def find_retype_host(self, context, request_spec, filter_properties=None,
migration_policy='never'):
"""Find a host that can accept the volume with its new type."""
filter_properties = filter_properties or {}
current_host = request_spec['volume_properties']['host']
# The volume already exists on this host, and so we shouldn't check if
# it can accept the volume again in the CapacityFilter.
filter_properties['vol_exists_on'] = current_host
weighed_hosts = self._get_weighted_candidates(context, request_spec,
filter_properties)
if not weighed_hosts:
msg = (_('No valid hosts for volume %(id)s with type %(type)s')
% {'id': request_spec['volume_id'],
'type': request_spec['volume_type']})
raise exception.NoValidHost(reason=msg)
for weighed_host in weighed_hosts:
host_state = weighed_host.obj
if host_state.host == current_host:
return host_state
if migration_policy == 'never':
msg = (_('Current host not valid for volume %(id)s with type '
'%(type)s, migration not allowed')
% {'id': request_spec['volume_id'],
'type': request_spec['volume_type']})
raise exception.NoValidHost(reason=msg)
top_host = self._choose_top_host(weighed_hosts, request_spec)
return top_host.obj
def schedule_create_volume(self, context, volume_id, **_kwargs):
"""Picks a host that is up and has the fewest volumes."""
elevated = context.elevated()
volume_ref = db.volume_get(context, volume_id)
availability_zone = volume_ref.get('availability_zone')
zone, host = None, None
if availability_zone:
zone, _x, host = availability_zone.partition(':')
if host and context.is_admin:
service = db.service_get_by_args(elevated, host, 'cinder-volume')
if not utils.service_is_up(service):
raise exception.WillNotSchedule(host=host)
driver.cast_to_volume_host(context, host, 'create_volume',
volume_id=volume_id, **_kwargs)
return None
results = db.service_get_all_volume_sorted(elevated)
if zone:
results = [(service, gigs) for (service, gigs) in results
if service['availability_zone'] == zone]
for result in results:
(service, volume_gigabytes) = result
if volume_gigabytes + volume_ref['size'] > FLAGS.max_gigabytes:
msg = _("Not enough allocatable volume gigabytes remaining")
raise exception.NoValidHost(reason=msg)
if utils.service_is_up(service) and not service['disabled']:
driver.cast_to_volume_host(context, service['host'],
'create_volume', volume_id=volume_id, **_kwargs)
return None
msg = _("Is the appropriate service running?")
raise exception.NoValidHost(reason=msg)
def find_retype_host(self,
context,
request_spec,
filter_properties=None,
migration_policy='never'):
"""Find a host that can accept the volume with its new type."""
filter_properties = filter_properties or {}
current_host = request_spec['volume_properties']['host']
# The volume already exists on this host, and so we shouldn't check if
# it can accept the volume again in the CapacityFilter.
filter_properties['vol_exists_on'] = current_host
weighed_hosts = self._get_weighted_candidates(context, request_spec,
filter_properties)
if not weighed_hosts:
raise exception.NoValidHost(
reason=_('No valid hosts for volume '
'%(id)s with type %(type)s') % {
'id': request_spec['volume_id'],
'type': request_spec['volume_type']
})
for weighed_host in weighed_hosts:
host_state = weighed_host.obj
if host_state.host == current_host:
return host_state
if utils.extract_host(current_host, 'pool') is None:
# legacy volumes created before pool is introduced has no pool
# info in host. But host_state.host always include pool level
# info. In this case if above exact match didn't work out, we
# find host_state that are of the same host of volume being
# retyped. In other words, for legacy volumes, retyping could
# cause migration between pools on same host, which we consider
# it is different from migration between hosts thus allow that
# to happen even migration policy is 'never'.
for weighed_host in weighed_hosts:
host_state = weighed_host.obj
backend = utils.extract_host(host_state.host, 'backend')
if backend == current_host:
return host_state
if migration_policy == 'never':
raise exception.NoValidHost(
reason=_('Current host not valid for '
'volume %(id)s with type '
'%(type)s, migration not '
'allowed') % {
'id': request_spec['volume_id'],
'type': request_spec['volume_type']
})
top_host = self._choose_top_host(weighed_hosts, request_spec)
return top_host.obj
def schedule_create_volume(self, context, request_spec, filter_properties):
"""Picks a host that is up and has the fewest volumes."""
elevated = context.elevated()
volume_id = request_spec.get('volume_id')
snapshot_id = request_spec.get('snapshot_id')
image_id = request_spec.get('image_id')
volume_properties = request_spec.get('volume_properties')
volume_size = volume_properties.get('size')
availability_zone = volume_properties.get('availability_zone')
zone, host = None, None
if availability_zone:
zone, _x, host = availability_zone.partition(':')
if host and context.is_admin:
topic = CONF.volume_topic
service = db.service_get_by_args(elevated, host, topic)
if not utils.service_is_up(service):
raise exception.WillNotSchedule(host=host)
updated_volume = driver.volume_update_db(context, volume_id, host)
self.volume_rpcapi.create_volume(context,
updated_volume,
host,
request_spec,
filter_properties,
snapshot_id=snapshot_id,
image_id=image_id)
return None
results = db.service_get_all_volume_sorted(elevated)
if zone:
results = [(s, gigs) for (s, gigs) in results
if s['availability_zone'] == zone]
for result in results:
(service, volume_gigabytes) = result
if volume_gigabytes + volume_size > CONF.max_gigabytes:
msg = _("Not enough allocatable volume gigabytes remaining")
raise exception.NoValidHost(reason=msg)
if utils.service_is_up(service) and not service['disabled']:
updated_volume = driver.volume_update_db(
context, volume_id, service['host'])
self.volume_rpcapi.create_volume(context,
updated_volume,
service['host'],
request_spec,
filter_properties,
snapshot_id=snapshot_id,
image_id=image_id)
return None
msg = _("Is the appropriate service running?")
raise exception.NoValidHost(reason=msg)
def _schedule(self, context, topic, request_spec, **kwargs):
"""Picks a host that is up at random."""
elevated = context.elevated()
hosts = self.hosts_up(elevated, topic)
if not hosts:
msg = _("Is the appropriate service running?")
raise exception.NoValidHost(reason=msg)
hosts = self._filter_hosts(request_spec, hosts, **kwargs)
if not hosts:
msg = _("Could not find another host")
raise exception.NoValidHost(reason=msg)
return hosts[int(random.random() * len(hosts))]
def _populate_retry(self, filter_properties, properties):
"""Populate filter properties with history of retries for this
request. If maximum retries is exceeded, raise NoValidHost.
"""
max_attempts = self.max_attempts
retry = filter_properties.pop('retry', {})
if max_attempts == 1:
# re-scheduling is disabled.
return
# retry is enabled, update attempt count:
if retry:
retry['num_attempts'] += 1
else:
retry = {
'num_attempts': 1,
'hosts': [] # list of volume service hosts tried
}
filter_properties['retry'] = retry
volume_id = properties.get('volume_id')
self._log_volume_error(volume_id, retry)
if retry['num_attempts'] > max_attempts:
raise exception.NoValidHost(
reason=_("Exceeded max scheduling attempts %(max_attempts)d "
"for volume %(volume_id)s") %
{'max_attempts': max_attempts,
'volume_id': volume_id})
def test_create_consistencygroup_exceptions(self):
with mock.patch.object(filter_scheduler.FilterScheduler,
'schedule_create_consistencygroup') as mock_cg:
original_driver = self.manager.driver
self.manager.driver = filter_scheduler.FilterScheduler
LOG = logging.getLogger('cinder.scheduler.manager')
self.stubs.Set(LOG, 'error', mock.Mock())
self.stubs.Set(LOG, 'exception', mock.Mock())
self.stubs.Set(db, 'consistencygroup_update', mock.Mock())
ex = exception.CinderException('test')
mock_cg.side_effect = ex
group_id = '1'
self.assertRaises(exception.CinderException,
self.manager.create_consistencygroup,
self.context, 'volume', group_id)
self.assertTrue(LOG.exception.call_count > 0)
db.consistencygroup_update.assert_called_once_with(
self.context, group_id, {'status': 'error'})
mock_cg.reset_mock()
LOG.exception.reset_mock()
db.consistencygroup_update.reset_mock()
mock_cg.side_effect = exception.NoValidHost(
reason="No weighed hosts available")
self.manager.create_consistencygroup(self.context, 'volume',
group_id)
self.assertTrue(LOG.error.call_count > 0)
db.consistencygroup_update.assert_called_once_with(
self.context, group_id, {'status': 'error'})
self.manager.driver = original_driver
def _test_migrate_volume_exception_returns_volume_state(
self, _mock_volume_update, _mock_host_passes, _mock_volume_get,
status, fake_updates):
volume = tests_utils.create_volume(self.context,
status=status,
previous_status='available')
fake_volume_id = volume.id
topic = 'fake_topic'
request_spec = {'volume_id': fake_volume_id}
_mock_host_passes.side_effect = exception.NoValidHost(reason="")
_mock_volume_get.return_value = volume
self.manager.migrate_volume_to_host(self.context,
topic,
fake_volume_id,
'host',
True,
request_spec=request_spec,
filter_properties={},
volume=volume)
_mock_volume_update.assert_called_once_with(self.context,
fake_volume_id,
fake_updates)
_mock_host_passes.assert_called_once_with(self.context, 'host',
request_spec, {})
def test_retype_volume_exception_returns_volume_state(self, _mock_vol_get,
_mock_vol_update):
# Test NoValidHost exception behavior for retype.
# Puts the volume in original state and eats the exception.
fake_volume_id = 1
topic = 'fake_topic'
volume_id = fake_volume_id
request_spec = {'volume_id': fake_volume_id, 'volume_type': {'id': 3},
'migration_policy': 'on-demand'}
vol_info = {'id': fake_volume_id, 'status': 'in-use',
'instance_uuid': 'foo', 'attached_host': None}
_mock_vol_get.return_value = vol_info
_mock_vol_update.return_value = {'status': 'in-use'}
_mock_find_retype_host = mock.Mock(
side_effect=exception.NoValidHost(reason=""))
orig_retype = self.manager.driver.find_retype_host
self.manager.driver.find_retype_host = _mock_find_retype_host
self.manager.retype(self.context, topic, volume_id,
request_spec=request_spec,
filter_properties={})
_mock_vol_get.assert_called_once_with(self.context, fake_volume_id)
_mock_find_retype_host.assert_called_once_with(self.context,
request_spec, {},
'on-demand')
_mock_vol_update.assert_called_once_with(self.context, fake_volume_id,
{'status': 'in-use'})
self.manager.driver.find_retype_host = orig_retype
def test_create_volume_exception_puts_volume_in_error_state(
self, _mock_volume_update, _mock_message_create,
_mock_sched_create):
# Test NoValidHost exception behavior for create_volume.
# Puts the volume in 'error' state and eats the exception.
_mock_sched_create.side_effect = exception.NoValidHost(reason="")
volume = fake_volume.fake_volume_obj(self.context)
topic = 'fake_topic'
request_spec = {'volume_id': volume.id,
'volume': {'id': volume.id, '_name_id': None,
'metadata': {}, 'admin_metadata': {},
'glance_metadata': {}}}
request_spec_obj = objects.RequestSpec.from_primitives(request_spec)
self.manager.create_volume(self.context, topic, volume.id,
request_spec=request_spec,
filter_properties={},
volume=volume)
_mock_volume_update.assert_called_once_with(self.context,
volume.id,
{'status': 'error'})
_mock_sched_create.assert_called_once_with(self.context,
request_spec_obj, {})
_mock_message_create.assert_called_once_with(
self.context, defined_messages.UNABLE_TO_ALLOCATE,
self.context.project_id, resource_type='VOLUME',
resource_uuid=volume.id)
def test_migrate_volume_exception_returns_volume_state(self):
"""Test NoValidHost exception behavior for migrate_volume_to_host.
Puts the volume in 'error_migrating' state and eats the exception.
"""
fake_volume_id = 1
self._mox_schedule_method_helper('host_passes_filters')
self.mox.StubOutWithMock(db, 'volume_update')
topic = 'fake_topic'
volume_id = fake_volume_id
request_spec = {'volume_id': fake_volume_id}
self.manager.driver.host_passes_filters(
self.context, 'host', request_spec,
{}).AndRaise(exception.NoValidHost(reason=""))
db.volume_update(self.context, fake_volume_id,
{'migration_status': None})
self.mox.ReplayAll()
self.manager.migrate_volume_to_host(self.context,
topic,
volume_id,
'host',
True,
request_spec=request_spec,
filter_properties={})
def test_retype_volume_exception_returns_volume_state(
self, _mock_vol_attachment_get, _mock_vol_update):
# Test NoValidHost exception behavior for retype.
# Puts the volume in original state and eats the exception.
volume = tests_utils.create_volume(self.context,
status='retyping',
previous_status='in-use')
instance_uuid = '12345678-1234-5678-1234-567812345678'
volume_attach = tests_utils.attach_volume(self.context, volume.id,
instance_uuid, None,
'/dev/fake')
_mock_vol_attachment_get.return_value = [volume_attach]
topic = 'fake_topic'
request_spec = {'volume_id': volume.id, 'volume_type': {'id': 3},
'migration_policy': 'on-demand'}
_mock_vol_update.return_value = {'status': 'in-use'}
_mock_find_retype_host = mock.Mock(
side_effect=exception.NoValidHost(reason=""))
orig_retype = self.manager.driver.find_retype_host
self.manager.driver.find_retype_host = _mock_find_retype_host
self.manager.retype(self.context, topic, volume.id,
request_spec=request_spec,
filter_properties={},
volume=volume)
_mock_find_retype_host.assert_called_once_with(self.context,
request_spec, {},
'on-demand')
_mock_vol_update.assert_called_once_with(self.context, volume.id,
{'status': 'in-use'})
self.manager.driver.find_retype_host = orig_retype
def _check_mode_get_or_register_storage_system(self):
"""Does validity checks for storage system registry and health."""
def _resolve_host(host):
try:
ip = utils.resolve_hostname(host)
return ip
except socket.gaierror as e:
LOG.error(
_('Error resolving host %(host)s. Error - %(e)s.') % {
'host': host,
'e': e
})
return None
ips = self.configuration.netapp_controller_ips
ips = [i.strip() for i in ips.split(",")]
ips = [x for x in ips if _resolve_host(x)]
host = utils.resolve_hostname(
self.configuration.netapp_server_hostname)
if not ips:
msg = _('Controller ips not valid after resolution.')
raise exception.NoValidHost(reason=msg)
if host in ips:
LOG.info(_('Embedded mode detected.'))
system = self._client.list_storage_systems()[0]
else:
LOG.info(_('Proxy mode detected.'))
system = self._client.register_storage_system(
ips, password=self.configuration.netapp_sa_password)
self._client.set_system_id(system.get('id'))
def test_create_consistencygroup_exceptions(self):
with mock.patch.object(filter_scheduler.FilterScheduler,
'schedule_create_consistencygroup') as mock_cg:
original_driver = self.manager.driver
consistencygroup_obj = \
fake_consistencygroup.fake_consistencyobject_obj(self.context)
self.manager.driver = filter_scheduler.FilterScheduler
LOG = self.mock_object(manager, 'LOG')
self.stubs.Set(db, 'consistencygroup_update', mock.Mock())
ex = exception.CinderException('test')
mock_cg.side_effect = ex
group_id = fake.consistency_group_id
self.assertRaises(exception.CinderException,
self.manager.create_consistencygroup,
self.context, 'volume', consistencygroup_obj)
self.assertTrue(LOG.exception.call_count > 0)
db.consistencygroup_update.assert_called_once_with(
self.context, group_id,
{'status': (fields.ConsistencyGroupStatus.ERROR)})
mock_cg.reset_mock()
LOG.exception.reset_mock()
db.consistencygroup_update.reset_mock()
mock_cg.side_effect = exception.NoValidHost(
reason="No weighed hosts available")
self.manager.create_consistencygroup(self.context, 'volume',
consistencygroup_obj)
self.assertTrue(LOG.error.call_count > 0)
db.consistencygroup_update.assert_called_once_with(
self.context, group_id,
{'status': (fields.ConsistencyGroupStatus.ERROR)})
self.manager.driver = original_driver
def schedule_create_volume(self, context, request_spec, filter_properties):
weighed_host = self._schedule(context, request_spec,
filter_properties)
if not weighed_host:
raise exception.NoValidHost(reason="")
host = weighed_host.obj.host
volume_id = request_spec['volume_id']
snapshot_id = request_spec['snapshot_id']
image_id = request_spec['image_id']
updated_volume = driver.volume_update_db(context, volume_id, host)
self._post_select_populate_filter_properties(filter_properties,
weighed_host.obj)
# context is not serializable
filter_properties.pop('context', None)
self.volume_rpcapi.create_volume(context, updated_volume, host,
request_spec=request_spec,
filter_properties=filter_properties,
allow_reschedule=True,
snapshot_id=snapshot_id,
image_id=image_id)
def _schedule(self, context, topic, request_spec, **kwargs):
"""Picks a host that is up at random."""
hosts = self._get_weighted_candidates(context, topic, request_spec,
**kwargs)
if not hosts:
msg = _("Could not find another host")
raise exception.NoValidHost(reason=msg)
return hosts[int(random.random() * len(hosts))]
def _resolve_host(host):
try:
ip = na_utils.resolve_hostname(host)
return ip
except socket.gaierror as e:
LOG.error(_LE('Error resolving host %(host)s. Error - %(e)s.'),
{'host': host, 'e': e})
raise exception.NoValidHost(
_("Controller IP '%(host)s' could not be resolved: %(e)s.")
% {'host': host, 'e': e})
def _get_weighted_candidates(self, context, topic, request_spec, **kwargs):
"""Returns a list of the available hosts."""
elevated = context.elevated()
hosts = self.hosts_up(elevated, topic)
if not hosts:
msg = _("Is the appropriate service running?")
raise exception.NoValidHost(reason=msg)
return self._filter_hosts(request_spec, hosts, **kwargs)
def host_passes_filters(self, context, host, request_spec,
filter_properties):
"""Check if the specified host passes the filters."""
weighed_hosts = self._get_weighted_candidates(context, request_spec,
filter_properties)
for weighed_host in weighed_hosts:
host_state = weighed_host.obj
if host_state.host == host:
return host_state
msg = (_('cannot place volume %(id)s on %(host)s')
% {'id': request_spec['volume_id'], 'host': host})
raise exception.NoValidHost(reason=msg)
def host_passes_filters(self, context, host, request_spec,
filter_properties):
"""Check if the specified host passes the filters."""
weighed_hosts = self._get_weighted_candidates(context, request_spec,
filter_properties)
for weighed_host in weighed_hosts:
host_state = weighed_host.obj
if host_state.host == host:
return host_state
volume_id = request_spec.get('volume_id', '??volume_id missing??')
raise exception.NoValidHost(reason=_('Cannot place volume %(id)s on '
'%(host)s') %
{'id': volume_id,
'host': host})
def schedule_create_consistencygroup(self, context, group_id,
request_spec_list,
filter_properties_list):
weighed_host = self._schedule_group(context, request_spec_list,
filter_properties_list)
if not weighed_host:
raise exception.NoValidHost(reason="No weighed hosts available")
host = weighed_host.obj.host
updated_group = driver.group_update_db(context, group_id, host)
self.volume_rpcapi.create_consistencygroup(context, updated_group,
host)
def test_create_volume_exception_puts_volume_in_error_state(
self, _mock_volume_update, _mock_sched_create):
# Test NoValidHost exception behavior for create_volume.
# Puts the volume in 'error' state and eats the exception.
_mock_sched_create.side_effect = exception.NoValidHost(reason="")
fake_volume_id = 1
topic = 'fake_topic'
request_spec = {'volume_id': fake_volume_id}
self.manager.create_volume(self.context, topic, fake_volume_id,
request_spec=request_spec,
filter_properties={})
_mock_volume_update.assert_called_once_with(self.context,
fake_volume_id,
{'status': 'error'})
_mock_sched_create.assert_called_once_with(self.context, request_spec,
{})
def test_migrate_volume_exception_returns_volume_state(
self, _mock_volume_update, _mock_host_passes):
# Test NoValidHost exception behavior for migrate_volume_to_host.
# Puts the volume in 'error_migrating' state and eats the exception.
_mock_host_passes.side_effect = exception.NoValidHost(reason="")
fake_volume_id = 1
topic = 'fake_topic'
request_spec = {'volume_id': fake_volume_id}
self.manager.migrate_volume_to_host(self.context, topic,
fake_volume_id, 'host', True,
request_spec=request_spec,
filter_properties={})
_mock_volume_update.assert_called_once_with(self.context,
fake_volume_id,
{'migration_status': None})
_mock_host_passes.assert_called_once_with(self.context, 'host',
request_spec, {})
def schedule_create_volume(self, context, request_spec, filter_properties):
resource_type = request_spec.get('volume_type', None)
extra_specs = {}
if resource_type:
extra_specs = resource_type.get('extra_specs', {})
volume_id = request_spec.get('volume_id')
snapshot_id = request_spec.get('snapshot_id')
image_id = request_spec.get('image_id')
replica_vol = None
if 'True' in extra_specs.get('replication_enabled', ''):
weighed_host, weighed_host2 = self._get_replication_hosts(
context, request_spec, filter_properties)
else:
weighed_host = self._schedule(context, request_spec,
filter_properties)
weighed_host2 = None
if not weighed_host:
raise exception.NoValidHost(reason="")
if weighed_host2:
az = weighed_host2.obj.service['availability_zone']
replica_vol = {'status': 'secondary_replica',
'attach_status': 'detached',
'host': weighed_host2.obj.host,
'availability_zone': az}
host = weighed_host.obj.host
updated_volume = driver.volume_update_db(context, volume_id, host,
replica=replica_vol)
self._post_select_populate_filter_properties(filter_properties,
weighed_host.obj)
# context is not serializable
filter_properties.pop('context', None)
self.volume_rpcapi.create_volume(context, updated_volume, host,
request_spec, filter_properties,
allow_reschedule=True,
snapshot_id=snapshot_id,
image_id=image_id)
def _connector_to_hostname_prefix(self, connector):
"""Translate connector info to storage system host name.
Translate a host's name and IP to the prefix of its hostname on the
storage subsystem. We create a host name from the host and
IP address, replacing any invalid characters (at most 55 characters),
and adding a random 8-character suffix to avoid collisions. The total
length should be at most 63 characters.
"""
# Build cleanup translation tables for host names
invalid_ch_in_host = ''
for num in range(0, 128):
ch = six.text_type(chr(num))
if not ch.isalnum() and ch not in [' ', '.', '-', '_']:
invalid_ch_in_host = invalid_ch_in_host + ch
host_name = connector['host']
if isinstance(host_name, six.text_type):
unicode_host_name_filter = {
ord(six.text_type(char)): u'-'
for char in invalid_ch_in_host
}
host_name = host_name.translate(unicode_host_name_filter)
elif isinstance(host_name, str):
string_host_name_filter = string.maketrans(
invalid_ch_in_host, '-' * len(invalid_ch_in_host))
host_name = host_name.translate(string_host_name_filter)
else:
msg = _('_create_host: Can not translate host name. Host name '
'is not unicode or string.')
LOG.error(msg)
raise exception.NoValidHost(reason=msg)
host_name = six.text_type(host_name)
# FlashSystem family doesn't like hostname that starts with number.
if not re.match('^[A-Za-z]', host_name):
host_name = '_' + host_name
return host_name[:55]
def host_passes_filters(self, context, host, request_spec,
filter_properties):
"""Check if the specified host passes the filters."""
weighed_hosts = self._get_weighted_candidates(
context,
CONF.volume_topic,
request_spec,
filter_properties=filter_properties)
for weighed_host in weighed_hosts:
if weighed_host == host:
elevated = context.elevated()
host_states = self.host_manager.get_all_host_states(elevated)
for host_state in host_states:
if host_state.host == host:
return host_state
msg = (_('cannot place volume %(id)s on %(host)s') % {
'id': request_spec['volume_id'],
'host': host
})
raise exception.NoValidHost(reason=msg)
def test_create_volume_exception_puts_volume_in_error_state(self):
"""Test NoValidHost exception behavior for create_volume.
Puts the volume in 'error' state and eats the exception.
"""
fake_volume_id = 1
self._mox_schedule_method_helper('schedule_create_volume')
self.mox.StubOutWithMock(db, 'volume_update')
topic = 'fake_topic'
volume_id = fake_volume_id
request_spec = {'volume_id': fake_volume_id}
self.manager.driver.schedule_create_volume(
self.context, request_spec,
{}).AndRaise(exception.NoValidHost(reason=""))
db.volume_update(self.context, fake_volume_id, {'status': 'error'})
self.mox.ReplayAll()
self.manager.create_volume(self.context,
topic,
volume_id,
request_spec=request_spec,
filter_properties={})