deb-cinder/cinder/volume/api.py

1767 lines
78 KiB
Python

# Copyright 2010 United States Government as represented by the
# Administrator of the National Aeronautics and Space Administration.
# 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 requests relating to volumes."""
import ast
import collections
import datetime
import functools
from oslo_config import cfg
from oslo_log import log as logging
from oslo_utils import excutils
from oslo_utils import strutils
from oslo_utils import timeutils
from oslo_utils import uuidutils
import six
from cinder.api import common
from cinder import context
from cinder import db
from cinder.db import base
from cinder import exception
from cinder import flow_utils
from cinder.i18n import _, _LE, _LI, _LW
from cinder.image import cache as image_cache
from cinder.image import glance
from cinder import keymgr
from cinder import objects
from cinder.objects import base as objects_base
from cinder.objects import fields
import cinder.policy
from cinder import quota
from cinder import quota_utils
from cinder.scheduler import rpcapi as scheduler_rpcapi
from cinder import utils
from cinder.volume.flows.api import create_volume
from cinder.volume.flows.api import manage_existing
from cinder.volume import qos_specs
from cinder.volume import rpcapi as volume_rpcapi
from cinder.volume import utils as volume_utils
from cinder.volume import volume_types
allow_force_upload_opt = cfg.BoolOpt('enable_force_upload',
default=False,
help='Enables the Force option on '
'upload_to_image. This enables '
'running upload_volume on in-use '
'volumes for backends that '
'support it.')
volume_host_opt = cfg.BoolOpt('snapshot_same_host',
default=True,
help='Create volume from snapshot at the host '
'where snapshot resides')
volume_same_az_opt = cfg.BoolOpt('cloned_volume_same_az',
default=True,
help='Ensure that the new volumes are the '
'same AZ as snapshot or source volume')
az_cache_time_opt = cfg.IntOpt('az_cache_duration',
default=3600,
help='Cache volume availability zones in '
'memory for the provided duration in '
'seconds')
CONF = cfg.CONF
CONF.register_opt(allow_force_upload_opt)
CONF.register_opt(volume_host_opt)
CONF.register_opt(volume_same_az_opt)
CONF.register_opt(az_cache_time_opt)
CONF.import_opt('glance_core_properties', 'cinder.image.glance')
LOG = logging.getLogger(__name__)
QUOTAS = quota.QUOTAS
def wrap_check_policy(func):
"""Check policy corresponding to the wrapped methods prior to execution
This decorator requires the first 3 args of the wrapped function
to be (self, context, volume)
"""
@functools.wraps(func)
def wrapped(self, context, target_obj, *args, **kwargs):
check_policy(context, func.__name__, target_obj)
return func(self, context, target_obj, *args, **kwargs)
return wrapped
def check_policy(context, action, target_obj=None):
target = {
'project_id': context.project_id,
'user_id': context.user_id,
}
if isinstance(target_obj, objects_base.CinderObject):
# Turn object into dict so target.update can work
target.update(
target_obj.obj_to_primitive()['versioned_object.data'] or {})
else:
target.update(target_obj or {})
_action = 'volume:%s' % action
cinder.policy.enforce(context, _action, target)
class API(base.Base):
"""API for interacting with the volume manager."""
AVAILABLE_MIGRATION_STATUS = (None, 'deleting', 'error', 'success')
def __init__(self, db_driver=None, image_service=None):
self.image_service = (image_service or
glance.get_default_image_service())
self.scheduler_rpcapi = scheduler_rpcapi.SchedulerAPI()
self.volume_rpcapi = volume_rpcapi.VolumeAPI()
self.availability_zones = []
self.availability_zones_last_fetched = None
self.key_manager = keymgr.API()
super(API, self).__init__(db_driver)
def list_availability_zones(self, enable_cache=False):
"""Describe the known availability zones
:retval tuple of dicts, each with a 'name' and 'available' key
"""
refresh_cache = False
if enable_cache:
if self.availability_zones_last_fetched is None:
refresh_cache = True
else:
cache_age = timeutils.delta_seconds(
self.availability_zones_last_fetched,
timeutils.utcnow())
if cache_age >= CONF.az_cache_duration:
refresh_cache = True
if refresh_cache or not enable_cache:
topic = CONF.volume_topic
ctxt = context.get_admin_context()
services = objects.ServiceList.get_all_by_topic(ctxt, topic)
az_data = [(s.availability_zone, s.disabled)
for s in services]
disabled_map = {}
for (az_name, disabled) in az_data:
tracked_disabled = disabled_map.get(az_name, True)
disabled_map[az_name] = tracked_disabled and disabled
azs = [{'name': name, 'available': not disabled}
for (name, disabled) in disabled_map.items()]
if refresh_cache:
now = timeutils.utcnow()
self.availability_zones = azs
self.availability_zones_last_fetched = now
LOG.debug("Availability zone cache updated, next update will"
" occur around %s.", now + datetime.timedelta(
seconds=CONF.az_cache_duration))
else:
azs = self.availability_zones
LOG.info(_LI("Availability Zones retrieved successfully."))
return tuple(azs)
def _retype_is_possible(self, context,
first_type_id, second_type_id,
first_type=None, second_type=None):
safe = False
elevated = context.elevated()
services = objects.ServiceList.get_all_by_topic(elevated,
'cinder-volume',
disabled=True)
if len(services.objects) == 1:
safe = True
else:
type_a = first_type or volume_types.get_volume_type(
elevated,
first_type_id)
type_b = second_type or volume_types.get_volume_type(
elevated,
second_type_id)
if (volume_utils.matching_backend_name(type_a['extra_specs'],
type_b['extra_specs'])):
safe = True
return safe
def _is_volume_migrating(self, volume):
# The migration status 'none' means no migration has ever been done
# before. The migration status 'error' means the previous migration
# failed. The migration status 'success' means the previous migration
# succeeded. The migration status 'deleting' means the source volume
# fails to delete after a migration.
# All of the statuses above means the volume is not in the process
# of a migration.
return (volume['migration_status'] not in
self.AVAILABLE_MIGRATION_STATUS)
def create(self, context, size, name, description, snapshot=None,
image_id=None, volume_type=None, metadata=None,
availability_zone=None, source_volume=None,
scheduler_hints=None,
source_replica=None, consistencygroup=None,
cgsnapshot=None, multiattach=False, source_cg=None):
check_policy(context, 'create')
# NOTE(jdg): we can have a create without size if we're
# doing a create from snap or volume. Currently
# the taskflow api will handle this and pull in the
# size from the source.
# NOTE(jdg): cinderclient sends in a string representation
# of the size value. BUT there is a possibility that somebody
# could call the API directly so the is_int_like check
# handles both cases (string representation of true float or int).
if size and (not strutils.is_int_like(size) or int(size) <= 0):
msg = _('Invalid volume size provided for create request: %s '
'(size argument must be an integer (or string '
'representation of an integer) and greater '
'than zero).') % size
raise exception.InvalidInput(reason=msg)
if consistencygroup and (not cgsnapshot and not source_cg):
if not volume_type:
msg = _("volume_type must be provided when creating "
"a volume in a consistency group.")
raise exception.InvalidInput(reason=msg)
cg_voltypeids = consistencygroup.get('volume_type_id')
if volume_type.get('id') not in cg_voltypeids:
msg = _("Invalid volume_type provided: %s (requested "
"type must be supported by this consistency "
"group).") % volume_type
raise exception.InvalidInput(reason=msg)
if volume_type and 'extra_specs' not in volume_type:
extra_specs = volume_types.get_volume_type_extra_specs(
volume_type['id'])
volume_type['extra_specs'] = extra_specs
if source_volume and volume_type:
if volume_type['id'] != source_volume['volume_type_id']:
if not self._retype_is_possible(
context,
volume_type['id'],
source_volume['volume_type_id'],
volume_type):
msg = _("Invalid volume_type provided: %s (requested type "
"is not compatible; either match source volume, "
"or omit type argument).") % volume_type['id']
raise exception.InvalidInput(reason=msg)
# When cloning replica (for testing), volume type must be omitted
if source_replica and volume_type:
msg = _("No volume_type should be provided when creating test "
"replica.")
raise exception.InvalidInput(reason=msg)
if snapshot and volume_type:
if volume_type['id'] != snapshot.volume_type_id:
if not self._retype_is_possible(context,
volume_type['id'],
snapshot.volume_type_id,
volume_type):
msg = _("Invalid volume_type provided: %s (requested "
"type is not compatible; recommend omitting "
"the type argument).") % volume_type['id']
raise exception.InvalidInput(reason=msg)
# Determine the valid availability zones that the volume could be
# created in (a task in the flow will/can use this information to
# ensure that the availability zone requested is valid).
raw_zones = self.list_availability_zones(enable_cache=True)
availability_zones = set([az['name'] for az in raw_zones])
if CONF.storage_availability_zone:
availability_zones.add(CONF.storage_availability_zone)
create_what = {
'context': context,
'raw_size': size,
'name': name,
'description': description,
'snapshot': snapshot,
'image_id': image_id,
'raw_volume_type': volume_type,
'metadata': metadata or {},
'raw_availability_zone': availability_zone,
'source_volume': source_volume,
'scheduler_hints': scheduler_hints,
'key_manager': self.key_manager,
'source_replica': source_replica,
'optional_args': {'is_quota_committed': False},
'consistencygroup': consistencygroup,
'cgsnapshot': cgsnapshot,
'multiattach': multiattach,
}
try:
sched_rpcapi = (self.scheduler_rpcapi if (not cgsnapshot and
not source_cg) else None)
volume_rpcapi = (self.volume_rpcapi if (not cgsnapshot and
not source_cg) else None)
flow_engine = create_volume.get_flow(self.db,
self.image_service,
availability_zones,
create_what,
sched_rpcapi,
volume_rpcapi)
except Exception:
msg = _('Failed to create api volume flow.')
LOG.exception(msg)
raise exception.CinderException(msg)
# Attaching this listener will capture all of the notifications that
# taskflow sends out and redirect them to a more useful log for
# cinders debugging (or error reporting) usage.
with flow_utils.DynamicLogListener(flow_engine, logger=LOG):
flow_engine.run()
vref = flow_engine.storage.fetch('volume')
LOG.info(_LI("Volume created successfully."), resource=vref)
return vref
@wrap_check_policy
def delete(self, context, volume,
force=False,
unmanage_only=False,
cascade=False):
if context.is_admin and context.project_id != volume.project_id:
project_id = volume.project_id
else:
project_id = context.project_id
if not volume.host:
volume_utils.notify_about_volume_usage(context,
volume, "delete.start")
# NOTE(vish): scheduling failed, so delete it
# Note(zhiteng): update volume quota reservation
try:
reserve_opts = {'volumes': -1, 'gigabytes': -volume.size}
QUOTAS.add_volume_type_opts(context,
reserve_opts,
volume.volume_type_id)
reservations = QUOTAS.reserve(context,
project_id=project_id,
**reserve_opts)
except Exception:
reservations = None
LOG.exception(_LE("Failed to update quota while "
"deleting volume."))
volume.destroy()
if reservations:
QUOTAS.commit(context, reservations, project_id=project_id)
volume_utils.notify_about_volume_usage(context,
volume, "delete.end")
LOG.info(_LI("Delete volume request issued successfully."),
resource={'type': 'volume',
'id': volume.id})
return
# Build required conditions for conditional update
expected = {'attach_status': db.Not('attached'),
'migration_status': self.AVAILABLE_MIGRATION_STATUS,
'consistencygroup_id': None}
# If not force deleting we have status conditions
if not force:
expected['status'] = ('available', 'error', 'error_restoring',
'error_extending')
if cascade:
# Allow deletion if all snapshots are in an expected state
filters = [~db.volume_has_undeletable_snapshots_filter()]
else:
# Don't allow deletion of volume with snapshots
filters = [~db.volume_has_snapshots_filter()]
values = {'status': 'deleting', 'terminated_at': timeutils.utcnow()}
result = volume.conditional_update(values, expected, filters)
if not result:
status = utils.build_or_str(expected.get('status'),
_('status must be %s and'))
msg = _('Volume %s must not be migrating, attached, belong to a '
'consistency group or have snapshots.') % status
LOG.info(msg)
raise exception.InvalidVolume(reason=msg)
if cascade:
values = {'status': 'deleting'}
expected = {'status': ('available', 'error', 'deleting'),
'cgsnapshot_id': None}
snapshots = objects.snapshot.SnapshotList.get_all_for_volume(
context, volume.id)
for s in snapshots:
result = s.conditional_update(values, expected, filters)
if not result:
volume.update({'status': 'error_deleting'})
volume.save()
msg = _('Failed to update snapshot.')
raise exception.InvalidVolume(reason=msg)
cache = image_cache.ImageVolumeCache(self.db, self)
entry = cache.get_by_image_volume(context, volume.id)
if entry:
cache.evict(context, entry)
# If the volume is encrypted, delete its encryption key from the key
# manager. This operation makes volume deletion an irreversible process
# because the volume cannot be decrypted without its key.
encryption_key_id = volume.get('encryption_key_id', None)
if encryption_key_id is not None:
try:
self.key_manager.delete_key(context, encryption_key_id)
except Exception as e:
LOG.warning(_LW("Unable to delete encryption key for "
"volume: %s."), e.msg, resource=volume)
self.volume_rpcapi.delete_volume(context,
volume,
unmanage_only,
cascade)
LOG.info(_LI("Delete volume request issued successfully."),
resource=volume)
@wrap_check_policy
def update(self, context, volume, fields):
if volume['status'] == 'maintenance':
LOG.info(_LI("Unable to update volume, "
"because it is in maintenance."), resource=volume)
msg = _("The volume cannot be updated during maintenance.")
raise exception.InvalidVolume(reason=msg)
# NOTE(thangp): Update is called by various APIs, some of which are
# not yet using oslo_versionedobjects. We need to handle the case
# where volume is either a dict or an oslo_versionedobject.
if isinstance(volume, objects_base.CinderObject):
volume.update(fields)
volume.save()
LOG.info(_LI("Volume updated successfully."), resource=volume)
else:
vref = self.db.volume_update(context, volume['id'], fields)
LOG.info(_LI("Volume updated successfully."), resource=vref)
def get(self, context, volume_id, viewable_admin_meta=False):
volume = objects.Volume.get_by_id(context, volume_id)
if viewable_admin_meta:
ctxt = context.elevated()
admin_metadata = self.db.volume_admin_metadata_get(ctxt,
volume_id)
volume.admin_metadata = admin_metadata
volume.obj_reset_changes()
try:
check_policy(context, 'get', volume)
except exception.PolicyNotAuthorized:
# raise VolumeNotFound instead to make sure Cinder behaves
# as it used to
raise exception.VolumeNotFound(volume_id=volume_id)
LOG.info(_LI("Volume info retrieved successfully."), resource=volume)
return volume
def get_all(self, context, marker=None, limit=None, sort_keys=None,
sort_dirs=None, filters=None, viewable_admin_meta=False,
offset=None):
check_policy(context, 'get_all')
if filters is None:
filters = {}
allTenants = utils.get_bool_param('all_tenants', filters)
try:
if limit is not None:
limit = int(limit)
if limit < 0:
msg = _('limit param must be positive')
raise exception.InvalidInput(reason=msg)
except ValueError:
msg = _('limit param must be an integer')
raise exception.InvalidInput(reason=msg)
# Non-admin shouldn't see temporary target of a volume migration, add
# unique filter data to reflect that only volumes with a NULL
# 'migration_status' or a 'migration_status' that does not start with
# 'target:' should be returned (processed in db/sqlalchemy/api.py)
if not context.is_admin:
filters['no_migration_targets'] = True
if filters:
LOG.debug("Searching by: %s.", six.text_type(filters))
if context.is_admin and allTenants:
# Need to remove all_tenants to pass the filtering below.
del filters['all_tenants']
volumes = objects.VolumeList.get_all(context, marker, limit,
sort_keys=sort_keys,
sort_dirs=sort_dirs,
filters=filters,
offset=offset)
else:
if viewable_admin_meta:
context = context.elevated()
volumes = objects.VolumeList.get_all_by_project(
context, context.project_id, marker, limit,
sort_keys=sort_keys, sort_dirs=sort_dirs, filters=filters,
offset=offset)
LOG.info(_LI("Get all volumes completed successfully."))
return volumes
def get_snapshot(self, context, snapshot_id):
check_policy(context, 'get_snapshot')
snapshot = objects.Snapshot.get_by_id(context, snapshot_id)
# FIXME(jdg): The objects don't have the db name entries
# so build the resource tag manually for now.
LOG.info(_LI("Snapshot retrieved successfully."),
resource={'type': 'snapshot',
'id': snapshot.id})
return snapshot
def get_volume(self, context, volume_id):
check_policy(context, 'get_volume')
volume = objects.Volume.get_by_id(context, volume_id)
LOG.info(_LI("Volume retrieved successfully."), resource=volume)
return volume
def get_all_snapshots(self, context, search_opts=None, marker=None,
limit=None, sort_keys=None, sort_dirs=None,
offset=None):
check_policy(context, 'get_all_snapshots')
search_opts = search_opts or {}
if context.is_admin and 'all_tenants' in search_opts:
# Need to remove all_tenants to pass the filtering below.
del search_opts['all_tenants']
snapshots = objects.SnapshotList.get_all(
context, search_opts, marker, limit, sort_keys, sort_dirs,
offset)
else:
snapshots = objects.SnapshotList.get_all_by_project(
context, context.project_id, search_opts, marker, limit,
sort_keys, sort_dirs, offset)
LOG.info(_LI("Get all snapshots completed successfully."))
return snapshots
@wrap_check_policy
def reserve_volume(self, context, volume):
expected = {'multiattach': volume.multiattach,
'status': (('available', 'in-use') if volume.multiattach
else 'available')}
result = volume.conditional_update({'status': 'attaching'}, expected)
if not result:
expected_status = utils.build_or_str(expected['status'])
msg = _('Volume status must be %s to reserve.') % expected_status
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
LOG.info(_LI("Reserve volume completed successfully."),
resource=volume)
@wrap_check_policy
def unreserve_volume(self, context, volume):
expected = {'status': 'attaching'}
# Status change depends on whether it has attachments (in-use) or not
# (available)
value = {'status': db.Case([(db.volume_has_attachments_filter(),
'in-use')],
else_='available')}
volume.conditional_update(value, expected)
LOG.info(_LI("Unreserve volume completed successfully."),
resource=volume)
@wrap_check_policy
def begin_detaching(self, context, volume):
# If we are in the middle of a volume migration, we don't want the
# user to see that the volume is 'detaching'. Having
# 'migration_status' set will have the same effect internally.
expected = {'status': 'in-use',
'attach_status': 'attached',
'migration_status': self.AVAILABLE_MIGRATION_STATUS}
result = volume.conditional_update({'status': 'detaching'}, expected)
if not (result or self._is_volume_migrating(volume)):
msg = _("Unable to detach volume. Volume status must be 'in-use' "
"and attach_status must be 'attached' to detach.")
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
LOG.info(_LI("Begin detaching volume completed successfully."),
resource=volume)
@wrap_check_policy
def roll_detaching(self, context, volume):
volume.conditional_update({'status': 'in-use'},
{'status': 'detaching'})
LOG.info(_LI("Roll detaching of volume completed successfully."),
resource=volume)
@wrap_check_policy
def attach(self, context, volume, instance_uuid, host_name,
mountpoint, mode):
if volume['status'] == 'maintenance':
LOG.info(_LI('Unable to attach volume, '
'because it is in maintenance.'), resource=volume)
msg = _("The volume cannot be attached in maintenance mode.")
raise exception.InvalidVolume(reason=msg)
# We add readonly metadata if it doesn't already exist
readonly = self.update_volume_admin_metadata(context.elevated(),
volume,
{'readonly': 'False'},
update=False)['readonly']
if readonly == 'True' and mode != 'ro':
raise exception.InvalidVolumeAttachMode(mode=mode,
volume_id=volume['id'])
attach_results = self.volume_rpcapi.attach_volume(context,
volume,
instance_uuid,
host_name,
mountpoint,
mode)
LOG.info(_LI("Attach volume completed successfully."),
resource=volume)
return attach_results
@wrap_check_policy
def detach(self, context, volume, attachment_id):
if volume['status'] == 'maintenance':
LOG.info(_LI('Unable to detach volume, '
'because it is in maintenance.'), resource=volume)
msg = _("The volume cannot be detached in maintenance mode.")
raise exception.InvalidVolume(reason=msg)
detach_results = self.volume_rpcapi.detach_volume(context, volume,
attachment_id)
LOG.info(_LI("Detach volume completed successfully."),
resource=volume)
return detach_results
@wrap_check_policy
def initialize_connection(self, context, volume, connector):
if volume['status'] == 'maintenance':
LOG.info(_LI('Unable to initialize the connection for '
'volume, because it is in '
'maintenance.'), resource=volume)
msg = _("The volume connection cannot be initialized in "
"maintenance mode.")
raise exception.InvalidVolume(reason=msg)
init_results = self.volume_rpcapi.initialize_connection(context,
volume,
connector)
LOG.info(_LI("Initialize volume connection completed successfully."),
resource=volume)
return init_results
@wrap_check_policy
def terminate_connection(self, context, volume, connector, force=False):
self.volume_rpcapi.terminate_connection(context,
volume,
connector,
force)
LOG.info(_LI("Terminate volume connection completed successfully."),
resource=volume)
self.unreserve_volume(context, volume)
@wrap_check_policy
def accept_transfer(self, context, volume, new_user, new_project):
if volume['status'] == 'maintenance':
LOG.info(_LI('Unable to accept transfer for volume, '
'because it is in maintenance.'), resource=volume)
msg = _("The volume cannot accept transfer in maintenance mode.")
raise exception.InvalidVolume(reason=msg)
results = self.volume_rpcapi.accept_transfer(context,
volume,
new_user,
new_project)
LOG.info(_LI("Transfer volume completed successfully."),
resource=volume)
return results
def _create_snapshot(self, context,
volume, name, description,
force=False, metadata=None,
cgsnapshot_id=None):
snapshot = self.create_snapshot_in_db(
context, volume, name,
description, force, metadata, cgsnapshot_id)
self.volume_rpcapi.create_snapshot(context, volume, snapshot)
return snapshot
def create_snapshot_in_db(self, context,
volume, name, description,
force, metadata,
cgsnapshot_id):
check_policy(context, 'create_snapshot', volume)
if volume['status'] == 'maintenance':
LOG.info(_LI('Unable to create the snapshot for volume, '
'because it is in maintenance.'), resource=volume)
msg = _("The snapshot cannot be created when the volume is in "
"maintenance mode.")
raise exception.InvalidVolume(reason=msg)
if self._is_volume_migrating(volume):
# Volume is migrating, wait until done
msg = _("Snapshot cannot be created while volume is migrating.")
raise exception.InvalidVolume(reason=msg)
if volume['status'].startswith('replica_'):
# Can't snapshot secondary replica
msg = _("Snapshot of secondary replica is not allowed.")
raise exception.InvalidVolume(reason=msg)
if ((not force) and (volume['status'] != "available")):
msg = _("Volume %(vol_id)s status must be available, "
"but current status is: "
"%(vol_status)s.") % {'vol_id': volume['id'],
'vol_status': volume['status']}
raise exception.InvalidVolume(reason=msg)
try:
if CONF.no_snapshot_gb_quota:
reserve_opts = {'snapshots': 1}
else:
reserve_opts = {'snapshots': 1, 'gigabytes': volume['size']}
QUOTAS.add_volume_type_opts(context,
reserve_opts,
volume.get('volume_type_id'))
reservations = QUOTAS.reserve(context, **reserve_opts)
except exception.OverQuota as e:
overs = e.kwargs['overs']
usages = e.kwargs['usages']
quotas = e.kwargs['quotas']
def _consumed(name):
return (usages[name]['reserved'] + usages[name]['in_use'])
for over in overs:
if 'gigabytes' in over:
msg = _LW("Quota exceeded for %(s_pid)s, tried to create "
"%(s_size)sG snapshot (%(d_consumed)dG of "
"%(d_quota)dG already consumed).")
LOG.warning(msg, {'s_pid': context.project_id,
's_size': volume['size'],
'd_consumed': _consumed(over),
'd_quota': quotas[over]})
raise exception.VolumeSizeExceedsAvailableQuota(
requested=volume['size'],
consumed=_consumed('gigabytes'),
quota=quotas['gigabytes'])
elif 'snapshots' in over:
msg = _LW("Quota exceeded for %(s_pid)s, tried to create "
"snapshot (%(d_consumed)d snapshots "
"already consumed).")
LOG.warning(msg, {'s_pid': context.project_id,
'd_consumed': _consumed(over)})
raise exception.SnapshotLimitExceeded(
allowed=quotas[over])
self._check_metadata_properties(metadata)
snapshot = None
try:
kwargs = {
'volume_id': volume['id'],
'cgsnapshot_id': cgsnapshot_id,
'user_id': context.user_id,
'project_id': context.project_id,
'status': 'creating',
'progress': '0%',
'volume_size': volume['size'],
'display_name': name,
'display_description': description,
'volume_type_id': volume['volume_type_id'],
'encryption_key_id': volume['encryption_key_id'],
'metadata': metadata or {}
}
snapshot = objects.Snapshot(context=context, **kwargs)
snapshot.create()
QUOTAS.commit(context, reservations)
except Exception:
with excutils.save_and_reraise_exception():
try:
if snapshot.obj_attr_is_set('id'):
snapshot.destroy()
finally:
QUOTAS.rollback(context, reservations)
return snapshot
def create_snapshots_in_db(self, context,
volume_list,
name, description,
force, cgsnapshot_id):
snapshot_list = []
for volume in volume_list:
self._create_snapshot_in_db_validate(context, volume, force)
reservations = self._create_snapshots_in_db_reserve(
context, volume_list)
options_list = []
for volume in volume_list:
options = self._create_snapshot_in_db_options(
context, volume, name, description, cgsnapshot_id)
options_list.append(options)
try:
for options in options_list:
snapshot = objects.Snapshot(context=context, **options)
snapshot.create()
snapshot_list.append(snapshot)
QUOTAS.commit(context, reservations)
except Exception:
with excutils.save_and_reraise_exception():
try:
for snap in snapshot_list:
snap.destroy()
finally:
QUOTAS.rollback(context, reservations)
return snapshot_list
def _create_snapshot_in_db_validate(self, context, volume, force):
check_policy(context, 'create_snapshot', volume)
if volume['status'] == 'maintenance':
LOG.info(_LI('Unable to create the snapshot for volume, '
'because it is in maintenance.'), resource=volume)
msg = _("The snapshot cannot be created when the volume is in "
"maintenance mode.")
raise exception.InvalidVolume(reason=msg)
if self._is_volume_migrating(volume):
# Volume is migrating, wait until done
msg = _("Snapshot cannot be created while volume is migrating.")
raise exception.InvalidVolume(reason=msg)
if ((not force) and (volume['status'] != "available")):
msg = _("Snapshot cannot be created because volume %(vol_id)s "
"is not available, current volume status: "
"%(vol_status)s.") % {'vol_id': volume['id'],
'vol_status': volume['status']}
raise exception.InvalidVolume(reason=msg)
def _create_snapshots_in_db_reserve(self, context, volume_list):
reserve_opts_list = []
total_reserve_opts = {}
try:
for volume in volume_list:
if CONF.no_snapshot_gb_quota:
reserve_opts = {'snapshots': 1}
else:
reserve_opts = {'snapshots': 1,
'gigabytes': volume['size']}
QUOTAS.add_volume_type_opts(context,
reserve_opts,
volume.get('volume_type_id'))
reserve_opts_list.append(reserve_opts)
for reserve_opts in reserve_opts_list:
for (key, value) in reserve_opts.items():
if key not in total_reserve_opts.keys():
total_reserve_opts[key] = value
else:
total_reserve_opts[key] = \
total_reserve_opts[key] + value
reservations = QUOTAS.reserve(context, **total_reserve_opts)
except exception.OverQuota as e:
overs = e.kwargs['overs']
usages = e.kwargs['usages']
quotas = e.kwargs['quotas']
volume_utils.process_reserve_over_quota(context, overs, usages,
quotas, volume['size'])
return reservations
def _create_snapshot_in_db_options(self, context, volume,
name, description,
cgsnapshot_id):
options = {'volume_id': volume['id'],
'cgsnapshot_id': cgsnapshot_id,
'user_id': context.user_id,
'project_id': context.project_id,
'status': "creating",
'progress': '0%',
'volume_size': volume['size'],
'display_name': name,
'display_description': description,
'volume_type_id': volume['volume_type_id'],
'encryption_key_id': volume['encryption_key_id']}
return options
def create_snapshot(self, context,
volume, name, description,
metadata=None, cgsnapshot_id=None):
result = self._create_snapshot(context, volume, name, description,
False, metadata, cgsnapshot_id)
LOG.info(_LI("Snapshot create request issued successfully."),
resource=result)
return result
def create_snapshot_force(self, context,
volume, name,
description, metadata=None):
result = self._create_snapshot(context, volume, name, description,
True, metadata)
LOG.info(_LI("Snapshot force create request issued successfully."),
resource=result)
return result
@wrap_check_policy
def delete_snapshot(self, context, snapshot, force=False,
unmanage_only=False):
# Build required conditions for conditional update
expected = {'cgsnapshot_id': None}
# If not force deleting we have status conditions
if not force:
expected['status'] = ('available', 'error')
result = snapshot.conditional_update({'status': 'deleting'}, expected)
if not result:
status = utils.build_or_str(expected.get('status'),
_('status must be %s and'))
msg = (_('Snapshot %s must not be part of a consistency group.') %
status)
LOG.error(msg)
raise exception.InvalidSnapshot(reason=msg)
# Make RPC call to the right host
volume = objects.Volume.get_by_id(context, snapshot.volume_id)
self.volume_rpcapi.delete_snapshot(context, snapshot, volume.host,
unmanage_only=unmanage_only)
LOG.info(_LI("Snapshot delete request issued successfully."),
resource=snapshot)
@wrap_check_policy
def update_snapshot(self, context, snapshot, fields):
snapshot.update(fields)
snapshot.save()
@wrap_check_policy
def get_volume_metadata(self, context, volume):
"""Get all metadata associated with a volume."""
rv = self.db.volume_metadata_get(context, volume['id'])
LOG.info(_LI("Get volume metadata completed successfully."),
resource=volume)
return dict(rv)
@wrap_check_policy
def delete_volume_metadata(self, context, volume,
key, meta_type=common.METADATA_TYPES.user):
"""Delete the given metadata item from a volume."""
if volume['status'] == 'maintenance':
LOG.info(_LI('Unable to delete the volume metadata, '
'because it is in maintenance.'), resource=volume)
msg = _("The volume metadata cannot be deleted when the volume "
"is in maintenance mode.")
raise exception.InvalidVolume(reason=msg)
self.db.volume_metadata_delete(context, volume['id'], key, meta_type)
LOG.info(_LI("Delete volume metadata completed successfully."),
resource=volume)
def _check_metadata_properties(self, metadata=None):
if not metadata:
metadata = {}
for k, v in metadata.items():
if len(k) == 0:
msg = _("Metadata property key blank.")
LOG.warning(msg)
raise exception.InvalidVolumeMetadata(reason=msg)
if len(k) > 255:
msg = _("Metadata property key greater than 255 characters.")
LOG.warning(msg)
raise exception.InvalidVolumeMetadataSize(reason=msg)
if len(v) > 255:
msg = _("Metadata property value greater than 255 characters.")
LOG.warning(msg)
raise exception.InvalidVolumeMetadataSize(reason=msg)
@wrap_check_policy
def update_volume_metadata(self, context, volume,
metadata, delete=False,
meta_type=common.METADATA_TYPES.user):
"""Updates or creates volume metadata.
If delete is True, metadata items that are not specified in the
`metadata` argument will be deleted.
"""
if volume['status'] == 'maintenance':
LOG.info(_LI('Unable to update the metadata for volume, '
'because it is in maintenance.'), resource=volume)
msg = _("The volume metadata cannot be updated when the volume "
"is in maintenance mode.")
raise exception.InvalidVolume(reason=msg)
self._check_metadata_properties(metadata)
db_meta = self.db.volume_metadata_update(context, volume['id'],
metadata,
delete,
meta_type)
# TODO(jdg): Implement an RPC call for drivers that may use this info
LOG.info(_LI("Update volume metadata completed successfully."),
resource=volume)
return db_meta
@wrap_check_policy
def get_volume_admin_metadata(self, context, volume):
"""Get all administration metadata associated with a volume."""
rv = self.db.volume_admin_metadata_get(context, volume['id'])
LOG.info(_LI("Get volume admin metadata completed successfully."),
resource=volume)
return dict(rv)
@wrap_check_policy
def update_volume_admin_metadata(self, context, volume, metadata,
delete=False, add=True, update=True):
"""Updates or creates volume administration metadata.
If delete is True, metadata items that are not specified in the
`metadata` argument will be deleted.
"""
self._check_metadata_properties(metadata)
db_meta = self.db.volume_admin_metadata_update(context, volume['id'],
metadata, delete, add,
update)
# TODO(jdg): Implement an RPC call for drivers that may use this info
LOG.info(_LI("Update volume admin metadata completed successfully."),
resource=volume)
return db_meta
@wrap_check_policy
def get_snapshot_metadata(self, context, snapshot):
"""Get all metadata associated with a snapshot."""
snapshot_obj = self.get_snapshot(context, snapshot.id)
LOG.info(_LI("Get snapshot metadata completed successfully."),
resource=snapshot)
return snapshot_obj.metadata
@wrap_check_policy
def delete_snapshot_metadata(self, context, snapshot, key):
"""Delete the given metadata item from a snapshot."""
snapshot_obj = self.get_snapshot(context, snapshot.id)
snapshot_obj.delete_metadata_key(context, key)
LOG.info(_LI("Delete snapshot metadata completed successfully."),
resource=snapshot)
@wrap_check_policy
def update_snapshot_metadata(self, context,
snapshot, metadata,
delete=False):
"""Updates or creates snapshot metadata.
If delete is True, metadata items that are not specified in the
`metadata` argument will be deleted.
"""
if delete:
_metadata = metadata
else:
orig_meta = snapshot.metadata
_metadata = orig_meta.copy()
_metadata.update(metadata)
self._check_metadata_properties(_metadata)
snapshot.metadata = _metadata
snapshot.save()
# TODO(jdg): Implement an RPC call for drivers that may use this info
LOG.info(_LI("Update snapshot metadata completed successfully."),
resource=snapshot)
return snapshot.metadata
def get_snapshot_metadata_value(self, snapshot, key):
LOG.info(_LI("Get snapshot metadata value not implemented."),
resource=snapshot)
# FIXME(jdg): Huh? Pass?
pass
def get_volumes_image_metadata(self, context):
check_policy(context, 'get_volumes_image_metadata')
db_data = self.db.volume_glance_metadata_get_all(context)
results = collections.defaultdict(dict)
for meta_entry in db_data:
results[meta_entry['volume_id']].update({meta_entry['key']:
meta_entry['value']})
return results
@wrap_check_policy
def get_volume_image_metadata(self, context, volume):
db_data = self.db.volume_glance_metadata_get(context, volume['id'])
LOG.info(_LI("Get volume image-metadata completed successfully."),
resource=volume)
return {meta_entry.key: meta_entry.value for meta_entry in db_data}
def get_list_volumes_image_metadata(self, context, volume_id_list):
db_data = self.db.volume_glance_metadata_list_get(context,
volume_id_list)
results = collections.defaultdict(dict)
for meta_entry in db_data:
results[meta_entry['volume_id']].update({meta_entry['key']:
meta_entry['value']})
return results
def _check_volume_availability(self, volume, force):
"""Check if the volume can be used."""
if volume['status'] not in ['available', 'in-use']:
msg = _('Volume %(vol_id)s status must be '
'available or in-use, but current status is: '
'%(vol_status)s.') % {'vol_id': volume['id'],
'vol_status': volume['status']}
raise exception.InvalidVolume(reason=msg)
if not force and 'in-use' == volume['status']:
msg = _('Volume status is in-use.')
raise exception.InvalidVolume(reason=msg)
@wrap_check_policy
def copy_volume_to_image(self, context, volume, metadata, force):
"""Create a new image from the specified volume."""
if not CONF.enable_force_upload and force:
LOG.info(_LI("Force upload to image is disabled, "
"Force option will be ignored."),
resource={'type': 'volume', 'id': volume['id']})
force = False
self._check_volume_availability(volume, force)
glance_core_properties = CONF.glance_core_properties
if glance_core_properties:
try:
volume_image_metadata = self.get_volume_image_metadata(context,
volume)
custom_property_set = (set(volume_image_metadata).difference
(set(glance_core_properties)))
if custom_property_set:
properties = {custom_property:
volume_image_metadata[custom_property]
for custom_property in custom_property_set}
metadata.update(dict(properties=properties))
except exception.GlanceMetadataNotFound:
# If volume is not created from image, No glance metadata
# would be available for that volume in
# volume glance metadata table
pass
recv_metadata = self.image_service.create(
context, self.image_service._translate_to_glance(metadata))
self.update(context, volume, {'status': 'uploading'})
self.volume_rpcapi.copy_volume_to_image(context,
volume,
recv_metadata)
response = {"id": volume['id'],
"updated_at": volume['updated_at'],
"status": 'uploading',
"display_description": volume['display_description'],
"size": volume['size'],
"volume_type": volume['volume_type'],
"image_id": recv_metadata['id'],
"container_format": recv_metadata['container_format'],
"disk_format": recv_metadata['disk_format'],
"image_name": recv_metadata.get('name', None)}
if 'protected' in recv_metadata:
response['protected'] = recv_metadata.get('protected')
if 'is_public' in recv_metadata:
response['is_public'] = recv_metadata.get('is_public')
elif 'visibility' in recv_metadata:
response['visibility'] = recv_metadata.get('visibility')
LOG.info(_LI("Copy volume to image completed successfully."),
resource=volume)
return response
@wrap_check_policy
def extend(self, context, volume, new_size):
if volume.status != 'available':
msg = _('Volume %(vol_id)s status must be available '
'to extend, but current status is: '
'%(vol_status)s.') % {'vol_id': volume.id,
'vol_status': volume.status}
raise exception.InvalidVolume(reason=msg)
size_increase = (int(new_size)) - volume.size
if size_increase <= 0:
msg = (_("New size for extend must be greater "
"than current size. (current: %(size)s, "
"extended: %(new_size)s).") % {'new_size': new_size,
'size': volume.size})
raise exception.InvalidInput(reason=msg)
try:
values = {'per_volume_gigabytes': new_size}
QUOTAS.limit_check(context, project_id=context.project_id,
**values)
except exception.OverQuota as e:
quotas = e.kwargs['quotas']
raise exception.VolumeSizeExceedsLimit(
size=new_size, limit=quotas['per_volume_gigabytes'])
try:
reserve_opts = {'gigabytes': size_increase}
QUOTAS.add_volume_type_opts(context, reserve_opts,
volume.volume_type_id)
reservations = QUOTAS.reserve(context,
project_id=volume.project_id,
**reserve_opts)
except exception.OverQuota as exc:
usages = exc.kwargs['usages']
quotas = exc.kwargs['quotas']
def _consumed(name):
return (usages[name]['reserved'] + usages[name]['in_use'])
msg = _LE("Quota exceeded for %(s_pid)s, tried to extend volume "
"by %(s_size)sG, (%(d_consumed)dG of %(d_quota)dG "
"already consumed).")
LOG.error(msg, {'s_pid': context.project_id,
's_size': size_increase,
'd_consumed': _consumed('gigabytes'),
'd_quota': quotas['gigabytes']})
raise exception.VolumeSizeExceedsAvailableQuota(
requested=size_increase,
consumed=_consumed('gigabytes'),
quota=quotas['gigabytes'])
self.update(context, volume, {'status': 'extending'})
self.volume_rpcapi.extend_volume(context, volume, new_size,
reservations)
LOG.info(_LI("Extend volume request issued successfully."),
resource=volume)
@wrap_check_policy
def migrate_volume(self, context, volume, host, force_host_copy,
lock_volume):
"""Migrate the volume to the specified host."""
if volume.status not in ['available', 'in-use']:
msg = _('Volume %(vol_id)s status must be available or in-use, '
'but current status is: '
'%(vol_status)s.') % {'vol_id': volume.id,
'vol_status': volume.status}
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
# Make sure volume is not part of a migration.
if self._is_volume_migrating(volume):
msg = _("Volume %s is already part of an active "
"migration.") % volume.id
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
# We only handle volumes without snapshots for now
snaps = objects.SnapshotList.get_all_for_volume(context, volume.id)
if snaps:
msg = _("Volume %s must not have snapshots.") % volume.id
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
# We only handle non-replicated volumes for now
if (volume.replication_status is not None and
volume.replication_status != 'disabled'):
msg = _("Volume %s must not be replicated.") % volume.id
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
if volume.consistencygroup_id:
msg = _("Volume %s must not be part of a consistency "
"group.") % volume.id
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
# Make sure the host is in the list of available hosts
elevated = context.elevated()
topic = CONF.volume_topic
services = objects.ServiceList.get_all_by_topic(
elevated, topic, disabled=False)
found = False
for service in services:
svc_host = volume_utils.extract_host(host, 'backend')
if utils.service_is_up(service) and service.host == svc_host:
found = True
if not found:
msg = _('No available service named %s') % host
LOG.error(msg)
raise exception.InvalidHost(reason=msg)
# Make sure the destination host is different than the current one
if host == volume.host:
msg = _('Destination host must be different '
'than the current host.')
LOG.error(msg)
raise exception.InvalidHost(reason=msg)
# When the migration of an available volume starts, both the status
# and the migration status of the volume will be changed.
# If the admin sets lock_volume flag to True, the volume
# status is changed to 'maintenance', telling users
# that this volume is in maintenance mode, and no action is allowed
# on this volume, e.g. attach, detach, retype, migrate, etc.
updates = {'migration_status': 'starting',
'previous_status': volume.status}
if lock_volume and volume.status == 'available':
updates['status'] = 'maintenance'
self.update(context, volume, updates)
# Call the scheduler to ensure that the host exists and that it can
# accept the volume
volume_type = {}
if volume.volume_type_id:
volume_type = volume_types.get_volume_type(context.elevated(),
volume.volume_type_id)
request_spec = {'volume_properties': volume,
'volume_type': volume_type,
'volume_id': volume.id}
self.scheduler_rpcapi.migrate_volume_to_host(context,
CONF.volume_topic,
volume.id,
host,
force_host_copy,
request_spec,
volume=volume)
LOG.info(_LI("Migrate volume request issued successfully."),
resource=volume)
@wrap_check_policy
def migrate_volume_completion(self, context, volume, new_volume, error):
# This is a volume swap initiated by Nova, not Cinder. Nova expects
# us to return the new_volume_id.
if not (volume.migration_status or new_volume.migration_status):
# When we're not migrating and haven't hit any errors, we issue
# volume attach and detach requests so the volumes don't end in
# 'attaching' and 'detaching' state
if not error:
attachments = volume.volume_attachment
for attachment in attachments:
self.detach(context, volume, attachment.id)
self.attach(context, new_volume,
attachment.instance_uuid,
attachment.attached_host,
attachment.mountpoint,
'rw')
return new_volume.id
if not volume.migration_status:
msg = _('Source volume not mid-migration.')
raise exception.InvalidVolume(reason=msg)
if not new_volume.migration_status:
msg = _('Destination volume not mid-migration.')
raise exception.InvalidVolume(reason=msg)
expected_status = 'target:%s' % volume.id
if not new_volume.migration_status == expected_status:
msg = (_('Destination has migration_status %(stat)s, expected '
'%(exp)s.') % {'stat': new_volume.migration_status,
'exp': expected_status})
raise exception.InvalidVolume(reason=msg)
LOG.info(_LI("Migrate volume completion issued successfully."),
resource=volume)
return self.volume_rpcapi.migrate_volume_completion(context, volume,
new_volume, error)
@wrap_check_policy
def update_readonly_flag(self, context, volume, flag):
if volume['status'] != 'available':
msg = _('Volume %(vol_id)s status must be available '
'to update readonly flag, but current status is: '
'%(vol_status)s.') % {'vol_id': volume['id'],
'vol_status': volume['status']}
raise exception.InvalidVolume(reason=msg)
self.update_volume_admin_metadata(context.elevated(), volume,
{'readonly': six.text_type(flag)})
LOG.info(_LI("Update readonly setting on volume "
"completed successfully."),
resource=volume)
@wrap_check_policy
def retype(self, context, volume, new_type, migration_policy=None):
"""Attempt to modify the type associated with an existing volume."""
if volume.status not in ['available', 'in-use']:
msg = _('Unable to update type due to incorrect status: '
'%(vol_status)s on volume: %(vol_id)s. Volume status '
'must be available or '
'in-use.') % {'vol_status': volume.status,
'vol_id': volume.id}
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
if self._is_volume_migrating(volume):
msg = (_("Volume %s is already part of an active migration.")
% volume.id)
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
if migration_policy and migration_policy not in ['on-demand', 'never']:
msg = _('migration_policy must be \'on-demand\' or \'never\', '
'passed: %s') % new_type
LOG.error(msg)
raise exception.InvalidInput(reason=msg)
if volume.consistencygroup_id:
msg = _("Volume must not be part of a consistency group.")
LOG.error(msg)
raise exception.InvalidVolume(reason=msg)
# Support specifying volume type by ID or name
try:
if uuidutils.is_uuid_like(new_type):
vol_type = volume_types.get_volume_type(context.elevated(),
new_type)
else:
vol_type = volume_types.get_volume_type_by_name(
context.elevated(), new_type)
except exception.InvalidVolumeType:
msg = _('Invalid volume_type passed: %s.') % new_type
LOG.error(msg)
raise exception.InvalidInput(reason=msg)
vol_type_id = vol_type['id']
vol_type_qos_id = vol_type['qos_specs_id']
old_vol_type = None
old_vol_type_id = volume.volume_type_id
old_vol_type_qos_id = None
# Error if the original and new type are the same
if volume.volume_type_id == vol_type_id:
msg = _('New volume_type same as original: %s.') % new_type
LOG.error(msg)
raise exception.InvalidInput(reason=msg)
if volume.volume_type_id:
old_vol_type = volume_types.get_volume_type(
context, old_vol_type_id)
old_vol_type_qos_id = old_vol_type['qos_specs_id']
# We don't support changing encryption requirements yet
old_enc = volume_types.get_volume_type_encryption(context,
old_vol_type_id)
new_enc = volume_types.get_volume_type_encryption(context,
vol_type_id)
if old_enc != new_enc:
msg = _('Retype cannot change encryption requirements.')
raise exception.InvalidInput(reason=msg)
# We don't support changing QoS at the front-end yet for in-use volumes
# TODO(avishay): Call Nova to change QoS setting (libvirt has support
# - virDomainSetBlockIoTune() - Nova does not have support yet).
if (volume.status != 'available' and
old_vol_type_qos_id != vol_type_qos_id):
for qos_id in [old_vol_type_qos_id, vol_type_qos_id]:
if qos_id:
specs = qos_specs.get_qos_specs(context.elevated(), qos_id)
if specs['consumer'] != 'back-end':
msg = _('Retype cannot change front-end qos specs for '
'in-use volume: %s.') % volume.id
raise exception.InvalidInput(reason=msg)
# We're checking here in so that we can report any quota issues as
# early as possible, but won't commit until we change the type. We
# pass the reservations onward in case we need to roll back.
reservations = quota_utils.get_volume_type_reservation(
context, volume, vol_type_id, reserve_vol_type_only=True)
# Get old reservations
try:
reserve_opts = {'volumes': -1, 'gigabytes': -volume.size}
QUOTAS.add_volume_type_opts(context,
reserve_opts,
old_vol_type_id)
# NOTE(wanghao): We don't need to reserve volumes and gigabytes
# quota for retyping operation since they didn't changed, just
# reserve volume_type and type gigabytes is fine.
reserve_opts.pop('volumes')
reserve_opts.pop('gigabytes')
old_reservations = QUOTAS.reserve(context,
project_id=volume.project_id,
**reserve_opts)
except Exception:
volume.status = volume.previous_status
volume.save()
msg = _("Failed to update quota usage while retyping volume.")
LOG.exception(msg, resource=volume)
raise exception.CinderException(msg)
self.update(context, volume, {'status': 'retyping',
'previous_status': volume.status})
request_spec = {'volume_properties': volume,
'volume_id': volume.id,
'volume_type': vol_type,
'migration_policy': migration_policy,
'quota_reservations': reservations,
'old_reservations': old_reservations}
self.scheduler_rpcapi.retype(context, CONF.volume_topic, volume.id,
request_spec=request_spec,
filter_properties={}, volume=volume)
LOG.info(_LI("Retype volume request issued successfully."),
resource=volume)
def manage_existing(self, context, host, ref, name=None, description=None,
volume_type=None, metadata=None,
availability_zone=None, bootable=False):
if volume_type and 'extra_specs' not in volume_type:
extra_specs = volume_types.get_volume_type_extra_specs(
volume_type['id'])
volume_type['extra_specs'] = extra_specs
if availability_zone is None:
elevated = context.elevated()
try:
svc_host = volume_utils.extract_host(host, 'backend')
service = objects.Service.get_by_args(
elevated, svc_host, 'cinder-volume')
except exception.ServiceNotFound:
with excutils.save_and_reraise_exception():
LOG.error(_LE('Unable to find service: %(service)s for '
'given host: %(host)s.'),
{'service': CONF.volume_topic, 'host': host})
if service.disabled:
LOG.error(_LE('Unable to manage_existing volume on a disabled '
'service.'))
raise exception.ServiceUnavailable()
availability_zone = service.get('availability_zone')
manage_what = {
'context': context,
'name': name,
'description': description,
'host': host,
'ref': ref,
'volume_type': volume_type,
'metadata': metadata,
'availability_zone': availability_zone,
'bootable': bootable,
}
try:
flow_engine = manage_existing.get_flow(self.scheduler_rpcapi,
self.db,
manage_what)
except Exception:
msg = _('Failed to manage api volume flow.')
LOG.exception(msg)
raise exception.CinderException(msg)
# Attaching this listener will capture all of the notifications that
# taskflow sends out and redirect them to a more useful log for
# cinder's debugging (or error reporting) usage.
with flow_utils.DynamicLogListener(flow_engine, logger=LOG):
flow_engine.run()
vol_ref = flow_engine.storage.fetch('volume')
LOG.info(_LI("Manage volume request issued successfully."),
resource=vol_ref)
return vol_ref
def manage_existing_snapshot(self, context, ref, volume,
name=None, description=None,
metadata=None):
host = volume_utils.extract_host(volume['host'])
try:
# NOTE(jdg): We don't use this, we just make sure it's valid
# and exists before sending off the call
service = objects.Service.get_by_args(
context.elevated(), host, 'cinder-volume')
except exception.ServiceNotFound:
with excutils.save_and_reraise_exception():
LOG.error(_LE('Unable to find service: %(service)s for '
'given host: %(host)s.'),
{'service': CONF.volume_topic, 'host': host})
if service.disabled:
LOG.error(_LE('Unable to manage_existing snapshot on a disabled '
'service.'))
raise exception.ServiceUnavailable()
snapshot_object = self.create_snapshot_in_db(context, volume, name,
description, False,
metadata, None)
self.volume_rpcapi.manage_existing_snapshot(context, snapshot_object,
ref, host)
return snapshot_object
# FIXME(jdg): Move these Cheesecake methods (freeze, thaw and failover)
# to a services API because that's what they are
def failover_host(self,
ctxt,
host,
secondary_id=None):
check_policy(ctxt, 'failover_host')
ctxt = ctxt if ctxt.is_admin else ctxt.elevated()
svc_host = volume_utils.extract_host(host, 'backend')
service = objects.Service.get_by_args(
ctxt, svc_host, 'cinder-volume')
expected = {'replication_status': [fields.ReplicationStatus.ENABLED,
fields.ReplicationStatus.FAILED_OVER]}
result = service.conditional_update(
{'replication_status': fields.ReplicationStatus.FAILING_OVER},
expected)
if not result:
expected_status = utils.build_or_str(
expected['replication_status'])
msg = (_('Host replication_status must be %s to failover.')
% expected_status)
LOG.error(msg)
raise exception.InvalidInput(reason=msg)
self.volume_rpcapi.failover_host(ctxt, host, secondary_id)
def freeze_host(self, ctxt, host):
check_policy(ctxt, 'freeze_host')
ctxt = ctxt if ctxt.is_admin else ctxt.elevated()
svc_host = volume_utils.extract_host(host, 'backend')
service = objects.Service.get_by_args(
ctxt, svc_host, 'cinder-volume')
expected = {'frozen': False}
result = service.conditional_update(
{'frozen': True}, expected)
if not result:
msg = _('Host is already Frozen.')
LOG.error(msg)
raise exception.InvalidInput(reason=msg)
# Should we set service status to disabled to keep
# scheduler calls from being sent? Just use existing
# `cinder service-disable reason=freeze`
self.volume_rpcapi.freeze_host(ctxt, host)
def thaw_host(self, ctxt, host):
check_policy(ctxt, 'thaw_host')
ctxt = ctxt if ctxt.is_admin else ctxt.elevated()
svc_host = volume_utils.extract_host(host, 'backend')
service = objects.Service.get_by_args(
ctxt, svc_host, 'cinder-volume')
expected = {'frozen': True}
result = service.conditional_update(
{'frozen': False}, expected)
if not result:
msg = _('Host is NOT Frozen.')
LOG.error(msg)
raise exception.InvalidInput(reason=msg)
if not self.volume_rpcapi.thaw_host(ctxt, host):
return "Backend reported error during thaw_host operation."
def check_volume_filters(self, filters, strict=False):
"""Sets the user filter value to accepted format"""
booleans = self.db.get_booleans_for_table('volume')
# To translate any true/false equivalent to True/False
# which is only acceptable format in database queries.
for key, val in filters.items():
try:
if key in booleans:
filters[key] = self._check_boolean_filter_value(
key, val, strict)
elif key == 'display_name':
# Use the raw value of display name as is for the filter
# without passing it through ast.literal_eval(). If the
# display name is a properly quoted string (e.g. '"foo"')
# then literal_eval() strips the quotes (i.e. 'foo'), so
# the filter becomes different from the user input.
continue
else:
filters[key] = ast.literal_eval(val)
except (ValueError, SyntaxError):
LOG.debug('Could not evaluate value %s, assuming string', val)
def _check_boolean_filter_value(self, key, val, strict=False):
"""Boolean filter values in Volume GET.
Before V3.2, all values other than 'False', 'false', 'FALSE' were
trated as True for specific boolean filter parameters in Volume
GET request.
But V3.2 onwards, only true/True/0/1/False/false parameters are
supported.
All other input values to specific boolean filter parameter will
lead to raising exception.
This changes API behavior. So, micro version introduced for V3.2
onwards.
"""
if strict:
# for updated behavior, from V3.2 onwards.
# To translate any true/false/t/f/0/1 to True/False
# which is only acceptable format in database queries.
try:
return strutils.bool_from_string(val, strict=True)
except ValueError:
msg = _('\'%(key)s = %(value)s\'') % {'key': key,
'value': val}
raise exception.InvalidInput(reason=msg)
else:
# For existing behavior(before version 3.2)
accepted_true = ['True', 'true', 'TRUE']
accepted_false = ['False', 'false', 'FALSE']
if val in accepted_false:
return False
elif val in accepted_true:
return True
else:
return bool(val)
class HostAPI(base.Base):
def __init__(self):
super(HostAPI, self).__init__()
"""Sub-set of the Volume Manager API for managing host operations."""
def set_host_enabled(self, context, host, enabled):
"""Sets the specified host's ability to accept new volumes."""
raise NotImplementedError()
def get_host_uptime(self, context, host):
"""Returns the result of calling "uptime" on the target host."""
raise NotImplementedError()
def host_power_action(self, context, host, action):
raise NotImplementedError()
def set_host_maintenance(self, context, host, mode):
"""Start/Stop host maintenance window.
On start, it triggers volume evacuation.
"""
raise NotImplementedError()