def stop(self, graceful=False):
if self.server:
LOG.info(_LI("Stop the Notification server..."))
self.server.stop()
if graceful:
LOG.info(_LI("Notification server stopped successfully. Waiting for final message to be processed..."))
self.server.wait()
def _get_vol_byuuid(self, voluuid):
volkey = self.volumeroot + voluuid
result = self.client.read(volkey)
volval = json.loads(result.value)
LOG.info(_LI('Read key: %s from etcd, result is: %s'), volkey, volval)
return volval
def _get_vol_byuuid(self, voluuid):
volkey = self.volumeroot + voluuid
result = self.client.read(volkey)
volval = json.loads(result.value)
LOG.info(_LI('Read key: %s from etcd, result is: %s'), volkey, volval)
return volval
def start(self):
LOG.info(_LI("Start RPC server..."))
self.server = oslo_messaging.get_rpc_server(self.transport,
self.target,
self.endpoints)
self.server.start()
self.server.wait()
def _clear_fault(self, alarm_id, entity_instance_id=None):
# Only clear alarm if there is one already raised
if (alarm_id == fm_constants.FM_ALARM_ID_STORAGE_CEPH
and self.current_health_alarm):
LOG.info(_LI("Clearing health alarm"))
self.service.fm_api.clear_fault(
fm_constants.FM_ALARM_ID_STORAGE_CEPH,
self.service.entity_instance_id)
elif (alarm_id == fm_constants.FM_ALARM_ID_STORAGE_CEPH_FREE_SPACE
and entity_instance_id in self.current_quota_alarms):
LOG.info(
_LI("Clearing quota alarm with entity_instance_id %s") %
entity_instance_id)
self.service.fm_api.clear_fault(
fm_constants.FM_ALARM_ID_STORAGE_CEPH_FREE_SPACE,
entity_instance_id)
def _set_upgrade(self, upgrade):
state = upgrade.get('state')
from_version = upgrade.get('from_version')
if (state and state != constants.UPGRADE_COMPLETED
and from_version == constants.TITANIUM_SERVER_VERSION_16_10):
LOG.info(_LI("Surpress require_jewel_osds health warning"))
self.surpress_require_jewel_osds_warning = True
def osd_pool_create(ceph_api, pool_name, pg_num, pgp_num):
if pool_name.endswith("-cache"):
# ruleset 1: is the ruleset for the cache tier
# Name: cache_tier_ruleset
ruleset = 1
else:
# ruleset 0: is the default ruleset if no crushmap is loaded or
# the ruleset for the backing tier if loaded:
# Name: storage_tier_ruleset
ruleset = 0
response, body = ceph_api.osd_pool_create(pool_name,
pg_num,
pgp_num,
pool_type="replicated",
ruleset=ruleset,
body='json')
if response.ok:
LOG.info(
_LI("Created OSD pool: "
"pool_name={}, pg_num={}, pgp_num={}, "
"pool_type=replicated, ruleset={}").format(
pool_name, pg_num, pgp_num, ruleset))
else:
e = exception.CephPoolCreateFailure(name=pool_name,
reason=response.reason)
LOG.error(e)
raise e
# Explicitly assign the ruleset to the pool on creation since it is
# ignored in the create call
response, body = ceph_api.osd_set_pool_param(pool_name,
"crush_ruleset",
ruleset,
body='json')
if response.ok:
LOG.info(
_LI("Assigned crush ruleset to OS pool: "
"pool_name={}, ruleset={}").format(pool_name, ruleset))
else:
e = exception.CephPoolRulesetFailure(name=pool_name,
reason=response.reason)
LOG.error(e)
ceph_api.osd_pool_delete(pool_name,
pool_name,
sure='--yes-i-really-really-mean-it',
body='json')
raise e
def auto_heal(self, health):
if (health['health'] == constants.CEPH_HEALTH_WARN
and (constants.CEPH_HEALTH_WARN_REQUIRE_JEWEL_OSDS_NOT_SET
in health['detail'])):
try:
upgrade = self.service.get_software_upgrade_status()
except Exception as ex:
LOG.warn(
_LW("Getting software upgrade status failed "
"with: %s. Skip auto-heal attempt "
"(will retry on next ceph status poll).") % str(ex))
return
state = upgrade.get('state')
# surpress require_jewel_osds in case upgrade is
# in progress but not completed or aborting
if (not self.surpress_require_jewel_osds_warning
and (upgrade.get('from_version')
== constants.TITANIUM_SERVER_VERSION_16_10)
and state not in [
None, constants.UPGRADE_COMPLETED,
constants.UPGRADE_ABORTING,
constants.UPGRADE_ABORT_COMPLETING,
constants.UPGRADE_ABORTING_ROLLBACK
]):
LOG.info(_LI("Surpress require_jewel_osds health warning"))
self.surpress_require_jewel_osds_warning = True
# set require_jewel_osds in case upgrade is
# not in progress or completed
if (state in [None, constants.UPGRADE_COMPLETED]):
LOG.warn(
_LW("No upgrade in progress or update completed "
"and require_jewel_osds health warning raised. "
"Set require_jewel_osds flag."))
self.set_flag_require_jewel_osds()
health = self._remove_require_jewel_osds_warning(health)
LOG.info(_LI("Unsurpress require_jewel_osds health warning"))
self.surpress_require_jewel_osds_warning = False
# unsurpress require_jewel_osds in case upgrade
# is aborting
if (self.surpress_require_jewel_osds_warning and state in [
constants.UPGRADE_ABORTING,
constants.UPGRADE_ABORT_COMPLETING,
constants.UPGRADE_ABORTING_ROLLBACK
]):
LOG.info(_LI("Unsurpress require_jewel_osds health warning"))
self.surpress_require_jewel_osds_warning = False
return health
def get_tiers_size(self, _):
"""Get the ceph cluster tier sizes.
returns: a dict of sizes (in GB) by tier name
"""
tiers_size = self.service.monitor.tiers_size
LOG.debug(_LI("Ceph cluster tiers (size in GB): %s") % str(tiers_size))
return tiers_size
def osd_pool_set_quota(ceph_api, pool_name, max_bytes=0, max_objects=0):
"""Set the quota for an OSD pool_name
Setting max_bytes or max_objects to 0 will disable that quota param
:param pool_name: OSD pool_name
:param max_bytes: maximum bytes for OSD pool_name
:param max_objects: maximum objects for OSD pool_name
"""
# Update quota if needed
prev_quota = osd_pool_get_quota(ceph_api, pool_name)
if prev_quota["max_bytes"] != max_bytes:
resp, b = ceph_api.osd_set_pool_quota(pool_name,
'max_bytes',
max_bytes,
body='json')
if resp.ok:
LOG.info(
_LI("Set OSD pool_name quota: "
"pool_name={}, max_bytes={}").format(pool_name, max_bytes))
else:
e = exception.CephPoolSetQuotaFailure(pool=pool_name,
name='max_bytes',
value=max_bytes,
reason=resp.reason)
LOG.error(e)
raise e
if prev_quota["max_objects"] != max_objects:
resp, b = ceph_api.osd_set_pool_quota(pool_name,
'max_objects',
max_objects,
body='json')
if resp.ok:
LOG.info(
_LI("Set OSD pool_name quota: "
"pool_name={}, max_objects={}").format(
pool_name, max_objects))
else:
e = exception.CephPoolSetQuotaFailure(pool=pool_name,
name='max_objects',
value=max_objects,
reason=resp.reason)
LOG.error(e)
raise e
def _set_upgrade(self, upgrade):
state = upgrade.get('state')
from_version = upgrade.get('from_version')
if (state and state != constants.UPGRADE_COMPLETED
and from_version == constants.TITANIUM_SERVER_VERSION_18_03):
LOG.info(
_LI("Wait for ceph upgrade to complete before monitoring cluster."
))
self.wait_for_upgrade_complete = True
def update_vol(self, volid, key, val):
volkey = self.volumeroot + volid
result = self.client.read(volkey)
volval = json.loads(result.value)
volval[key] = val
volval = json.dumps(volval)
result.value = volval
self.client.update(result)
LOG.info(_LI('Update key: %s to etcd, value is: %s'), volkey, volval)
def update_vol(self, volid, key, val):
volkey = self.volumeroot + volid
result = self.client.read(volkey)
volval = json.loads(result.value)
volval[key] = val
volval = json.dumps(volval)
result.value = volval
self.client.update(result)
LOG.info(_LI('Update key: %s to etcd, value is: %s'), volkey, volval)
def cache_tiering_enable_cache(self, _, new_config, applied_config):
LOG.info(_LI("Enabling cache"))
try:
self.service.cache_tiering.enable_cache(new_config, applied_config)
except exception.CephManagerException as e:
self.service.sysinv_conductor.call(
{},
'cache_tiering_enable_cache_complete',
success=False,
exception=str(e.message),
new_config=new_config,
applied_config=applied_config)
def get_primary_tier_size(self, _):
"""Get the ceph size for the primary tier.
returns: an int for the size (in GB) of the tier
"""
tiers_size = self.service.monitor.tiers_size
primary_tier_size = tiers_size.get(
self.service.monitor.primary_tier_name, 0)
LOG.debug(_LI("Ceph cluster primary tier size: %s GB") %
str(primary_tier_size))
return primary_tier_size
def ceph_poll_status(self):
# get previous data every time in case:
# * daemon restarted
# * alarm was cleared manually but stored as raised in daemon
self._get_current_alarms()
if self.current_health_alarm:
LOG.info(
_LI("Current alarm: %s") %
str(self.current_health_alarm.__dict__))
# get ceph health
health = self._get_health()
LOG.info(
_LI("Current Ceph health: "
"%(health)s detail: %(detail)s") % health)
health = self.filter_health_status(health)
if health['health'] != constants.CEPH_HEALTH_OK:
self._report_fault(health, fm_constants.FM_ALARM_ID_STORAGE_CEPH)
self._report_alarm_osds_health()
else:
self._clear_fault(fm_constants.FM_ALARM_ID_STORAGE_CEPH)
self.clear_all_major_critical()
def get_vol_byname(self, volname):
volumes = self.client.read(self.volumeroot, recursive=True)
LOG.info(_LI('Get volbyname: volname is %s'), volname)
for child in volumes.children:
if child.key != VOLUMEROOT:
volmember = json.loads(child.value)
vol = volmember['display_name']
if vol.startswith(volname, 0, len(volname)):
if volmember['display_name'] == volname:
return volmember
elif volmember['name'] == volname:
return volmember
return None
def get_vol_byname(self, volname):
volumes = self.client.read(self.volumeroot, recursive=True)
LOG.info(_LI('Get volbyname: volname is %s'), volname)
for child in volumes.children:
if child.key != VOLUMEROOT:
volmember = json.loads(child.value)
vol = volmember['display_name']
if vol.startswith(volname, 0, len(volname)):
if volmember['display_name'] == volname:
return volmember
elif volmember['name'] == volname:
return volmember
return None
def __init__(self, reactor, hpepluginconfig):
"""
:param IReactorTime reactor: Reactor time interface implementation.
:param Ihpepluginconfig : hpedefaultconfig configuration
"""
LOG.info(_LI('Initialize Volume Plugin'))
self._reactor = reactor
self._hpepluginconfig = hpepluginconfig
hpeplugin_driver = hpepluginconfig.hpedockerplugin_driver
self.hpeplugin_driver = \
importutils.import_object(hpeplugin_driver, self._hpepluginconfig)
if self.hpeplugin_driver is None:
msg = (_('hpeplugin_driver import driver failed'))
LOG.error(msg)
raise exception.HPEPluginNotInitializedException(reason=msg)
try:
self.hpeplugin_driver.do_setup()
self.hpeplugin_driver.check_for_setup_error()
except Exception as ex:
msg = (_('hpeplugin_driver do_setup failed, error is: %s'),
six.text_type(ex))
LOG.error(msg)
raise exception.HPEPluginNotInitializedException(reason=msg)
self._voltracker = {}
self._path_info = []
self._my_ip = netutils.get_my_ipv4()
self._etcd = util.EtcdUtil(self._hpepluginconfig.host_etcd_ip_address,
self._hpepluginconfig.host_etcd_port_number,
self._hpepluginconfig.host_etcd_client_cert,
self._hpepluginconfig.host_etcd_client_key)
# TODO: make device_scan_attempts configurable
# see nova/virt/libvirt/volume/iscsi.py
root_helper = 'sudo'
self.use_multipath = self._hpepluginconfig.use_multipath
self.enforce_multipath = self._hpepluginconfig.enforce_multipath
self.connector = connector.InitiatorConnector.factory(
'ISCSI',
root_helper,
use_multipath=self.use_multipath,
device_scan_attempts=5,
transport='default')
def osd_pool_delete(ceph_api, pool_name):
"""Delete an osd pool
:param pool_name: pool name
"""
response, body = ceph_api.osd_pool_delete(
pool_name,
pool_name,
sure='--yes-i-really-really-mean-it',
body='json')
if response.ok:
LOG.info(_LI("Deleted OSD pool {}").format(pool_name))
else:
e = exception.CephPoolDeleteFailure(name=pool_name,
reason=response.reason)
LOG.warn(e)
raise e
def rados_cache_flush_evict_all(pool):
backing_pool = pool['pool_name']
cache_pool = backing_pool + '-cache'
try:
subprocess.check_call(
['/usr/bin/rados', '-p', cache_pool, 'cache-flush-evict-all'])
LOG.info(
_LI("Flushed OSD cache pool:"
"cache_pool={}").format(cache_pool))
except subprocess.CalledProcessError as e:
_e = exception.CephCacheFlushFailure(cache_pool=cache_pool,
return_code=str(e.returncode),
cmd=" ".join(e.cmd),
output=e.output)
LOG.warn(_e)
raise _e
def __init__(self, reactor, hpepluginconfig):
"""
:param IReactorTime reactor: Reactor time interface implementation.
:param Ihpepluginconfig : hpedefaultconfig configuration
"""
LOG.info(_LI('Initialize Volume Plugin'))
self._reactor = reactor
self._hpepluginconfig = hpepluginconfig
hpeplugin_driver = hpepluginconfig.hpedockerplugin_driver
self.hpeplugin_driver = \
importutils.import_object(hpeplugin_driver, self._hpepluginconfig)
if self.hpeplugin_driver is None:
msg = (_('hpeplugin_driver import driver failed'))
LOG.error(msg)
raise exception.HPEPluginNotInitializedException(reason=msg)
try:
self.hpeplugin_driver.do_setup()
self.hpeplugin_driver.check_for_setup_error()
except Exception as ex:
msg = (_('hpeplugin_driver do_setup failed, error is: %s'),
six.text_type(ex))
LOG.error(msg)
raise exception.HPEPluginNotInitializedException(reason=msg)
self._voltracker = {}
self._path_info = []
self._my_ip = netutils.get_my_ipv4()
self._etcd = util.EtcdUtil(
self._hpepluginconfig.host_etcd_ip_address,
self._hpepluginconfig.host_etcd_port_number,
self._hpepluginconfig.host_etcd_client_cert,
self._hpepluginconfig.host_etcd_client_key)
# TODO: make device_scan_attempts configurable
# see nova/virt/libvirt/volume/iscsi.py
root_helper = 'sudo'
self.use_multipath = self._hpepluginconfig.use_multipath
self.enforce_multipath = self._hpepluginconfig.enforce_multipath
self.connector = connector.InitiatorConnector.factory(
'ISCSI', root_helper, use_multipath=self.use_multipath,
device_scan_attempts=5, transport='default')
def set_flag_require_jewel_osds(self):
try:
response, body = self.service.ceph_api.osd_set_key(
constants.CEPH_FLAG_REQUIRE_JEWEL_OSDS, body='json')
LOG.info(_LI("Set require_jewel_osds flag"))
except IOError as e:
raise exception.CephApiFailure(call="osd_set_key",
reason=e.message)
else:
if not response.ok:
raise exception.CephSetKeyFailure(
flag=constants.CEPH_FLAG_REQUIRE_JEWEL_OSDS,
extra=_("needed to complete upgrade to Jewel"),
response_status_code=response.status_code,
response_reason=response.reason,
status=body.get('status'),
output=body.get('output'))
def volumedriver_remove(self, name):
"""
Remove a Docker volume.
:param unicode name: The name of the volume.
:return: Result indicating success.
"""
contents = json.loads(name.content.getvalue())
volname = contents['Name']
# Only 1 node in a multinode cluster can try to remove the volume.
# Grab lock for volume name. If lock is inuse, just return with no
# error.
self._lock_volume(volname, 'Remove')
vol = self._etcd.get_vol_byname(volname)
if vol is None:
# Just log an error, but don't fail the docker rm command
msg = (_LE('Volume remove name not found %s'), volname)
LOG.error(msg)
self._unlock_volume(volname)
return json.dumps({u"Err": ''})
try:
self.hpeplugin_driver.delete_volume(vol)
LOG.info(_LI('volume: %(name)s,'
'was successfully deleted'), {'name': volname})
except Exception as ex:
msg = (_LE('Err: Failed to remove volume %s, error is %s'),
volname, six.text_type(ex))
LOG.error(msg)
self._unlock_volume(volname)
raise exception.HPEPluginRemoveException(reason=msg)
try:
self._etcd.delete_vol(vol)
except KeyError:
msg = (_LW('Warning: Failed to delete volume key: %s from '
'etcd due to KeyError'), volname)
LOG.warning(msg)
pass
self._unlock_volume(volname)
return json.dumps({u"Err": ''})
def volumedriver_remove(self, name):
"""
Remove a Docker volume.
:param unicode name: The name of the volume.
:return: Result indicating success.
"""
contents = json.loads(name.content.getvalue())
volname = contents['Name']
# Only 1 node in a multinode cluster can try to remove the volume.
# Grab lock for volume name. If lock is inuse, just return with no
# error.
self._lock_volume(volname, 'Remove')
vol = self._etcd.get_vol_byname(volname)
if vol is None:
# Just log an error, but don't fail the docker rm command
msg = (_LE('Volume remove name not found %s'), volname)
LOG.error(msg)
self._unlock_volume(volname)
return json.dumps({u"Err": ''})
try:
self.hpeplugin_driver.delete_volume(vol)
LOG.info(_LI('volume: %(name)s,' 'was successfully deleted'),
{'name': volname})
except Exception as ex:
msg = (_LE('Err: Failed to remove volume %s, error is %s'),
volname, six.text_type(ex))
LOG.error(msg)
self._unlock_volume(volname)
raise exception.HPEPluginRemoveException(reason=msg)
try:
self._etcd.delete_vol(vol)
except KeyError:
msg = (_LW('Warning: Failed to delete volume key: %s from '
'etcd due to KeyError'), volname)
LOG.warning(msg)
pass
self._unlock_volume(volname)
return json.dumps({u"Err": ''})
def cache_overlay_delete(self, pool):
backing_pool = pool['pool_name']
cache_pool = pool['pool_name']
response, body = self.service.ceph_api.osd_tier_remove_overlay(
backing_pool, body='json')
if response.ok:
LOG.info(
_LI("Removed OSD tier overlay: "
"backing_pool={}").format(backing_pool))
else:
e = exception.CephCacheDeleteOverlayFailure(
backing_pool=backing_pool,
cache_pool=cache_pool,
response_status_code=response.status_code,
response_reason=response.reason,
status=body.get('status'),
output=body.get('output'))
LOG.warn(e)
raise e
def ceph_get_fsid(self):
# Check whether an alarm has already been raised
self._get_current_alarms()
if self.current_health_alarm:
LOG.info(
_LI("Current alarm: %s") %
str(self.current_health_alarm.__dict__))
fsid = self._get_fsid()
if not fsid:
# Raise alarm - it will not have an entity_instance_id
self._report_fault(
{
'health': constants.CEPH_HEALTH_DOWN,
'detail': 'Ceph cluster is down.'
}, fm_constants.FM_ALARM_ID_STORAGE_CEPH)
else:
# Clear alarm with no entity_instance_id
self._clear_fault(fm_constants.FM_ALARM_ID_STORAGE_CEPH)
self.service.entity_instance_id = 'cluster=%s' % fsid
def cache_mode_set(self, pool, mode):
backing_pool = pool['pool_name']
cache_pool = backing_pool + '-cache'
response, body = self.service.ceph_api.osd_tier_cachemode(cache_pool,
mode,
body='json')
if response.ok:
LOG.info(
_LI("Set OSD tier cache mode: "
"cache_pool={}, mode={}").format(cache_pool, mode))
else:
e = exception.CephCacheSetModeFailure(
cache_pool=cache_pool,
mode=mode,
response_status_code=response.status_code,
response_reason=response.reason,
status=body.get('status'),
output=body.get('output'))
LOG.warn(e)
raise e
def cache_tier_add(self, pool):
backing_pool = pool['pool_name']
cache_pool = backing_pool + '-cache'
response, body = self.service.ceph_api.osd_tier_add(
backing_pool,
cache_pool,
force_nonempty="--force-nonempty",
body='json')
if response.ok:
LOG.info(
_LI("Added OSD tier: "
"backing_pool={}, cache_pool={}").format(
backing_pool, cache_pool))
else:
e = exception.CephPoolAddTierFailure(
backing_pool=backing_pool,
cache_pool=cache_pool,
response_status_code=response.status_code,
response_reason=response.reason,
status=body.get('status'),
output=body.get('output'))
LOG.warn(e)
raise e
def delete_vol(self, vol):
volkey = self.volumeroot + vol['id']
self.client.delete(volkey)
LOG.info(_LI('Deleted key: %s from etcd'), volkey)
def save_vol(self, vol):
volkey = self.volumeroot + vol['id']
volval = json.dumps(vol)
self.client.write(volkey, volval)
LOG.info(_LI('Write key: %s to etc, value is: %s'), volkey, volval)
def delete_vol(self, vol):
volkey = self.volumeroot + vol['id']
self.client.delete(volkey)
LOG.info(_LI('Deleted key: %s from etcd'), volkey)
def volumedriver_unmount(self, name):
"""
The Docker container is no longer using the given volume,
so unmount it.
NOTE: Since Docker will automatically call Unmount if the Mount
fails, make sure we properly handle partially completed Mounts.
:param unicode name: The name of the volume.
:return: Result indicating success.
"""
LOG.info(_LI('In volumedriver_unmount'))
contents = json.loads(name.content.getvalue())
volname = contents['Name']
vol = self._etcd.get_vol_byname(volname)
if vol is not None:
volid = vol['id']
else:
msg = (_LE('Volume unmount name not found %s'), volname)
LOG.error(msg)
raise exception.HPEPluginUMountException(reason=msg)
vol_mount = DEFAULT_MOUNT_VOLUME
if ('Opts' in contents and contents['Opts']
and 'mount-volume' in contents['Opts']):
vol_mount = str(contents['Opts']['mount-volume'])
path_info = self._etcd.get_vol_path_info(volname)
if path_info:
path_name = path_info['path']
connection_info = path_info['connection_info']
mount_dir = path_info['mount_dir']
else:
msg = (_LE('Volume unmount path info not found %s'), volname)
LOG.error(msg)
raise exception.HPEPluginUMountException(reason=msg)
# Get connector info from OS Brick
# TODO: retrieve use_multipath and enforce_multipath from config file
root_helper = 'sudo'
connector_info = connector.get_connector_properties(
root_helper,
self._my_ip,
multipath=self.use_multipath,
enforce_multipath=self.enforce_multipath)
# Determine if we need to unmount a previously mounted volume
if vol_mount is DEFAULT_MOUNT_VOLUME:
# unmount directory
fileutil.umount_dir(mount_dir)
# remove directory
fileutil.remove_dir(mount_dir)
# We're deferring the execution of the disconnect_volume as it can take
# substantial
# time (over 2 minutes) to cleanup the iscsi files
if connection_info:
LOG.info(_LI('call os brick to disconnect volume'))
d = threads.deferToThread(self.connector.disconnect_volume,
connection_info['data'], None)
d.addCallbacks(self.disconnect_volume_callback,
self.disconnect_volume_error_callback)
try:
# Call driver to terminate the connection
self.hpeplugin_driver.terminate_connection(vol, connector_info)
LOG.info(
_LI('connection_info: %(connection_info)s, '
'was successfully terminated'),
{'connection_info': json.dumps(connection_info)})
except Exception as ex:
msg = (_LE('connection info termination failed %s'),
six.text_type(ex))
LOG.error(msg)
# Not much we can do here, so just continue on with unmount
# We need to ensure we update etcd path_info so the stale
# path does not stay around
# raise exception.HPEPluginUMountException(reason=msg)
# TODO: Create path_info list as we can mount the volume to multiple
# hosts at the same time.
self._etcd.update_vol(volid, 'path_info', None)
LOG.info(
_LI('path for volume: %(name)s, was successfully removed: '
'%(path_name)s'), {
'name': volname,
'path_name': path_name
})
response = json.dumps({u"Err": ''})
return response
def plugin_activate(self, ignore_body=True):
"""
Return which Docker plugin APIs this object supports.
"""
LOG.info(_LI('In Plugin Activate'))
return json.dumps({u"Implements": [u"VolumeDriver"]})
def disconnect_volume_error_callback(self, connector_info):
LOG.info(
_LI('In disconnect_volume_error_callback: '
'connector info is %s'), json.dumps(connector_info))
def do_disable_cache(self, new_config, applied_config, lock_ownership):
LOG.info(
_LI("cache_tiering_disable_cache: "
"new_config={}, applied_config={}").format(
new_config, applied_config))
with lock_ownership():
success = False
_exception = None
try:
self.config_desired.cache_enabled = False
for pool in CEPH_POOLS:
if (pool['pool_name']
== constants.CEPH_POOL_OBJECT_GATEWAY_NAME_JEWEL
or pool['pool_name']
== constants.CEPH_POOL_OBJECT_GATEWAY_NAME_HAMMER):
object_pool_name = \
self.service.monitor._get_object_pool_name()
pool['pool_name'] = object_pool_name
with self.ignore_ceph_failure():
self.cache_mode_set(pool, 'forward')
for pool in CEPH_POOLS:
if (pool['pool_name']
== constants.CEPH_POOL_OBJECT_GATEWAY_NAME_JEWEL
or pool['pool_name']
== constants.CEPH_POOL_OBJECT_GATEWAY_NAME_HAMMER):
object_pool_name = \
self.service.monitor._get_object_pool_name()
pool['pool_name'] = object_pool_name
retries_left = 3
while True:
try:
self.cache_flush(pool)
break
except exception.CephCacheFlushFailure:
retries_left -= 1
if not retries_left:
# give up
break
else:
time.sleep(1)
for pool in CEPH_POOLS:
if (pool['pool_name']
== constants.CEPH_POOL_OBJECT_GATEWAY_NAME_JEWEL
or pool['pool_name']
== constants.CEPH_POOL_OBJECT_GATEWAY_NAME_HAMMER):
object_pool_name = \
self.service.monitor._get_object_pool_name()
pool['pool_name'] = object_pool_name
with self.ignore_ceph_failure():
self.cache_overlay_delete(pool)
self.cache_tier_remove(pool)
for pool in CEPH_POOLS:
if (pool['pool_name']
== constants.CEPH_POOL_OBJECT_GATEWAY_NAME_JEWEL
or pool['pool_name']
== constants.CEPH_POOL_OBJECT_GATEWAY_NAME_HAMMER):
object_pool_name = \
self.service.monitor._get_object_pool_name()
pool['pool_name'] = object_pool_name
with self.ignore_ceph_failure():
self.cache_pool_delete(pool)
success = True
except Exception as e:
LOG.warn(
_LE('Failed to disable cache: reason=%s') %
traceback.format_exc())
_exception = str(e)
finally:
self.service.monitor.monitor_check_cache_tier(False)
if success:
self.config_desired.cache_enabled = False
self.config_applied.cache_enabled = False
self.service.sysinv_conductor.call(
{},
'cache_tiering_disable_cache_complete',
success=success,
exception=_exception,
new_config=new_config.to_dict(),
applied_config=applied_config.to_dict())
def volumedriver_unmount(self, name):
"""
The Docker container is no longer using the given volume,
so unmount it.
NOTE: Since Docker will automatically call Unmount if the Mount
fails, make sure we properly handle partially completed Mounts.
:param unicode name: The name of the volume.
:return: Result indicating success.
"""
LOG.info(_LI('In volumedriver_unmount'))
contents = json.loads(name.content.getvalue())
volname = contents['Name']
vol = self._etcd.get_vol_byname(volname)
if vol is not None:
volid = vol['id']
else:
msg = (_LE('Volume unmount name not found %s'), volname)
LOG.error(msg)
raise exception.HPEPluginUMountException(reason=msg)
path_info = self._etcd.get_vol_path_info(volname)
if path_info:
path_name = path_info['path']
connection_info = path_info['connection_info']
mount_dir = path_info['mount_dir']
else:
msg = (_LE('Volume unmount path info not found %s'), volname)
LOG.error(msg)
raise exception.HPEPluginUMountException(reason=msg)
# Get connector info from OS Brick
# TODO: retrieve use_multipath and enforce_multipath from config file
root_helper = 'sudo'
use_multipath = False
connector_info = connector.get_connector_properties(
root_helper, self._my_ip, use_multipath, enforce_multipath=False)
# unmount directory
fileutil.umount_dir(mount_dir)
# remove directory
fileutil.remove_dir(mount_dir)
try:
# Call driver to terminate the connection
self.hpeplugin_driver.terminate_connection(vol, connector_info)
LOG.info(_LI('connection_info: %(connection_info)s, '
'was successfully terminated'),
{'connection_info': json.dumps(connection_info)})
except Exception as ex:
msg = (_LE('connection info termination failed %s'),
six.text_type(ex))
LOG.error(msg)
raise exception.HPEPluginUMountException(reason=msg)
# We're deferring the execution of the disconnect_volume as it can take
# substantial
# time (over 2 minutes) to cleanup the iscsi files
if connection_info:
LOG.info(_LI('call os brick to disconnect volume'))
d = threads.deferToThread(self.connector.disconnect_volume,
connection_info['data'], None)
d.addCallbacks(self.disconnect_volume_callback,
self.disconnect_volume_error_callback)
# TODO(leeantho) Without this sleep the volume is sometimes not
# removed after the unmount. There must be a different way to fix
# the issue?
time.sleep(1)
# TODO: Create path_info list as we can mount the volume to multiple
# hosts at the same time.
self._etcd.update_vol(volid, 'path_info', None)
LOG.info(_LI('path for volume: %(name)s, was successfully removed: '
'%(path_name)s'), {'name': volname,
'path_name': path_name})
response = json.dumps({u"Err": ''})
return response
def save_vol(self, vol):
volkey = self.volumeroot + vol['id']
volval = json.dumps(vol)
self.client.write(volkey, volval)
LOG.info(_LI('Write key: %s to etc, value is: %s'), volkey, volval)
def plugin_activate(self, ignore_body=True):
"""
Return which Docker plugin APIs this object supports.
"""
LOG.info(_LI('In Plugin Activate'))
return json.dumps({u"Implements": [u"VolumeDriver"]})
def disconnect_volume_error_callback(self, connector_info):
LOG.info(_LI('In disconnect_volume_error_callback: '
'connector info is %s'), json.dumps(connector_info))
six.text_type(ex))
LOG.error(msg)
raise exception.HPEPluginStartPluginException(reason=msg)
# Set Logging level
logging_level = hpedefaultconfig.logging
setup_logging('hpe_storage_api', logging_level)
self._create_listening_directory(PLUGIN_PATH.parent())
endpoint = serverFromString(self._reactor, "unix:{}:mode=600".
format(PLUGIN_PATH.path))
servicename = StreamServerEndpointService(endpoint, Site(
VolumePlugin(self._reactor, hpedefaultconfig).app.resource()))
return servicename
hpedockerplugin = HPEDockerPluginService()
# this will hold the services that combine to form the poetry server
top_service = service.MultiService()
hpepluginservice = hpedockerplugin.setupservice()
hpepluginservice.setServiceParent(top_service)
# this variable has to be named 'application'
application = service.Application("hpedockerplugin")
# this hooks the collection we made to the application
top_service.setServiceParent(application)
LOG.info(_LI('HPE Docker Volume Plugin Successfully Started'))
def volumedriver_unmount(self, name):
"""
The Docker container is no longer using the given volume,
so unmount it.
NOTE: Since Docker will automatically call Unmount if the Mount
fails, make sure we properly handle partially completed Mounts.
:param unicode name: The name of the volume.
:return: Result indicating success.
"""
LOG.info(_LI('In volumedriver_unmount'))
contents = json.loads(name.content.getvalue())
volname = contents['Name']
vol = self._etcd.get_vol_byname(volname)
if vol is not None:
volid = vol['id']
else:
msg = (_LE('Volume unmount name not found %s'), volname)
LOG.error(msg)
raise exception.HPEPluginUMountException(reason=msg)
vol_mount = DEFAULT_MOUNT_VOLUME
if ('Opts' in contents and contents['Opts'] and
'mount-volume' in contents['Opts']):
vol_mount = str(contents['Opts']['mount-volume'])
path_info = self._etcd.get_vol_path_info(volname)
if path_info:
path_name = path_info['path']
connection_info = path_info['connection_info']
mount_dir = path_info['mount_dir']
else:
msg = (_LE('Volume unmount path info not found %s'), volname)
LOG.error(msg)
raise exception.HPEPluginUMountException(reason=msg)
# Get connector info from OS Brick
# TODO: retrieve use_multipath and enforce_multipath from config file
root_helper = 'sudo'
connector_info = connector.get_connector_properties(
root_helper, self._my_ip, multipath=self.use_multipath,
enforce_multipath=self.enforce_multipath)
# Determine if we need to unmount a previously mounted volume
if vol_mount is DEFAULT_MOUNT_VOLUME:
# unmount directory
fileutil.umount_dir(mount_dir)
# remove directory
fileutil.remove_dir(mount_dir)
# We're deferring the execution of the disconnect_volume as it can take
# substantial
# time (over 2 minutes) to cleanup the iscsi files
if connection_info:
LOG.info(_LI('call os brick to disconnect volume'))
d = threads.deferToThread(self.connector.disconnect_volume,
connection_info['data'], None)
d.addCallbacks(self.disconnect_volume_callback,
self.disconnect_volume_error_callback)
try:
# Call driver to terminate the connection
self.hpeplugin_driver.terminate_connection(vol, connector_info)
LOG.info(_LI('connection_info: %(connection_info)s, '
'was successfully terminated'),
{'connection_info': json.dumps(connection_info)})
except Exception as ex:
msg = (_LE('connection info termination failed %s'),
six.text_type(ex))
LOG.error(msg)
# Not much we can do here, so just continue on with unmount
# We need to ensure we update etcd path_info so the stale
# path does not stay around
# raise exception.HPEPluginUMountException(reason=msg)
# TODO: Create path_info list as we can mount the volume to multiple
# hosts at the same time.
self._etcd.update_vol(volid, 'path_info', None)
LOG.info(_LI('path for volume: %(name)s, was successfully removed: '
'%(path_name)s'), {'name': volname,
'path_name': path_name})
response = json.dumps({u"Err": ''})
return response
