductor.API()
self.compute_task_api = conductor.ComputeTaskAPI()
self.is_neutron_security_groups = (
openstack_driver.is_neutron_security_groups())
self.consoleauth_rpcapi = consoleauth.rpcapi.ConsoleAuthAPI()
self.cells_rpcapi = cells_rpcapi.CellsAPI()
self.scheduler_client = scheduler_client.SchedulerClient()
self._resource_tracker_dict = {}
self.instance_events = InstanceEvents()
self._sync_power_pool = eventlet.GreenPool()
self._syncs_in_progress = {}
self.send_instance_updates = CONF.scheduler_tracks_instance_changes
if CONF.max_concurrent_builds != 0:
self._build_semaphore = eventlet.semaphore.Semaphore(
CONF.max_concurrent_builds)
else:
self._build_semaphore = compute_utils.UnlimitedSemaphore()
if max(CONF.max_concurrent_live_migrations, 0) != 0:
self._live_migration_semaphore = eventlet.semaphore.Semaphore(
CONF.max_concurrent_live_migrations)
else:
self._live_migration_semaphore = compute_utils.UnlimitedSemaphore()
super(ComputeManager, self).__init__(service_name="compute",
*args, **kwargs)
# NOTE(russellb) Load the driver last. It may call back into the
# compute manager via the virtapi, so we want it to be fully
# initialized before that happens.
self.driver = driver.load_compute_driver(self.virtapi, compute_driver)
self.use_legacy_block_device_info = \
self.driver.need_legacy_block_device_info
def _get_resource_tracker(self, nodename):
rt = self._resource_tracker_dict.get(nodename)
if not rt:
if not self.driver.node_is_available(nodename):
raise exception.NovaException(
_("%s is not a valid node managed by this "
"compute host.") % nodename)
rt = resource_tracker.ResourceTracker(self.host,
self.driver,
nodename)
self._resource_tracker_dict[nodename] = rt
return rt
def _update_resource_tracker(self, context, instance):
"""Let the resource tracker know that an instance has changed state."""
if (instance.host == self.host and
self.driver.node_is_available(instance.node)):
rt = self._get_resource_tracker(instance.node)
rt.update_usage(context, instance)
def _instance_update(self, context, instance, **kwargs):
"""Update an instance in the database using kwargs as value."""
for k, v in kwargs.items():
setattr(instance, k, v)
instance.save()
self._update_resource_tracker(context, instance)
def _nil_out_instance_obj_host_and_node(self, instance):
# NOTE(jwcroppe): We don't do instance.save() here for performance
# reasons; a call to this is expected to be immediately followed by
# another call that does instance.save(), thus avoiding two writes
# to the database layer.
instance.host = None
instance.node = None
def _set_instance_obj_error_state(self, context, instance,
clean_task_state=False):
try:
instance.vm_state = vm_states.ERROR
if clean_task_state:
instance.task_state = None
instance.save()
except exception.InstanceNotFound:
LOG.debug('Instance has been destroyed from under us while '
'trying to set it to ERROR', instance=instance)
def _get_instances_on_driver(self, context, filters=None):
"""Return a list of instance records for the instances found
on the hypervisor which satisfy the specified filters. If filters=None
return a list of instance records for all the instances found on the
hypervisor.
"""
if not filters:
filters = {}
try:
driver_uuids = self.driver.list_instance_uuids()
if len(driver_uuids) == 0:
# Short circuit, don't waste a DB call
return objects.InstanceList()
filters['uuid'] = driver_uuids
local_instances = objects.InstanceList.get_by_filters(
context, filters, use_slave=True)
return local_instances
except NotImplementedError:
pass
# The driver doesn't support uuids listing, so we'll have
# to brute force.
driver_instances = self.driver.list_instances()
instances = objects.InstanceList.get_by_filters(context, filters,
use_slave=True)
name_map = {instance.name: instance for instance in instances}
local_instances = []
for driver_instance in driver_instances:
instance = name_map.get(driver_instance)
if not instance:
continue
local_instances.append(instance)
return local_instances
def _destroy_evacuated_instances(self, context):
"""Destroys evacuated instances.
While nova-compute was down, the instances running on it could be
evacuated to another host. Check that the instances reported
by the driver are still associated with this host. If they are
not, destroy them, with the exception of instances which are in
the MIGRATING, RESIZE_MIGRATING, RESIZE_MIGRATED, RESIZE_FINISH
task state or RESIZED vm state.
"""
filters = {
'source_compute': self.host,
'status': 'accepted',
'migration_type': 'evacuation',
}
evacuations = objects.MigrationList.get_by_filters(context, filters)
if not evacuations:
return
evacuations = {mig.instance_uuid: mig for mig in evacuations}
filters = {'deleted': False}
local_instances = self._get_instances_on_driver(context, filters)
evacuated = [inst for inst in local_instances
if inst.uuid in evacuations]
for instance in evacuated:
migration = evacuations[instance.uuid]
LOG.info(_LI('Deleting instance as it has been evacuated from '
'this host'), instance=instance)
try:
network_info = self.network_api.get_instance_nw_info(
context, instance)
bdi = self._get_instance_block_device_info(context,
instance)
destroy_disks = not (self._is_instance_storage_shared(
context, instance))
except exception.InstanceNotFound:
network_info = network_model.NetworkInfo()
bdi = {}
LOG.info(_LI('Instance has been marked deleted already, '
'removing it from the hypervisor.'),
instance=instance)
# always destroy disks if the instance was deleted
destroy_disks = True
self.driver.destroy(context, instance,
network_info,
bdi, destroy_disks)
migration.status = 'completed'
migration.save()
def _is_instance_storage_shared(self, context, instance, host=None):
shared_storage = True
data = None
try:
data = self.driver.check_instance_shared_storage_local(context,
instance)
if data:
shared_storage = (self.compute_rpcapi.
check_instance_shared_storage(context,
instance, data, host=host))
except NotImplementedError:
LOG.debug('Hypervisor driver does not support '
'instance shared storage check, '
'assuming it\'s not on shared storage',
instance=instance)
shared_storage = False
except Exception:
LOG.exception(_LE('Failed to check if instance shared'),
instance=instance)
finally:
if data:
self.driver.check_instance_shared_storage_cleanup(context,
data)
return shared_storage
def _complete_partial_deletion(self, context, instance):
"""Complete deletion for instances in DELETED status but not marked as
deleted in the DB
"""
instance.destroy()
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
quotas = objects.Quotas(context=context)
project_id, user_id = objects.quotas.ids_from_instance(context,
instance)
quotas.reserve(project_id=project_id, user_id=user_id, instances=-1,
cores=-instance.vcpus, ram=-instance.memory_mb)
self._complete_deletion(context,
instance,
bdms,
quotas,
instance.system_metadata)
def _complete_deletion(self, context, instance, bdms,
quotas, system_meta):
if quotas:
quotas.commit()
# ensure block device mappings are not leaked
for bdm in bdms:
bdm.destroy()
self._notify_about_instance_usage(context, instance, "delete.end",
system_metadata=system_meta)
self._clean_instance_console_tokens(context, instance)
self._delete_scheduler_instance_info(context, instance.uuid)
def _create_reservations(self, context, instance, project_id, user_id):
vcpus = instance.vcpus
mem_mb = instance.memory_mb
quotas = objects.Quotas(context=context)
quotas.reserve(project_id=project_id,
user_id=user_id,
instances=-1,
cores=-vcpus,
ram=-mem_mb)
return quotas
def _init_instance(self, context, instance):
'''Initialize this instance during service init.'''
# NOTE(danms): If the instance appears to not be owned by this
# host, it may have been evacuated away, but skipped by the
# evacuation cleanup code due to configuration. Thus, if that
# is a possibility, don't touch the instance in any way, but
# log the concern. This will help avoid potential issues on
# startup due to misconfiguration.
if instance.host != self.host:
LOG.warning(_LW('Instance %(uuid)s appears to not be owned '
'by this host, but by %(host)s. Startup '
'processing is being skipped.'),
{'uuid': instance.uuid,
'host': instance.host})
return
# Instances that are shut down, or in an error state can not be
# initialized and are not attempted to be recovered. The exception
# to this are instances that are in RESIZE_MIGRATING or DELETING,
# which are dealt with further down.
if (instance.vm_state == vm_states.SOFT_DELETED or
(instance.vm_state == vm_states.ERROR and
instance.task_state not in
(task_states.RESIZE_MIGRATING, task_states.DELETING))):
LOG.debug("Instance is in %s state.",
instance.vm_state, instance=instance)
return
if instance.vm_state == vm_states.DELETED:
try:
self._complete_partial_deletion(context, instance)
except Exception:
# we don't want that an exception blocks the init_host
msg = _LE('Failed to complete a deletion')
LOG.exception(msg, instance=instance)
return
if (instance.vm_state == vm_states.BUILDING or
instance.task_state in [task_states.SCHEDULING,
task_states.BLOCK_DEVICE_MAPPING,
task_states.NETWORKING,
task_states.SPAWNING]):
# NOTE(dave-mcnally) compute stopped before instance was fully
# spawned so set to ERROR state. This is safe to do as the state
# may be set by the api but the host is not so if we get here the
# instance has already been scheduled to this particular host.
LOG.debug("Instance failed to spawn correctly, "
"setting to ERROR state", instance=instance)
instance.task_state = None
instance.vm_state = vm_states.ERROR
instance.save()
return
if (instance.vm_state in [vm_states.ACTIVE, vm_states.STOPPED] and
instance.task_state in [task_states.REBUILDING,
task_states.REBUILD_BLOCK_DEVICE_MAPPING,
task_states.REBUILD_SPAWNING]):
# NOTE(jichenjc) compute stopped before instance was fully
# spawned so set to ERROR state. This is consistent to BUILD
LOG.debug("Instance failed to rebuild correctly, "
"setting to ERROR state", instance=instance)
instance.task_state = None
instance.vm_state = vm_states.ERROR
instance.save()
return
if (instance.vm_state != vm_states.ERROR and
instance.task_state in [task_states.IMAGE_SNAPSHOT_PENDING,
task_states.IMAGE_PENDING_UPLOAD,
task_states.IMAGE_UPLOADING,
task_states.IMAGE_SNAPSHOT]):
LOG.debug("Instance in transitional state %s at start-up "
"clearing task state",
instance.task_state, instance=instance)
try:
self._post_interrupted_snapshot_cleanup(context, instance)
except Exception:
# we don't want that an exception blocks the init_host
msg = _LE('Failed to cleanup snapshot.')
LOG.exception(msg, instance=instance)
instance.task_state = None
instance.save()
if (instance.vm_state != vm_states.ERROR and
instance.task_state in [task_states.RESIZE_PREP]):
LOG.debug("Instance in transitional state %s at start-up "
"clearing task state",
instance['task_state'], instance=instance)
instance.task_state = None
instance.save()
if instance.task_state == task_states.DELETING:
try:
LOG.info(_LI('Service started deleting the instance during '
'the previous run, but did not finish. Restarting'
' the deletion now.'), instance=instance)
instance.obj_load_attr('metadata')
instance.obj_load_attr('system_metadata')
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
project_id, user_id = objects.quotas.ids_from_instance(
context, instance)
quotas = self._create_reservations(context, instance,
project_id, user_id)
self._delete_instance(context, instance, bdms, quotas)
except Exception:
# we don't want that an exception blocks the init_host
msg = _LE('Failed to complete a deletion')
LOG.exception(msg, instance=instance)
self._set_instance_obj_error_state(context, instance)
return
try_reboot, reboot_type = self._retry_reboot(context, instance)
current_power_state = self._get_power_state(context, instance)
if try_reboot:
LOG.debug("Instance in transitional state (%(task_state)s) at "
"start-up and power state is (%(power_state)s), "
"triggering reboot",
{'task_state': instance.task_state,
'power_state': current_power_state},
instance=instance)
# NOTE(mikal): if the instance was doing a soft reboot that got as
# far as shutting down the instance but not as far as starting it
# again, then we've just become a hard reboot. That means the
# task state for the instance needs to change so that we're in one
# of the expected task states for a hard reboot.
soft_types = [task_states.REBOOT_STARTED,
task_states.REBOOT_PENDING,
task_states.REBOOTING]
if instance.task_state in soft_types and reboot_type == 'HARD':
instance.task_state = task_states.REBOOT_PENDING_HARD
self.reboot_instance(context, instance, block_device_info=None,
reboot_type=reboot_type)
return
elif (current_power_state == power_state.RUNNING and
instance.task_state in [task_states.REBOOT_STARTED,
task_states.REBOOT_STARTED_HARD,
task_states.PAUSING,
task_states.UNPAUSING]):
LOG.warning(_LW("Instance in transitional state "
"(%(task_state)s) at start-up and power state "
"is (%(power_state)s), clearing task state"),
{'task_state': instance.task_state,
'power_state': current_power_state},
instance=instance)
instance.task_state = None
instance.vm_state = vm_states.ACTIVE
instance.save()
elif (current_power_state == power_state.PAUSED and
instance.task_state == task_states.UNPAUSING):
LOG.warning(_LW("Instance in transitional state "
"(%(task_state)s) at start-up and power state "
"is (%(power_state)s), clearing task state "
"and unpausing the instance"),
{'task_state': instance.task_state,
'power_state': current_power_state},
instance=instance)
try:
self.unpause_instance(context, instance)
except NotImplementedError:
# Some virt driver didn't support pause and unpause
pass
except Exception:
LOG.exception(_LE('Failed to unpause instance'),
instance=instance)
return
if instance.task_state == task_states.POWERING_OFF:
try:
LOG.debug("Instance in transitional state %s at start-up "
"retrying stop request",
instance.task_state, instance=instance)
self.stop_instance(context, instance, True)
except Exception:
# we don't want that an exception blocks the init_host
msg = _LE('Failed to stop instance')
LOG.exception(msg, instance=instance)
return
if instance.task_state == task_states.POWERING_ON:
try:
LOG.debug("Instance in transitional state %s at start-up "
"retrying start request",
instance.task_state, instance=instance)
self.start_instance(context, instance)
except Exception:
# we don't want that an exception blocks the init_host
msg = _LE('Failed to start instance')
LOG.exception(msg, instance=instance)
return
net_info = compute_utils.get_nw_info_for_instance(instance)
try:
self.driver.plug_vifs(instance, net_info)
except NotImplementedError as e:
LOG.debug(e, instance=instance)
except exception.VirtualInterfacePlugException:
# we don't want an exception to block the init_host
LOG.exception(_LE("Vifs plug failed"), instance=instance)
self._set_instance_obj_error_state(context, instance)
return
if instance.task_state == task_states.RESIZE_MIGRATING:
# We crashed during resize/migration, so roll back for safety
try:
# NOTE(mriedem): check old_vm_state for STOPPED here, if it's
# not in system_metadata we default to True for backwards
# compatibility
power_on = (instance.system_metadata.get('old_vm_state') !=
vm_states.STOPPED)
block_dev_info = self._get_instance_block_device_info(context,
instance)
self.driver.finish_revert_migration(context,
instance, net_info, block_dev_info, power_on)
except Exception:
LOG.exception(_LE('Failed to revert crashed migration'),
instance=instance)
finally:
LOG.info(_LI('Instance found in migrating state during '
'startup. Resetting task_state'),
instance=instance)
instance.task_state = None
instance.save()
if instance.task_state == task_states.MIGRATING:
# Live migration did not complete, but instance is on this
# host, so reset the state.
instance.task_state = None
instance.save(expected_task_state=[task_states.MIGRATING])
db_state = instance.power_state
drv_state = self._get_power_state(context, instance)
expect_running = (db_state == power_state.RUNNING and
drv_state != db_state)
LOG.debug('Current state is %(drv_state)s, state in DB is '
'%(db_state)s.',
{'drv_state': drv_state, 'db_state': db_state},
instance=instance)
if expect_running and CONF.resume_guests_state_on_host_boot:
LOG.info(_LI('Rebooting instance after nova-compute restart.'),
instance=instance)
block_device_info = \
self._get_instance_block_device_info(context, instance)
try:
self.driver.resume_state_on_host_boot(
context, instance, net_info, block_device_info)
except NotImplementedError:
LOG.warning(_LW('Hypervisor driver does not support '
'resume guests'), instance=instance)
except Exception:
# NOTE(vish): The instance failed to resume, so we set the
# instance to error and attempt to continue.
LOG.warning(_LW('Failed to resume instance'),
instance=instance)
self._set_instance_obj_error_state(context, instance)
elif drv_state == power_state.RUNNING:
# VMwareAPI drivers will raise an exception
try:
self.driver.ensure_filtering_rules_for_instance(
instance, net_info)
except NotImplementedError:
LOG.warning(_LW('Hypervisor driver does not support '
'firewall rules'), instance=instance)
def _retry_reboot(self, context, instance):
current_power_state = self._get_power_state(context, instance)
current_task_state = instance.task_state
retry_reboot = False
reboot_type = compute_utils.get_reboot_type(current_task_state,
current_power_state)
pending_soft = (current_task_state == task_states.REBOOT_PENDING and
instance.vm_state in vm_states.ALLOW_SOFT_REBOOT)
pending_hard = (current_task_state == task_states.REBOOT_PENDING_HARD
and instance.vm_state in vm_states.ALLOW_HARD_REBOOT)
started_not_running = (current_task_state in
[task_states.REBOOT_STARTED,
task_states.REBOOT_STARTED_HARD] and
current_power_state != power_state.RUNNING)
if pending_soft or pending_hard or started_not_running:
retry_reboot = True
return retry_reboot, reboot_type
def handle_lifecycle_event(self, event):
LOG.info(_LI("VM %(state)s (Lifecycle Event)"),
{'state': event.get_name()},
instance_uuid=event.get_instance_uuid())
context = nova.context.get_admin_context(read_deleted='yes')
instance = objects.Instance.get_by_uuid(context,
event.get_instance_uuid(),
expected_attrs=[])
vm_power_state = None
if event.get_transition() == virtevent.EVENT_LIFECYCLE_STOPPED:
vm_power_state = power_state.SHUTDOWN
elif event.get_transition() == virtevent.EVENT_LIFECYCLE_STARTED:
vm_power_state = power_state.RUNNING
elif event.get_transition() == virtevent.EVENT_LIFECYCLE_PAUSED:
vm_power_state = power_state.PAUSED
elif event.get_transition() == virtevent.EVENT_LIFECYCLE_RESUMED:
vm_power_state = power_state.RUNNING
elif event.get_transition() == virtevent.EVENT_LIFECYCLE_SUSPENDED:
vm_power_state = power_state.SUSPENDED
else:
LOG.warning(_LW("Unexpected power state %d"),
event.get_transition())
# Note(lpetrut): The event may be delayed, thus not reflecting
# the current instance power state. In that case, ignore the event.
current_power_state = self._get_power_state(context, instance)
if current_power_state == vm_power_state:
LOG.debug('Synchronizing instance power state after lifecycle '
'event "%(event)s"; current vm_state: %(vm_state)s, '
'current task_state: %(task_state)s, current DB '
'power_state: %(db_power_state)s, VM power_state: '
'%(vm_power_state)s',
{'event': event.get_name(),
'vm_state': instance.vm_state,
'task_state': instance.task_state,
'db_power_state': instance.power_state,
'vm_power_state': vm_power_state},
instance_uuid=instance.uuid)
self._sync_instance_power_state(context,
instance,
vm_power_state)
def handle_events(self, event):
if isinstance(event, virtevent.LifecycleEvent):
try:
self.handle_lifecycle_event(event)
except exception.InstanceNotFound:
LOG.debug("Event %s arrived for non-existent instance. The "
"instance was probably deleted.", event)
else:
LOG.debug("Ignoring event %s", event)
def init_virt_events(self):
if CONF.workarounds.handle_virt_lifecycle_events:
self.driver.register_event_listener(self.handle_events)
else:
# NOTE(mriedem): If the _sync_power_states periodic task is
# disabled we should emit a warning in the logs.
if CONF.sync_power_state_interval < 0:
LOG.warn(_LW('Instance lifecycle events from the compute '
'driver have been disabled. Note that lifecycle '
'changes to an instance outside of the compute '
'service will not be synchronized '
'automatically since the _sync_power_states '
'periodic task is also disabled.'))
else:
LOG.info(_LI('Instance lifecycle events from the compute '
'driver have been disabled. Note that lifecycle '
'changes to an instance outside of the compute '
'service will only be synchronized by the '
'_sync_power_states periodic task.'))
def init_host(self):
"""Initialization for a standalone compute service."""
self.driver.init_host(host=self.host)
context = nova.context.get_admin_context()
instances = objects.InstanceList.get_by_host(
context, self.host, expected_attrs=['info_cache'])
if CONF.defer_iptables_apply:
self.driver.filter_defer_apply_on()
self.init_virt_events()
try:
# checking that instance was not already evacuated to other host
self._destroy_evacuated_instances(context)
for instance in instances:
self._init_instance(context, instance)
finally:
if CONF.defer_iptables_apply:
self.driver.filter_defer_apply_off()
self._update_scheduler_instance_info(context, instances)
def cleanup_host(self):
self.driver.register_event_listener(None)
self.instance_events.cancel_all_events()
self.driver.cleanup_host(host=self.host)
def pre_start_hook(self):
"""After the service is initialized, but before we fully bring
the service up by listening on RPC queues, make sure to update
our available resources (and indirectly our available nodes).
"""
self.update_available_resource(nova.context.get_admin_context())
def _get_power_state(self, context, instance):
"""Retrieve the power state for the given instance."""
LOG.debug('Checking state', instance=instance)
try:
return self.driver.get_info(instance).state
except exception.InstanceNotFound:
return power_state.NOSTATE
def get_console_topic(self, context):
"""Retrieves the console host for a project on this host.
Currently this is just set in the flags for each compute host.
"""
# TODO(mdragon): perhaps make this variable by console_type?
return '%s.%s' % (CONF.console_topic, CONF.console_host)
@wrap_exception()
def get_console_pool_info(self, context, console_type):
return self.driver.get_console_pool_info(console_type)
@wrap_exception()
def refresh_security_group_rules(self, context, security_group_id):
"""Tell the virtualization driver to refresh security group rules.
Passes straight through to the virtualization driver.
"""
return self.driver.refresh_security_group_rules(security_group_id)
@wrap_exception()
def refresh_security_group_members(self, context, security_group_id):
"""Tell the virtualization driver to refresh security group members.
Passes straight through to the virtualization driver.
"""
return self.driver.refresh_security_group_members(security_group_id)
@object_compat
@wrap_exception()
def refresh_instance_security_rules(self, context, instance):
"""Tell the virtualization driver to refresh security rules for
an instance.
Passes straight through to the virtualization driver.
Synchronise the call because we may still be in the middle of
creating the instance.
"""
@utils.synchronized(instance.uuid)
def _sync_refresh():
try:
return self.driver.refresh_instance_security_rules(instance)
except NotImplementedError:
LOG.warning(_LW('Hypervisor driver does not support '
'security groups.'), instance=instance)
return _sync_refresh()
@wrap_exception()
def refresh_provider_fw_rules(self, context):
"""This call passes straight through to the virtualization driver."""
return self.driver.refresh_provider_fw_rules()
def _await_block_device_map_created(self, context, vol_id):
# TODO(yamahata): creating volume simultaneously
# reduces creation time?
# TODO(yamahata): eliminate dumb polling
start = time.time()
retries = CONF.block_device_allocate_retries
if retries < 0:
LOG.warning(_LW("Treating negative config value (%(retries)s) for "
"'block_device_retries' as 0."),
{'retries': retries})
# (1) treat negative config value as 0
# (2) the configured value is 0, one attempt should be made
# (3) the configured value is > 0, then the total number attempts
# is (retries + 1)
attempts = 1
if retries >= 1:
attempts = retries + 1
for attempt in range(1, attempts + 1):
volume = self.volume_api.get(context, vol_id)
volume_status = volume['status']
if volume_status not in ['creating', 'downloading']:
if volume_status == 'available':
return attempt
LOG.warning(_LW("Volume id: %(vol_id)s finished being "
"created but its status is %(vol_status)s."),
{'vol_id': vol_id,
'vol_status': volume_status})
break
greenthread.sleep(CONF.block_device_allocate_retries_interval)
raise exception.VolumeNotCreated(volume_id=vol_id,
seconds=int(time.time() - start),
attempts=attempt,
volume_status=volume_status)
def _decode_files(self, injected_files):
"""Base64 decode the list of files to inject."""
if not injected_files:
return []
def _decode(f):
path, contents = f
try:
decoded = base64.b64decode(contents)
return path, decoded
except TypeError:
raise exception.Base64Exception(path=path)
return [_decode(f) for f in injected_files]
def _validate_instance_group_policy(self, context, instance,
filter_properties):
# NOTE(russellb) Instance group policy is enforced by the scheduler.
# However, there is a race condition with the enforcement of
# the policy. Since more than one instance may be scheduled at the
# same time, it's possible that more than one instance with an
# anti-affinity policy may end up here. It's also possible that
# multiple instances with an affinity policy could end up on different
# hosts. This is a validation step to make sure that starting the
# instance here doesn't violate the policy.
scheduler_hints = filter_properties.get('scheduler_hints') or {}
group_hint = scheduler_hints.get('group')
if not group_hint:
return
@utils.synchronized(group_hint)
def _do_validation(context, instance, group_hint):
group = objects.InstanceGroup.get_by_hint(context, group_hint)
if 'anti-affinity' in group.policies:
group_hosts = group.get_hosts(exclude=[instance.uuid])
if self.host in group_hosts:
msg = _("Anti-affinity instance group policy "
"was violated.")
raise exception.RescheduledException(
instance_uuid=instance.uuid,
reason=msg)
elif 'affinity' in group.policies:
group_hosts = group.get_hosts(exclude=[instance.uuid])
if group_hosts and self.host not in group_hosts:
msg = _("Affinity instance group policy was violated.")
raise exception.RescheduledException(
instance_uuid=instance.uuid,
reason=msg)
_do_validation(context, instance, group_hint)
def _log_original_error(self, exc_info, instance_uuid):
LOG.error(_LE('Error: %s'), exc_info[1], instance_uuid=instance_uuid,
exc_info=exc_info)
def _reschedule(self, context, request_spec, filter_properties,
instance, reschedule_method, method_args, task_state,
exc_info=None):
"""Attempt to re-schedule a compute operation."""
instance_uuid = instance.uuid
retry = filter_properties.get('retry')
if not retry:
# no retry information, do not reschedule.
LOG.debug("Retry info not present, will not reschedule",
instance_uuid=instance_uuid)
return
if not request_spec:
LOG.debug("No request spec, will not reschedule",
instance_uuid=instance_uuid)
return
LOG.debug("Re-scheduling %(method)s: attempt %(num)d",
{'method': reschedule_method.func_name,
'num': retry['num_attempts']}, instance_uuid=instance_uuid)
# reset the task state:
self._instance_update(context, instance, task_state=task_state)
if exc_info:
# stringify to avoid circular ref problem in json serialization:
retry['exc'] = traceback.format_exception_only(exc_info[0],
exc_info[1])
reschedule_method(context, *method_args)
return True
@periodic_task.periodic_task
def _check_instance_build_time(self, context):
"""Ensure that instances are not stuck in build."""
timeout = CONF.instance_build_timeout
if timeout == 0:
return
filters = {'vm_state': vm_states.BUILDING,
'host': self.host}
building_insts = objects.InstanceList.get_by_filters(context,
filters, expected_attrs=[], use_slave=True)
for instance in building_insts:
if timeutils.is_older_than(instance.created_at, timeout):
self._set_instance_obj_error_state(context, instance)
LOG.warning(_LW("Instance build timed out. Set to error "
"state."), instance=instance)
def _check_instance_exists(self, context, instance):
"""Ensure an instance with the same name is not already present."""
if self.driver.instance_exists(instance):
raise exception.InstanceExists(name=instance.name)
def _allocate_network_async(self, context, instance, requested_networks,
macs, security_groups, is_vpn, dhcp_options):
"""Method used to allocate networks in the background.
Broken out for testing.
"""
LOG.debug("Allocating IP information in the background.",
instance=instance)
retries = CONF.network_allocate_retries
if retries < 0:
LOG.warning(_LW("Treating negative config value (%(retries)s) for "
"'network_allocate_retries' as 0."),
{'retries': retries})
retries = 0
attempts = retries + 1
retry_time = 1
for attempt in range(1, attempts + 1):
try:
nwinfo = self.network_api.allocate_for_instance(
context, instance, vpn=is_vpn,
requested_networks=requested_networks,
macs=macs,
security_groups=security_groups,
dhcp_options=dhcp_options)
LOG.debug('Instance network_info: |%s|', nwinfo,
instance=instance)
instance.system_metadata['network_allocated'] = 'True'
# NOTE(JoshNang) do not save the instance here, as it can cause
# races. The caller shares a reference to instance and waits
# for this async greenthread to finish before calling
# instance.save().
return nwinfo
except Exception:
exc_info = sys.exc_info()
log_info = {'attempt': attempt,
'attempts': attempts}
if attempt == attempts:
LOG.exception(_LE('Instance failed network setup '
'after %(attempts)d attempt(s)'),
log_info)
six.reraise(*exc_info)
LOG.warning(_LW('Instance failed network setup '
'(attempt %(attempt)d of %(attempts)d)'),
log_info, instance=instance)
time.sleep(retry_time)
retry_time *= 2
if retry_time > 30:
retry_time = 30
# Not reached.
def _build_networks_for_instance(self, context, instance,
requested_networks, security_groups):
# If we're here from a reschedule the network may already be allocated.
if strutils.bool_from_string(
instance.system_metadata.get('network_allocated', 'False')):
# NOTE(alex_xu): The network_allocated is True means the network
# resource already allocated at previous scheduling, and the
# network setup is cleanup at previous. After rescheduling, the
# network resource need setup on the new host.
self.network_api.setup_instance_network_on_host(
context, instance, instance.host)
return self.network_api.get_instance_nw_info(context, instance)
if not self.is_neutron_security_groups:
security_groups = []
macs = self.driver.macs_for_instance(instance)
dhcp_options = self.driver.dhcp_options_for_instance(instance)
network_info = self._allocate_network(context, instance,
requested_networks, macs, security_groups, dhcp_options)
return network_info
def _allocate_network(self, context, instance, requested_networks, macs,
security_groups, dhcp_options):
"""Start network allocation asynchronously. Return an instance
of NetworkInfoAsyncWrapper that can be used to retrieve the
allocated networks when the operation has finished.
"""
# NOTE(comstud): Since we're allocating networks asynchronously,
# this task state has little meaning, as we won't be in this
# state for very long.
instance.vm_state = vm_states.BUILDING
instance.task_state = task_states.NETWORKING
instance.save(expected_task_state=[None])
self._update_resource_tracker(context, instance)
is_vpn = pipelib.is_vpn_image(instance.image_ref)
return network_model.NetworkInfoAsyncWrapper(
self._allocate_network_async, context, instance,
requested_networks, macs, security_groups, is_vpn,
dhcp_options)
def _default_root_device_name(self, instance, image_meta, root_bdm):
try:
return self.driver.default_root_device_name(instance,
image_meta,
root_bdm)
except NotImplementedError:
return compute_utils.get_next_device_name(instance, [])
def _default_device_names_for_instance(self, instance,
root_device_name,
*block_device_lists):
try:
self.driver.default_device_names_for_instance(instance,
root_device_name,
*block_device_lists)
except NotImplementedError:
compute_utils.default_device_names_for_instance(
instance, root_device_name, *block_device_lists)
def _get_device_name_for_instance(self, instance, bdms, block_device_obj):
# NOTE(ndipanov): Copy obj to avoid changing the original
block_device_obj = block_device_obj.obj_clone()
try:
return self.driver.get_device_name_for_instance(
instance, bdms, block_device_obj)
except NotImplementedError:
return compute_utils.get_device_name_for_instance(
instance, bdms, block_device_obj.get("device_name"))
def _default_block_device_names(self, context, instance,
image_meta, block_devices):
"""Verify that all the devices have the device_name set. If not,
provide a default name.
It also ensures that there is a root_device_name and is set to the
first block device in the boot sequence (boot_index=0).
"""
root_bdm = block_device.get_root_bdm(block_devices)
if not root_bdm:
return
# Get the root_device_name from the root BDM or the instance
root_device_name = None
update_root_bdm = False
if root_bdm.device_name:
root_device_name = root_bdm.device_name
instance.root_device_name = root_device_name
elif instance.root_device_name:
root_device_name = instance.root_device_name
root_bdm.device_name = root_device_name
update_root_bdm = True
else:
root_device_name = self._default_root_device_name(instance,
image_meta,
root_bdm)
instance.root_device_name = root_device_name
root_bdm.device_name = root_device_name
update_root_bdm = True
if update_root_bdm:
root_bdm.save()
ephemerals = filter(block_device.new_format_is_ephemeral,
block_devices)
swap = filter(block_device.new_format_is_swap,
block_devices)
block_device_mapping = filter(
driver_block_device.is_block_device_mapping, block_devices)
self._default_device_names_for_instance(instance,
root_device_name,
ephemerals,
swap,
block_device_mapping)
def _block_device_info_to_legacy(self, block_device_info):
"""Convert BDI to the old format for drivers that need it."""
if self.use_legacy_block_device_info:
ephemerals = driver_block_device.legacy_block_devices(
driver.block_device_info_get_ephemerals(block_device_info))
mapping = driver_block_device.legacy_block_devices(
driver.block_device_info_get_mapping(block_device_info))
swap = block_device_info['swap']
if swap:
swap = swap.legacy()
block_device_info.update({
'ephemerals': ephemerals,
'swap': swap,
'block_device_mapping': mapping})
def _prep_block_device(self, context, instance, bdms,
do_check_attach=True):
"""Set up the block device for an instance with error logging."""
try:
block_device_info = driver.get_block_device_info(instance, bdms)
mapping = driver.block_device_info_get_mapping(block_device_info)
driver_block_device.attach_block_devices(
mapping, context, instance, self.volume_api, self.driver,
do_check_attach=do_check_attach,
wait_func=self._await_block_device_map_created)
self._block_device_info_to_legacy(block_device_info)
return block_device_info
except exception.OverQuota:
msg = _LW('Failed to create block device for instance due to '
'being over volume resource quota')
LOG.warn(msg, instance=instance)
raise exception.VolumeLimitExceeded()
except Exception:
LOG.exception(_LE('Instance failed block device setup'),
instance=instance)
raise exception.InvalidBDM()
def _update_instance_after_spawn(self, context, instance):
instance.power_state = self._get_power_state(context, instance)
instance.vm_state = vm_states.ACTIVE
instance.task_state = None
instance.launched_at = timeutils.utcnow()
configdrive.update_instance(instance)
def _update_scheduler_instance_info(self, context, instance):
"""Sends an InstanceList with created or updated Instance objects to
the Scheduler client.
In the case of init_host, the value passed will already be an
InstanceList. Other calls will send individual Instance objects that
have been created or resized. In this case, we create an InstanceList
object containing that Instance.
"""
if not self.send_instance_updates:
return
if isinstance(instance, objects.Instance):
instance = objects.InstanceList(objects=[instance])
context = context.elevated()
self.scheduler_client.update_instance_info(context, self.host,
instance)
def _delete_scheduler_instance_info(self, context, instance_uuid):
"""Sends the uuid of the deleted Instance to the Scheduler client."""
if not self.send_instance_updates:
return
context = context.elevated()
self.scheduler_client.delete_instance_info(context, self.host,
instance_uuid)
@periodic_task.periodic_task(spacing=CONF.scheduler_instance_sync_interval)
def _sync_scheduler_instance_info(self, context):
if not self.send_instance_updates:
return
context = context.elevated()
instances = objects.InstanceList.get_by_host(context, self.host,
expected_attrs=[],
use_slave=True)
uuids = [instance.uuid for instance in instances]
self.scheduler_client.sync_instance_info(context, self.host, uuids)
def _notify_about_instance_usage(self, context, instance, event_suffix,
network_info=None, system_metadata=None,
extra_usage_info=None, fault=None):
compute_utils.notify_about_instance_usage(
self.notifier, context, instance, event_suffix,
network_info=network_info,
system_metadata=system_metadata,
extra_usage_info=extra_usage_info, fault=fault)
def _deallocate_network(self, context, instance,
requested_networks=None):
LOG.debug('Deallocating network for instance', instance=instance)
self.network_api.deallocate_for_instance(
context, instance, requested_networks=requested_networks)
def _get_instance_block_device_info(self, context, instance,
refresh_conn_info=False,
bdms=None):
"""Transform block devices to the driver block_device format."""
if not bdms:
bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
context, instance.uuid)
block_device_info = driver.get_block_device_info(instance, bdms)
if not refresh_conn_info:
# if the block_device_mapping has no value in connection_info
# (returned as None), don't include in the mapping
block_device_info['block_device_mapping'] = [
bdm for bdm in driver.block_device_info_get_mapping(
block_device_info)
if bdm.get('connection_info')]
else:
driver_block_device.refresh_conn_infos(
driver.block_device_info_get_mapping(block_device_info),
context, instance, self.volume_api, self.driver)
self._block_device_info_to_legacy(block_device_info)
return block_device_info
@wrap_exception()
@reverts_task_state
@wrap_instance_fault
def build_and_run_instance(self, context, instance, image, request_spec,
filter_properties, admin_password=None,
injected_files=None, requested_networks=None,
security_groups=None, block_device_mapping=None,
node=None, limits=None):
@utils.synchronized(instance.uuid)
def _locked_do_build_and_run_instance(*args, **kwargs):
# NOTE(danms): We grab the semaphore with the instance uuid
# locked because we could wait in line to build this instance
# for a while and we want to make sure that nothing else tries
# to do anything with this instance while we wait.
with self._build_semaphore:
self._do_build_and_run_instance(*args, **kwargs)
# NOTE(danms): We spawn here to return the RPC worker thread back to
# the pool. Since what follows could take a really long time, we don't
# want to tie up RPC workers.
utils.spawn_n(_locked_do_build_and_run_instance,
context, instance, image, request_spec,
filter_properties, admin_password, injected_files,
requested_networks, security_groups,
block_device_mapping, node, limits)
@hooks.add_hook('build_instance')
@wrap_exception()
@reverts_task_state
@wrap_instance_event
@wrap_instance_fault
def _do_build_and_run_instance(self, context, instance, image,
request_spec, filter_properties, admin_password, injected_files,
requested_networks, security_groups, block_device_mapping,
node=None, limits=None):
try:
LOG.info(_LI('Starting instance...'), context=context,
instance=instance)
instance.vm_state = vm_states.BUILDING
instance.task_state = None
instance.save(expected_task_state=
(task_states.SCHEDULING, None))
except exception.InstanceNotFound:
msg = 'Instance disappeared before build.'
LOG.debug(msg, instance=instance)
return build_results.FAILED
except exception.UnexpectedTaskStateError as e:
LOG.debug(e.format_message(), instance=instance)
return build_results.FAILED
# b64 decode the files to inject:
decoded_files = self._decode_files(injected_files)
if limits is None:
limits = {}
if node is None:
node = self.driver.get_available_nodes(refresh=True)[0]
LOG.debug('No node specified, defaulting to %s', node,
instance=instance)
try:
self._build_and_run_instance(context, instance, image,
decoded_files, admin_password, requested_networks,
security_groups, block_device_mapping, node, limits,
filter_properties)
return build_results.ACTIVE
except exception.RescheduledException as e:
retry = filter_properties.get('retry')
if not retry:
# no retry information, do not reschedule.
LOG.debug("Retry info not present, will not reschedule",
instance=instance)
self._cleanup_allocated_networks(context, instance,
requested_networks)
compute_utils.add_instance_fault_from_exc(context,
instance, e, sys.exc_info())
self._nil_out_instance_obj_host_and_node(instance)
self._set_instance_obj_error_state(context, instance,
clean_task_state=True)
return build_results.FAILED
LOG.debug(e.format_message(), instance=instance)
retry['exc'] = traceback.format_exception(*sys.exc_info())
# NOTE(comstud): Deallocate networks if the driver wants
# us to do so.
if self.driver.deallocate_networks_on_reschedule(instance):
self._cleanup_allocated_networks(context, instance,
requested_networks)
else:
# NOTE(alex_xu): Network already allocated and we don't
# want to deallocate them before rescheduling. But we need
# to cleanup those network resources setup on this host before
# rescheduling.
self.network_api.cleanup_instance_network_on_host(
context, instance, self.host)
self._nil_out_instance_obj_host_and_node(instance)
instance.task_state = task_states.SCHEDULING
instance.save()
self.compute_task_api.build_instances(context, [instance],
image, filter_properties, admin_password,
injected_files, requested_networks, security_groups,
block_device_mapping)
return build_results.RESCHEDULED
except (exception.InstanceNotFound,
exception.UnexpectedDeletingTaskStateError):
msg = 'Instance disappeared during build.'
LOG.debug(msg, instance=instance)
self._cleanup_allocated_networks(context, instance,
requested_networks)
return build_results.FAILED
except exception.BuildAbortException as e:
LOG.exception(e.format_message(), instance=instance)
self._cleanup_allocated_networks(context, instance,
requested_networks)
self._cleanup_volumes(context, instance.uuid,
block_device_mapping, raise_exc=False)
compute_utils.add_instance_fault_from_exc(context, instance,
e, sys.exc_info())
self._nil_out_instance_obj_host_and_node(instance)
self._set_instance_obj_error_state(context, instance,
clean_task_state=True)
return build_results.FAILED
except Exception as e:
# Should not reach here.
msg = _LE('Unexpected build failure, not rescheduling build.')
LOG.exception(msg, instance=instance)
self._cleanup_allocated_networks(context, instance,
requested_networks)
self._cleanup_volumes(context, instance.uuid,
block_device_mapping, raise_exc=False)
compute_utils.add_instance_fault_from_exc(context, instance,
e, sys.exc_info())
self._nil_out_instance_obj_host_and_node(instance)
self._set_instance_obj_error_state(context, instance,
clean_task_state=True)
return build_results.FAILED
def _build_and_run_instance(self, context, instance, image, injected_files,
admin_password, requested_networks, security_groups,
block_device_mapping, node, limits, filter_properties):
image_name = image.get('name')
self._notify_about_instance_usage(context, instance, 'create.start',
extra_usage_info={'image_name': image_name})
try:
rt = self._get_resource_tracker(node)
with rt.instance_claim(context, instance, limits):
# NOTE(russellb) It's important that this validation be done
# *after* the resource tracker instance claim, as that is where
# the host is set on the instance.
self._validate_instance_group_policy(context, instance,
filter_properties)
with self._build_resources(context, instance,
requested_networks, security_groups, image,
block_device_mapping) as resources:
instance.vm_state = vm_states.BUILDING
instance.task_state = task_states.SPAWNING
# NOTE(JoshNang) This also saves the changes to the
# instance from _allocate_network_async, as they aren't
# saved in that function to prevent races.
instance.save(expected_task_state=
task_states.BLOCK_DEVICE_MAPPING)
block_device_info = resources['block_device_info']
network_info = resources['network_info']
self.driver.spawn(context, instance, image,
injected_files, admin_password,
network_info=network_info,
block_device_info=block_device_info)
except (exception.InstanceNotFound,
exception.UnexpectedDeletingTaskStateError) as e:
with excutils.save_and_reraise_exception():
self._notify_about_instance_usage(context, instance,
'create.end', fault=e)
except exception.ComputeResourcesUnavailable as e:
LOG.debug(e.format_message(), instance=instance)
self._notify_about_instance_usage(context, instance,
'create.error', fault=e)
raise exception.RescheduledException(
instance_uuid=instance.uuid, reason=e.format_message())
except exception.BuildAbortException as e:
with excutils.save_and_reraise_exception():
LOG.debug(e.format_message(), instance=instance)
self._notify_about_instance_usage(context, instance,
'create.error', fault=e)
except (exception.FixedIpLimitExceeded,
exception.NoMoreNetworks, exception.NoMoreFixedIps) as e:
LOG.warning(_LW('No more network or fixed IP to be allocated'),
instance=instance)
self._notify_about_instance_usage(context, instance,
'create.error', fault=e)
msg = _('Failed to allocate the network(s) with error %s, '
'not rescheduling.') % e.format_message()
raise exception.BuildAbortException(instance_uuid=instance.uuid,
reason=msg)
except (exception.VirtualInterfaceCreateException,
exception.VirtualInterfaceMacAddressException) as e:
LOG.exception(_LE('Failed to allocate network(s)'),
instance=instance)
self._notify_about_instance_usage(context, instance,
'create.error', fault=e)
msg = _('Failed to allocate the network(s), not rescheduling.')
raise exception.BuildAbortException(instance_uuid=instance.uuid,
reason=msg)
except (exception.FlavorDiskTooSmall,
exception.FlavorMemoryTooSmall,
exception.ImageNotActive,
exception.ImageUnacceptable) as e:
self._notify_about_instance_usage(context, instance,
'create.error', fault=e)
raise exception.BuildAbortException(instance_uuid=instance.uuid,
reason=e.format_message())
except Exception as e:
self._notify_about_instance_usage(context, instance,
'create.error', fault=e)
raise exception.RescheduledException(
instance_uuid=instance.uuid, reason=six.text_type(e))
# NOTE(alaski): This is only useful during reschedules, remove it now.
instance.system_metadata.pop('network_allocated', None)
# If CONF.default_access_ip_network_name is set, grab the
# corresponding network and set the access ip values accordingly.
network_name = CONF.default_access_ip_network_name
if (network_name and not instance.access_ip_v4 and
not instance.access_ip_v6):
# Note that when there are multiple ips to choose from, an
# arbitrary one will be chosen.
for vif in network_info:
if vif['network']['label'] == network_name:
for ip in vif.fixed_ips():
if not instance.access_ip_v4 and ip['version'] == 4:
instance.access_ip_v4 = ip['address']
if not instance.access_ip_v6 and ip['version'] == 6:
instance.access_ip_v6 = ip['address']
break
self._update_instance_after_spawn(context, instance)
try:
instance.save(expected_task_state=task_states.SPAWNING)
except (exception.InstanceNotFound,
exception.UnexpectedDeletingTaskStateError) as e:
with excutils.save_and_reraise_exception():
self._notify_about_instance_usage(context, instance,
'create.end', fault=e)
self._update_scheduler_instance_info(context, instance)
self._notify_about_instance_usage(context, instance, 'create.end',
extra_usage_info={'message': _('Success')},
network_info=network_info)
@contextlib.contextmanager
def _build_resources(self, context, instance, requested_networks,
security_groups, image, block_device_mapping):
resources = {}
network_info = None
try:
network_info = self._build_networks_for_instance(context, instance,
requested_networks, security_groups)
resources['network_info'] = network_info
except (exception.InstanceNotFound,
exception.UnexpectedDeletingTaskStateError):
raise
except exception.UnexpectedTaskStateError as e:
raise exception.BuildAbortException(instance_uuid=instance.uuid,
reason=e.format_message())
except Exception:
# Because this allocation is async any failures are likely to occur
# when the driver accesses network_info during spawn().
LOG.exception(_LE('Failed to allocate network(s)'),
instance=instance)
msg = _('Failed to allocate the network(s), not rescheduling.')
raise exception.BuildAbortException(instance_uuid=instance.uuid,
reason=msg)
try:
# Verify that all the BDMs have a device_name set and assign a
# default to the ones missing it with the help of the driver.
self._default_block_device_names(context, instance, image,
block_device_mapping)
instance.vm_state = vm_states.BUILDING
instance.task_state = task_states.BLOCK_DEVICE_MAPPING
instance.save()
block_device_info = self._prep_block_device(context, instance,
block_device_mapping)
resources['block_device_info'] = block_device_info
except (exception.InstanceNotFound,
exception.UnexpectedDeletingTaskStateError):
with excutils.save_and_reraise_exception():
# Make sure the async call finishes
if network_info is not None:
network_info.wait(do_raise=False)
except (exception.UnexpectedTaskStateError,
exception.VolumeLimitExceeded,
exception.InvalidBDM) as e:
# Make sure the async call finishes
if network_info is not None:
network_info.wait(do_raise=False)
raise exception.BuildAbortException(instance_uuid=instance.uuid,
reason=e.format_message())
except Exception:
LOG.exception(_LE('Failure prepping block device'),
instance=instance)
# Make sure the async call finishes
if network_info is not None:
network_info.wait(do_raise=False)
msg = _('Failure prepping block device.')
raise exception.BuildAbortException(instance_uuid=instance.uuid,
reason=msg)
try:
yield resources
except Exception as exc:
with excutils.save_and_reraise_exception() as ctxt:
if not isinstance(exc, (exception.InstanceNotFound,
exception.UnexpectedDeletingTaskStateError)):
LOG.exception(_LE('Instance failed to spawn'),
instance=instance)
# Make sure the async call finishes
if network_info is not None:
network_info.wait(do_raise=False)
# if network_info is empty we're likely here because of
# network allocation failure. Since nothing can be reused on
# rescheduling it's better to deallocate network to eliminate
# the chance of orphaned ports in neutron
deallocate_networks = False if network_info else True
try:
self._shutdown_instance(context, instance,
block_device_mapping, requested_networks,
try_deallocate_networks=deallocate_networks)
except Exception:
ctxt.reraise = False
msg = _('Could not clean up failed build,'
' not rescheduling')
raise exception.BuildAbortException(
instance_uuid=instance.uuid, reason=msg)
def _cleanup_allocated_networks(self, context, instance,
requested_networks):
try:
self._deallocate_network(context, instance, requested_networks)
except Exception:
msg = _LE('Failed to deallocate networks')
LOG.exception(msg, instance=instance)
return
instance.system_metadata['network_allocated'] = 'False'
try:
instance.save()
except exception.InstanceNotFound:
# NOTE(alaski): 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
def _try_deallocate_network(self, context, instance,
requested_networks=None):
try:
# tear down allocated network structure
self._deallocate_network(context, instance, requested_networks)
except Exception:
with excutils.save_and_reraise_exception():
LOG.error(_LE('Failed to deallocate network for instance.'),
instance=instance)
self._set_instance_obj_error_state(context, instance)
def _get_power_off_values(self, context, instance, clean_shutdown):
"""Get the timing configuration for powering down this instance."""
if clean_shutdown:
timeout = compute_utils.get_value_from_system_metadata(instance,
key='image_os_shutdown_timeout', type=int,
default=CONF.shutdown_timeout)
retry_interval = self.SHUTDOWN_RETRY_INTERVAL
else:
timeout = 0
retry_interval = 0
return timeout, retry_interval
def _power_off_instance(self, context, instance, clean_shutdown=True):
"""Power off an instance on this host."""
timeout, retry_interval = self._get_power_off_values(context,
instance, clean_shutdown)
self.driver.power_off(instance, timeout, retry_interval)
def _shutdown_instance(self, context, instance,
bdms, requested_networks=None, notify=True,
try_deallocate_networks=True):
"""Shutdown an instance on this host.
:param:context: security context
:param:instance: a nova.objects.Instance object
:param:bdms: the block devices for the instance to be torn
down
:param:requested_networks: the networks on which the instance
has ports
:param:notify: true if a final usage notification should be
emitted
:param:try_deallocate_networks: false if we should avoid
trying to teardown networking
"""
context = context.elevated()
LOG.info(_LI('%(action_str)s instance') %
{'action_str': 'Terminating'},
context=context, instance=instance)
if notify:
self._notify_about_instance_usage(context, instance,
"shutdown.start")
network_info = compute_utils.get_nw_info_for_instance(instance)
# NOTE(vish) get bdms before destroying the instance
vol_bdms = [bdm for bdm in bdms if bdm.is_volume]
block_device_info = self._get_instance_block_device_info(
context, instance, bdms=bdms)
# NOTE(melwitt): attempt driver destroy before releasing ip, may
# want to keep ip allocated for certain failures
try:
self.driver.destroy(context, instance, network_info,
block_device_info)
except exception.InstancePowerOffFailure:
# if the instance can't power off, don't release the ip
with excutils.save_and_reraise_exception():
pass
except Exception:
with excutils.save_and_reraise_exception():
# deallocate ip and fail without proceeding to
# volume api calls, preserving current behavior
if try_deallocate_networks:
self._try_deallocate_network(context, instance,
requested_networks)
if try_deallocate_networks:
self._try_deallocate_network(context, instance, requested_networks)
for bdm in vol_bdms:
try:
# NOTE(vish): actual driver detach done in driver.destroy, so
# just tell cinder that we are done with it.
connector = self.driver.get_volume_connector(instance)
self.volume_api.terminate_connection(context,
bdm.volume_id,
connector)
self.volume_api.detach(context, bdm.volume_id)
except exception.DiskNotFound as exc:
LOG.debug('Ignoring DiskNotFound: %s', exc,
instance=instance)
except exception.VolumeNotFound as exc:
LOG.debug('Ignoring VolumeNotFound: %s', exc,
instance=instance)
except (cinder_exception.EndpointNotFound,
keystone_exception.EndpointNotFound) as exc:
LOG.warning(_LW('Ignoring EndpointNotFound: %s'), exc,
instance=instance)
except cinder_exception.ClientException as exc:
LOG.warning(_LW('Ignoring Unknown cinder exception: %s'), exc,
instance=instance)
if notify:
self._notify_about_instance_usage(context, instance,
"shutdown.end")
def _cleanup_volumes(self, context, instance_uuid, bdms, raise_exc=True):
exc_info = None
for bdm in bdms:
LOG.debug("terminating bdm %s", bdm,
instance_uuid=instance_uuid)
if bdm.volume_id and bdm.delete_on_termination:
try:
self.volume_api.delete(context, bdm.volume_id)
except Exception as exc:
exc_info = sys.exc_info()
LOG.warning(_LW('Failed to delete volume: %(volume_id)s '
'due to %(exc)s'),
{'volume_id': bdm.volume_id, 'exc': exc})
if exc_info is not None and raise_exc:
six.reraise(exc_info[0], exc_info[1], exc_info[2])
@hooks.add_hook("delete_instance")
def _delete_instance(self, context, instance, bdms, quotas):
"""Delete an instance on this host. Commit or rollback quotas
as necessary.
:param context: nova request context
:param instance: nova.objects.instance.Instance object
:param bdms: nova.objects.block_device.BlockDeviceMappingList object
:param quotas: nova.objects.quotas.Quotas object
"""
was_soft_deleted = instance.vm_state == vm_states.SOFT_DELETED
if was_soft_deleted:
# Instances in SOFT_DELETED vm_state have already had quotas
# decremented.
try:
quotas.rollback()
except Exception:
pass
try:
events = self.instance_events.clear_events_for_instance(instance)
if events:
LOG.debug('Events pending at deletion: %(events)s',
{'events': ','.join(events.keys())},
instance=instance)
self._notify_about_instance_usage(context, instance,
"delete.start")
self._shutdown_instance(context, instance, bdms)
# NOTE(dims): instance.info_cache.delete() should be called after
# _shutdown_instance in the compute manager as shutdown calls
# deallocate_for_instance so the info_cache is still needed
# at this point.
if instance.info_cache is not None:
instance.info_cache.delete()
else:
# NOTE(yoshimatsu): Avoid AttributeError if instance.info_cache
# is None. When the root cause that instance.info_cache becomes
# None is fixed, the log level should be reconsidered.
LOG.warning(_LW("Info cache for instance could not be found. "
"Ignore."), instance=instance)
# NOTE(vish): We have already deleted the instance, so we have
# to ignore problems cleaning up the volumes. It
# would be nice to let the user know somehow that
# the volume deletion failed, but it is not
# acceptable to have an instance that can not be
# deleted. Perhaps this could be reworked in the
# future to set an instance fault the first time
# and to only ignore the failure if the instance
# is already in ERROR.
self._cleanup_volumes(context, instance.uuid, bdms,
raise_exc=False)
# if a delete task succeeded, always update vm state and task
# state without expecting task state to be DELETING
instance.vm_state = vm_states.DELETED
instance.task_state = None
instance.power_state = power_state.NOSTATE
instance.terminated_at = timeutils.utcnow()
instance.save()
self._update_resource_tracker(context, instance)
system_meta = instance.system_metadata
instance.destroy()
except Exception:
with excutils.save_and_reraise_exception():
quotas.rollback()
self._complete_deletion(context,
instance,
bdms,
quotas,
system_meta)
@wrap_exception()
@reverts_task_state
@wrap_instance_event
@wrap_instance_fault
def terminate_instance(self, context, instance, bdms, reservations):
"""Terminate an instance on this host."""
quotas = objects.Quotas.from_reservations(context,
reservations,
instance=instance)
@utils.synchronized(instance.uuid)
def do_terminate_instance(instance, bdms):
try:
self._delete_instance(context, instance, bdms, quotas)
except exception.InstanceNotFound:
LOG.info(_LI("Instance disappeared during terminate"),
instance=instance)
except Exception:
# As we're trying to delete always go to Error if something
# goes wrong that _delete_instance can't handle.
with excutils.save_and_reraise_exception():
LOG.exception(_LE('Setting instance vm_state to ERROR'),
instance=instance)
self._set_instance_obj_error_state(context, instance)
do_terminate_instance(instance, bdms)
# NOTE(johannes): This is probably better named power_off_instance
# so it matches the driver method, but because of other issues, we
# can't use that name in grizzly.
@wrap_exception()
@reverts_task_state
@wrap_instance_event
@wrap_instance_fault
def stop_instance(self, context, instance, clean_shutdown):
"""Stopping an instance on this host."""
@utils.synchronized(instance.uuid)
def do_stop_instance():
current_power_state = self._get_power_state(context, instance)
LOG.debug('Stopping instance; current vm_state: %(vm_state)s, '
'current task_state: %(task_state)s, current DB '
'power_state: %(db_power_state)s, current VM '
'power_state: %(current_power_state)s',
{'vm_state': instance.vm_state,
'task_state': instance.task_state,
'db_power_state': instance.power_state,
'current_power_state': current_power_state},
instance_uuid=instance.uuid)
# NOTE(mriedem): If the instance is already powered off, we are
# possibly tearing down and racing with other operations, so we can
# expect the task_state to be None if something else updates the
# instance and we're not locking it.
expected_task_state = [task_states.POWERING_OFF]
# The list of power states is from _sync_instance_power_state.
if current_power_state in (power_state.NOSTATE,
power_state.SHUTDOWN,
power_state.CRASHED):
LOG.info(_LI('Instance is already powered off in the '
'hypervisor when stop is called.'),
instance=instance)
expected_task_state.append(None)
self._notify_about_instance_usage(context, instance,
"power_off.start")
self._power_off_instance(context, instance, clean_shutdown)
instance.power_state = self._get_power_state(context, instance)
instance.vm_state = vm_states.STOPPED
instance.task_state = None
instance.save(expected_task_state=expected_task_state)
self._notify_about_instance_usage(context, instance,
"power_off.end")
do_stop_instance()
def _power_on(self, context, instance):
network_info = self.network_api.get_instance_nw_info(context, instance)
block_device_info = self._get_instance_block_device_info(context,
instance)
self.driver.power_on(context, instance,
network_info,
block_device_info)
def _delete_snapshot_of_shelved_instance(self, context, instance,
snapshot_id):
"""Delete snapshot of shelved instance."""
try:
self.image_api.delete(context, snapshot_id)
except (exception.ImageNotFound,
exception.ImageNotAuthorized) as exc:
LOG.warning(_LW("Failed to delete snapshot "
"from shelved instance (%s)."),
exc.format_message(), instance=instance)
except Exception:
LOG.exception(_LE("Something wrong happened when trying to "
"delete snapshot from shelved instance."),
instance=instance)
# NOTE(johannes): This is probably better named power_on_instance
# so it matches the driver method, but because of other issues, we
# can't use that name in grizzly.
@wrap_exception()
@reverts_task_state
@wrap_instance_event
@wrap_instance_fault
def start_instance(self, context, instance):
"""Starting an instance on this host."""
self._notify_about_instance_usage(context, instance, "power_on.start")
self._power_on(context, instance)
instance.power_state = self._get_power_state(context, instance)
instance.vm_state = vm_states.ACTIVE
instance.task_state = None
# Delete an image(VM snapshot) for a shelved instance
snapshot_id = instance.system_metadata.get('shelved_image_id')
if snapshot_id:
self._delete_snapshot_of_shelved_instance(context, instance,
snapshot_id)
# Delete system_metadata for a shelved instance
compute_utils.remove_shelved_keys_from_system_metadata(instance)
instance.save(expected_task_state=task_states.POWERING_ON)
self._notify_about_instance_usage(context, instance, "power_on.end")
@wrap_exception()
@reverts_task_state
@wrap_instance_event
@wrap_instance_fault
def soft_delete_instance(self, context, instance, reservations):