def list(self, vpool_guid=None):
"""
Overview of all StorageDrivers
:param vpool_guid: Guid of the vPool
:type vpool_guid: str
"""
if vpool_guid is not None:
return VPool(vpool_guid).storagedrivers
return StorageDriverList.get_storagedrivers()
def getEdgeconnection(vpoolname=VPOOLNAME):
protocol = getEdgeProtocol()
storagedrivers = list(StorageDriverList.get_storagedrivers())
random.shuffle(storagedrivers)
for storagedriver in storagedrivers:
if storagedriver.status == 'FAILURE':
continue
if (vpoolname is not None and storagedriver.vpool.name == vpoolname) or \
(vpoolname is None and storagedriver.vpool.name != VPOOLNAME):
return storagedriver.storage_ip, storagedriver.ports['edge'], protocol
return None, None, protocol
def create(self):
"""
Prepares a new Storagedriver for a given vPool and Storagerouter
:return: None
:rtype: NoneType
"""
if self.sr_installer is None:
raise RuntimeError('No StorageRouterInstaller instance found')
machine_id = System.get_my_machine_id(client=self.sr_installer.root_client)
port_range = Configuration.get('/ovs/framework/hosts/{0}/ports|storagedriver'.format(machine_id))
storagerouter = self.sr_installer.storagerouter
with volatile_mutex('add_vpool_get_free_ports_{0}'.format(machine_id), wait=30):
model_ports_in_use = []
for sd in StorageDriverList.get_storagedrivers():
if sd.storagerouter_guid == storagerouter.guid:
model_ports_in_use += sd.ports.values()
for proxy in sd.alba_proxies:
model_ports_in_use.append(proxy.service.ports[0])
ports = System.get_free_ports(selected_range=port_range, exclude=model_ports_in_use, amount=4 + self.sr_installer.requested_proxies, client=self.sr_installer.root_client)
vpool = self.vp_installer.vpool
vrouter_id = '{0}{1}'.format(vpool.name, machine_id)
storagedriver = StorageDriver()
storagedriver.name = vrouter_id.replace('_', ' ')
storagedriver.ports = {'management': ports[0],
'xmlrpc': ports[1],
'dtl': ports[2],
'edge': ports[3]}
storagedriver.vpool = vpool
storagedriver.cluster_ip = Configuration.get('/ovs/framework/hosts/{0}/ip'.format(machine_id))
storagedriver.storage_ip = self.storage_ip
storagedriver.mountpoint = '/mnt/{0}'.format(vpool.name)
storagedriver.description = storagedriver.name
storagedriver.storagerouter = storagerouter
storagedriver.storagedriver_id = vrouter_id
storagedriver.save()
# ALBA Proxies
proxy_service_type = ServiceTypeList.get_by_name(ServiceType.SERVICE_TYPES.ALBA_PROXY)
for proxy_id in xrange(self.sr_installer.requested_proxies):
service = Service()
service.storagerouter = storagerouter
service.ports = [ports[4 + proxy_id]]
service.name = 'albaproxy_{0}_{1}'.format(vpool.name, proxy_id)
service.type = proxy_service_type
service.save()
alba_proxy = AlbaProxy()
alba_proxy.service = service
alba_proxy.storagedriver = storagedriver
alba_proxy.save()
self.storagedriver = storagedriver
def get_storagedrivers_in_same_domain(domain_guid):
"""
Get storagerouter guids in a domain
:param domain_guid: guid of a domain
:type domain_guid: str
:return: list of storagerouter guids
:rtype: list
"""
return [
storagedriver
for storagedriver in StorageDriverList.get_storagedrivers()
if domain_guid in storagedriver.storagerouter.regular_domains
]
def delete_snapshots(timestamp=None):
# type: (float) -> GroupResult
"""
Delete snapshots based on the retention policy
Offloads concurrency to celery
Returns a GroupResult. Waiting for the result can be done using result.get()
:param timestamp: Timestamp to determine whether snapshots should be kept or not, if none provided, current time will be used
:type timestamp: float
:return: The GroupResult
:rtype: GroupResult
"""
# The result cannot be fetched in this task
group_id = uuid()
return group(GenericController.delete_snapshots_storagedriver.s(storagedriver.guid, timestamp, group_id)
for storagedriver in StorageDriverList.get_storagedrivers()).apply_async(task_id=group_id)
def update_vmachine_config(vmachine, vm_object, pmachine=None):
"""
Update a vMachine configuration with a given vMachine configuration
:param vmachine: Virtual Machine to update
:param vm_object: New virtual machine info
:param pmachine: Physical machine of the virtual machine
"""
try:
vdisks_synced = 0
if vmachine.name is None:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vmachine_created',
'metadata': {'name': vm_object['name']}})
elif vmachine.name != vm_object['name']:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vmachine_renamed',
'metadata': {'old_name': vmachine.name,
'new_name': vm_object['name']}})
if pmachine is not None:
vmachine.pmachine = pmachine
vmachine.name = vm_object['name']
vmachine.hypervisor_id = vm_object['id']
vmachine.devicename = vm_object['backing']['filename']
vmachine.save()
# Updating and linking disks
storagedrivers = StorageDriverList.get_storagedrivers()
datastores = dict([('{0}:{1}'.format(storagedriver.storage_ip, storagedriver.mountpoint), storagedriver) for storagedriver in storagedrivers])
vdisk_guids = []
mutex = volatile_mutex('{0}_{1}'.format(vmachine.name, vmachine.devicename))
for disk in vm_object['disks']:
ensure_safety = False
if disk['datastore'] in vm_object['datastores']:
datastore = vm_object['datastores'][disk['datastore']]
if datastore in datastores:
try:
mutex.acquire(wait=10)
vdisk = VDiskList.get_by_devicename_and_vpool(disk['filename'], datastores[datastore].vpool)
if vdisk is None:
# The disk couldn't be located, but is in our datastore. We might be in a recovery scenario
vdisk = VDisk()
vdisk.vpool = datastores[datastore].vpool
vdisk.reload_client()
vdisk.devicename = disk['filename']
vdisk.volume_id = vdisk.storagedriver_client.get_volume_id(str(disk['backingfilename']))
vdisk.size = vdisk.info['volume_size']
vdisk.metadata = {'lba_size': vdisk.info['lba_size'],
'cluster_multiplier': vdisk.info['cluster_multiplier']}
# Create the disk in a locked context, but don't execute long running-task in same context
vdisk.save()
ensure_safety = True
finally:
mutex.release()
if ensure_safety:
MDSServiceController.ensure_safety(vdisk)
VDiskController.dtl_checkup(vdisk_guid=vdisk.guid)
# Update the disk with information from the hypervisor
if vdisk.vmachine is None:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vdisk_attached',
'metadata': {'vmachine_name': vmachine.name,
'vdisk_name': disk['name']}})
vdisk.vmachine = vmachine
vdisk.name = disk['name']
vdisk.order = disk['order']
vdisk.save()
vdisk_guids.append(vdisk.guid)
vdisks_synced += 1
for vdisk in vmachine.vdisks:
if vdisk.guid not in vdisk_guids:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vdisk_detached',
'metadata': {'vmachine_name': vmachine.name,
'vdisk_name': vdisk.name}})
vdisk.vmachine = None
vdisk.save()
VMachineController._logger.info('Updating vMachine finished (name {0}, {1} vdisks (re)linked)'.format(
vmachine.name, vdisks_synced
))
except Exception as ex:
VMachineController._logger.info('Error during vMachine update: {0}'.format(str(ex)))
raise
def migrate():
"""
Executes async migrations. It doesn't matter too much when they are executed, as long as they get eventually
executed. This code will typically contain:
* "dangerous" migration code (it needs certain running services)
* Migration code depending on a cluster-wide state
* ...
"""
MigrationController._logger.info('Preparing out of band migrations...')
from ovs.dal.lists.storagedriverlist import StorageDriverList
from ovs.dal.lists.storagerouterlist import StorageRouterList
from ovs.dal.lists.vpoollist import VPoolList
from ovs.extensions.generic.configuration import Configuration
from ovs.extensions.generic.sshclient import SSHClient
from ovs_extensions.generic.toolbox import ExtensionsToolbox
from ovs_extensions.services.interfaces.systemd import Systemd
from ovs.extensions.services.servicefactory import ServiceFactory
from ovs.extensions.storageserver.storagedriver import StorageDriverConfiguration
from ovs.lib.generic import GenericController
MigrationController._logger.info('Start out of band migrations...')
service_manager = ServiceFactory.get_manager()
sr_client_map = {}
for storagerouter in StorageRouterList.get_storagerouters():
sr_client_map[storagerouter.guid] = SSHClient(endpoint=storagerouter,
username='******')
#########################################################
# Addition of 'ExecReload' for AlbaProxy SystemD services
if ServiceFactory.get_service_type() == 'systemd':
changed_clients = set()
for storagedriver in StorageDriverList.get_storagedrivers():
root_client = sr_client_map[storagedriver.storagerouter_guid]
for alba_proxy in storagedriver.alba_proxies:
service = alba_proxy.service
service_name = 'ovs-{0}'.format(service.name)
if not service_manager.has_service(name=service_name, client=root_client):
continue
if 'ExecReload=' in root_client.file_read(filename='/lib/systemd/system/{0}.service'.format(service_name)):
continue
try:
service_manager.regenerate_service(name='ovs-albaproxy', client=root_client, target_name=service_name)
changed_clients.add(root_client)
except:
MigrationController._logger.exception('Error rebuilding service {0}'.format(service_name))
for root_client in changed_clients:
root_client.run(['systemctl', 'daemon-reload'])
##################################################################
# Adjustment of open file descriptors for Arakoon services to 8192
changed_clients = set()
for storagerouter in StorageRouterList.get_storagerouters():
root_client = sr_client_map[storagerouter.guid]
for service_name in service_manager.list_services(client=root_client):
if not service_name.startswith('ovs-arakoon-'):
continue
if ServiceFactory.get_service_type() == 'systemd':
path = '/lib/systemd/system/{0}.service'.format(service_name)
check = 'LimitNOFILE=8192'
else:
path = '/etc/init/{0}.conf'.format(service_name)
check = 'limit nofile 8192 8192'
if not root_client.file_exists(path):
continue
if check in root_client.file_read(path):
continue
try:
service_manager.regenerate_service(name='ovs-arakoon', client=root_client, target_name=service_name)
changed_clients.add(root_client)
ExtensionsToolbox.edit_version_file(client=root_client, package_name='arakoon', old_service_name=service_name)
except:
MigrationController._logger.exception('Error rebuilding service {0}'.format(service_name))
for root_client in changed_clients:
root_client.run(['systemctl', 'daemon-reload'])
#############################
# Migrate to multiple proxies
for storagedriver in StorageDriverList.get_storagedrivers():
vpool = storagedriver.vpool
root_client = sr_client_map[storagedriver.storagerouter_guid]
for alba_proxy in storagedriver.alba_proxies:
# Rename alba_proxy service in model
service = alba_proxy.service
old_service_name = 'albaproxy_{0}'.format(vpool.name)
new_service_name = 'albaproxy_{0}_0'.format(vpool.name)
if old_service_name != service.name:
continue
service.name = new_service_name
service.save()
if not service_manager.has_service(name=old_service_name, client=root_client):
continue
old_configuration_key = '/ovs/framework/hosts/{0}/services/{1}'.format(storagedriver.storagerouter.machine_id, old_service_name)
if not Configuration.exists(key=old_configuration_key):
continue
# Add '-reboot' to alba_proxy services (because of newly created services and removal of old service)
ExtensionsToolbox.edit_version_file(client=root_client,
package_name='alba',
old_service_name=old_service_name,
new_service_name=new_service_name)
# Register new service and remove old service
service_manager.add_service(name='ovs-albaproxy',
client=root_client,
params=Configuration.get(old_configuration_key),
target_name='ovs-{0}'.format(new_service_name))
# Update scrub proxy config
proxy_config_key = '/ovs/vpools/{0}/proxies/{1}/config/main'.format(vpool.guid, alba_proxy.guid)
proxy_config = None if Configuration.exists(key=proxy_config_key) is False else Configuration.get(proxy_config_key)
if proxy_config is not None:
fragment_cache = proxy_config.get('fragment_cache', ['none', {}])
if fragment_cache[0] == 'alba' and fragment_cache[1].get('cache_on_write') is True: # Accelerated ALBA configured
fragment_cache_scrub_info = copy.deepcopy(fragment_cache)
fragment_cache_scrub_info[1]['cache_on_read'] = False
proxy_scrub_config_key = '/ovs/vpools/{0}/proxies/scrub/generic_scrub'.format(vpool.guid)
proxy_scrub_config = None if Configuration.exists(key=proxy_scrub_config_key) is False else Configuration.get(proxy_scrub_config_key)
if proxy_scrub_config is not None and proxy_scrub_config['fragment_cache'] == ['none']:
proxy_scrub_config['fragment_cache'] = fragment_cache_scrub_info
Configuration.set(proxy_scrub_config_key,
json.dumps(proxy_scrub_config, indent=4),
raw=True)
# Update 'backend_connection_manager' section
changes = False
storagedriver_config = StorageDriverConfiguration('storagedriver', vpool.guid, storagedriver.storagedriver_id)
storagedriver_config.load()
if 'backend_connection_manager' not in storagedriver_config.configuration:
continue
current_config = storagedriver_config.configuration['backend_connection_manager']
if current_config.get('backend_type') != 'MULTI':
changes = True
backend_connection_manager = {'backend_type': 'MULTI'}
for index, proxy in enumerate(sorted(storagedriver.alba_proxies, key=lambda pr: pr.service.ports[0])):
backend_connection_manager[str(index)] = copy.deepcopy(current_config)
# noinspection PyUnresolvedReferences
backend_connection_manager[str(index)]['alba_connection_use_rora'] = True
# noinspection PyUnresolvedReferences
backend_connection_manager[str(index)]['alba_connection_rora_manifest_cache_capacity'] = 5000
# noinspection PyUnresolvedReferences
for key, value in backend_connection_manager[str(index)].items():
if key.startswith('backend_interface'):
backend_connection_manager[key] = value
# noinspection PyUnresolvedReferences
del backend_connection_manager[str(index)][key]
for key, value in {'backend_interface_retries_on_error': 5,
'backend_interface_retry_interval_secs': 1,
'backend_interface_retry_backoff_multiplier': 2.0}.iteritems():
if key not in backend_connection_manager:
backend_connection_manager[key] = value
else:
backend_connection_manager = current_config
for value in backend_connection_manager.values():
if isinstance(value, dict):
for key, val in value.items():
if key.startswith('backend_interface'):
backend_connection_manager[key] = val
changes = True
del value[key]
for key, value in {'backend_interface_retries_on_error': 5,
'backend_interface_retry_interval_secs': 1,
'backend_interface_retry_backoff_multiplier': 2.0}.iteritems():
if key not in backend_connection_manager:
changes = True
backend_connection_manager[key] = value
if changes is True:
storagedriver_config.clear_backend_connection_manager()
storagedriver_config.configure_backend_connection_manager(**backend_connection_manager)
storagedriver_config.save(root_client)
# Add '-reboot' to volumedriver services (because of updated 'backend_connection_manager' section)
ExtensionsToolbox.edit_version_file(client=root_client,
package_name='volumedriver',
old_service_name='volumedriver_{0}'.format(vpool.name))
if service_manager.ImplementationClass == Systemd:
root_client.run(['systemctl', 'daemon-reload'])
########################################
# Update metadata_store_bits information
for vpool in VPoolList.get_vpools():
bits = None
for storagedriver in vpool.storagedrivers:
key = '/ovs/framework/hosts/{0}/services/volumedriver_{1}'.format(storagedriver.storagerouter.machine_id, vpool.name)
if Configuration.exists(key=key) and 'METADATASTORE_BITS' not in Configuration.get(key=key):
if bits is None:
entries = service_manager.extract_from_service_file(name='ovs-volumedriver_{0}'.format(vpool.name),
client=sr_client_map[storagedriver.storagerouter_guid],
entries=['METADATASTORE_BITS='])
if len(entries) == 1:
bits = entries[0].split('=')[-1]
bits = int(bits) if bits.isdigit() else 5
if bits is not None:
try:
content = Configuration.get(key=key)
content['METADATASTORE_BITS'] = bits
Configuration.set(key=key, value=content)
except:
MigrationController._logger.exception('Error updating volumedriver info for vPool {0} on StorageRouter {1}'.format(vpool.name, storagedriver.storagerouter.name))
if bits is not None:
vpool.metadata_store_bits = bits
vpool.save()
MigrationController._logger.info('Finished out of band migrations')
GenericController.refresh_package_information()
def gather_scrub_work():
logger.info("Divide scrubbing work among allowed Storage Routers")
scrub_locations = {}
for storage_driver in StorageDriverList.get_storagedrivers():
for partition in storage_driver.partitions:
if DiskPartition.ROLES.SCRUB == partition.role:
logger.info(
"Scrub partition found on Storage Router {0}: {1}".format(storage_driver.name, partition.folder)
)
if storage_driver.storagerouter not in scrub_locations:
try:
_ = SSHClient(storage_driver.storagerouter.ip)
scrub_locations[storage_driver.storagerouter] = str(partition.path)
except UnableToConnectException:
logger.warning("StorageRouter {0} is not reachable".format(storage_driver.storagerouter.ip))
if len(scrub_locations) == 0:
raise RuntimeError("No scrub locations found")
vdisk_guids = set()
for vmachine in VMachineList.get_customer_vmachines():
for vdisk in vmachine.vdisks:
if vdisk.info["object_type"] in ["BASE"] and len(vdisk.child_vdisks) == 0:
vdisk_guids.add(vdisk.guid)
for vdisk in VDiskList.get_without_vmachine():
if vdisk.info["object_type"] in ["BASE"] and len(vdisk.child_vdisks) == 0:
vdisk_guids.add(vdisk.guid)
logger.info("Found {0} virtual disks which need to be check for scrub work".format(len(vdisk_guids)))
local_machineid = System.get_my_machine_id()
local_scrub_location = None
local_vdisks_to_scrub = []
result_set = ResultSet([])
storage_router_list = []
for index, scrub_info in enumerate(scrub_locations.items()):
start_index = index * len(vdisk_guids) / len(scrub_locations)
end_index = (index + 1) * len(vdisk_guids) / len(scrub_locations)
storage_router = scrub_info[0]
vdisk_guids_to_scrub = list(vdisk_guids)[start_index:end_index]
local = storage_router.machine_id == local_machineid
logger.info(
"Executing scrub work on {0} Storage Router {1} for {2} virtual disks".format(
"local" if local is True else "remote", storage_router.name, len(vdisk_guids_to_scrub)
)
)
if local is True:
local_scrub_location = scrub_info[1]
local_vdisks_to_scrub = vdisk_guids_to_scrub
else:
result_set.add(
ScheduledTaskController._execute_scrub_work.s(
scrub_location=scrub_info[1], vdisk_guids=vdisk_guids_to_scrub
).apply_async(routing_key="sr.{0}".format(storage_router.machine_id))
)
storage_router_list.append(storage_router)
logger.info("Launched scrub task on Storage Router {0}".format(storage_router.name))
# Remote tasks have been launched, now start the local task and then wait for remote tasks to finish
if local_scrub_location is not None and len(local_vdisks_to_scrub) > 0:
ScheduledTaskController._execute_scrub_work(
scrub_location=local_scrub_location, vdisk_guids=local_vdisks_to_scrub
)
all_results = result_set.join(
propagate=False
) # Propagate False makes sure all jobs are waited for even when 1 or more jobs fail
for index, result in enumerate(all_results):
if result is not None:
logger.error(
"Scrubbing failed on Storage Router {0} with error {1}".format(
storage_router_list[index].name, result
)
)
def migrate(previous_version):
"""
Migrates from any version to any version, running all migrations required
If previous_version is for example 0 and this script is at
version 3 it will execute two steps:
- 1 > 2
- 2 > 3
@param previous_version: The previous version from which to start the migration.
"""
working_version = previous_version
# Version 1 introduced:
# - The datastore is still empty, add defaults
if working_version < 1:
from ovs.dal.hybrids.user import User
from ovs.dal.hybrids.group import Group
from ovs.dal.hybrids.role import Role
from ovs.dal.hybrids.client import Client
from ovs.dal.hybrids.failuredomain import FailureDomain
from ovs.dal.hybrids.j_rolegroup import RoleGroup
from ovs.dal.hybrids.j_roleclient import RoleClient
from ovs.dal.hybrids.backendtype import BackendType
from ovs.dal.hybrids.servicetype import ServiceType
from ovs.dal.hybrids.branding import Branding
from ovs.dal.lists.backendtypelist import BackendTypeList
# Create groups
admin_group = Group()
admin_group.name = 'administrators'
admin_group.description = 'Administrators'
admin_group.save()
viewers_group = Group()
viewers_group.name = 'viewers'
viewers_group.description = 'Viewers'
viewers_group.save()
# Create users
admin = User()
admin.username = '******'
admin.password = hashlib.sha256('admin').hexdigest()
admin.is_active = True
admin.group = admin_group
admin.save()
# Create internal OAuth 2 clients
admin_pw_client = Client()
admin_pw_client.ovs_type = 'INTERNAL'
admin_pw_client.grant_type = 'PASSWORD'
admin_pw_client.user = admin
admin_pw_client.save()
admin_cc_client = Client()
admin_cc_client.ovs_type = 'INTERNAL'
admin_cc_client.grant_type = 'CLIENT_CREDENTIALS'
admin_cc_client.client_secret = ''.join(
random.choice(string.ascii_letters + string.digits +
'|_=+*#@!/-[]{}<>.?,\'";:~') for _ in range(128))
admin_cc_client.user = admin
admin_cc_client.save()
# Create roles
read_role = Role()
read_role.code = 'read'
read_role.name = 'Read'
read_role.description = 'Can read objects'
read_role.save()
write_role = Role()
write_role.code = 'write'
write_role.name = 'Write'
write_role.description = 'Can write objects'
write_role.save()
manage_role = Role()
manage_role.code = 'manage'
manage_role.name = 'Manage'
manage_role.description = 'Can manage the system'
manage_role.save()
# Attach groups to roles
mapping = [(admin_group, [read_role, write_role, manage_role]),
(viewers_group, [read_role])]
for setting in mapping:
for role in setting[1]:
rolegroup = RoleGroup()
rolegroup.group = setting[0]
rolegroup.role = role
rolegroup.save()
for user in setting[0].users:
for role in setting[1]:
for client in user.clients:
roleclient = RoleClient()
roleclient.client = client
roleclient.role = role
roleclient.save()
# Add backends
for backend_type_info in [('Ceph', 'ceph_s3'),
('Amazon', 'amazon_s3'),
('Swift', 'swift_s3'),
('Local', 'local'),
('Distributed', 'distributed'),
('ALBA', 'alba')]:
code = backend_type_info[1]
backend_type = BackendTypeList.get_backend_type_by_code(code)
if backend_type is None:
backend_type = BackendType()
backend_type.name = backend_type_info[0]
backend_type.code = code
backend_type.save()
# Add service types
for service_type_info in [
ServiceType.SERVICE_TYPES.MD_SERVER,
ServiceType.SERVICE_TYPES.ALBA_PROXY,
ServiceType.SERVICE_TYPES.ARAKOON
]:
service_type = ServiceType()
service_type.name = service_type_info
service_type.save()
# Branding
branding = Branding()
branding.name = 'Default'
branding.description = 'Default bootstrap theme'
branding.css = 'bootstrap-default.min.css'
branding.productname = 'Open vStorage'
branding.is_default = True
branding.save()
slate = Branding()
slate.name = 'Slate'
slate.description = 'Dark bootstrap theme'
slate.css = 'bootstrap-slate.min.css'
slate.productname = 'Open vStorage'
slate.is_default = False
slate.save()
# Failure Domain
failure_domain = FailureDomain()
failure_domain.name = 'Default'
failure_domain.save()
# We're now at version 1
working_version = 1
# Version 2 introduced:
# - new Descriptor format
if working_version < 2:
import imp
from ovs.dal.helpers import Descriptor
from ovs.extensions.storage.persistentfactory import PersistentFactory
client = PersistentFactory.get_client()
keys = client.prefix('ovs_data')
for key in keys:
data = client.get(key)
modified = False
for entry in data.keys():
if isinstance(data[entry], dict) and 'source' in data[
entry] and 'hybrids' in data[entry]['source']:
filename = data[entry]['source']
if not filename.startswith('/'):
filename = '/opt/OpenvStorage/ovs/dal/{0}'.format(
filename)
module = imp.load_source(data[entry]['name'], filename)
cls = getattr(module, data[entry]['type'])
new_data = Descriptor(cls, cached=False).descriptor
if 'guid' in data[entry]:
new_data['guid'] = data[entry]['guid']
data[entry] = new_data
modified = True
if modified is True:
data['_version'] += 1
client.set(key, data)
# We're now at version 2
working_version = 2
# Version 3 introduced:
# - new Descriptor format
if working_version < 3:
import imp
from ovs.dal.helpers import Descriptor
from ovs.extensions.storage.persistentfactory import PersistentFactory
client = PersistentFactory.get_client()
keys = client.prefix('ovs_data')
for key in keys:
data = client.get(key)
modified = False
for entry in data.keys():
if isinstance(data[entry],
dict) and 'source' in data[entry]:
module = imp.load_source(data[entry]['name'],
data[entry]['source'])
cls = getattr(module, data[entry]['type'])
new_data = Descriptor(cls, cached=False).descriptor
if 'guid' in data[entry]:
new_data['guid'] = data[entry]['guid']
data[entry] = new_data
modified = True
if modified is True:
data['_version'] += 1
client.set(key, data)
working_version = 3
# Version 4 introduced:
# - Flexible SSD layout
if working_version < 4:
import os
from ovs.dal.hybrids.diskpartition import DiskPartition
from ovs.dal.hybrids.j_storagedriverpartition import StorageDriverPartition
from ovs.dal.hybrids.servicetype import ServiceType
from ovs.dal.lists.servicetypelist import ServiceTypeList
from ovs.dal.lists.storagedriverlist import StorageDriverList
from ovs.extensions.generic.remote import remote
from ovs.extensions.generic.sshclient import SSHClient
from ovs.extensions.storageserver.storagedriver import StorageDriverConfiguration
for service in ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.MD_SERVER).services:
mds_service = service.mds_service
storagedriver = None
for current_storagedriver in service.storagerouter.storagedrivers:
if current_storagedriver.vpool_guid == mds_service.vpool_guid:
storagedriver = current_storagedriver
break
tasks = {}
if storagedriver._data.get('mountpoint_md'):
tasks['{0}/mds_{1}_{2}'.format(
storagedriver._data.get('mountpoint_md'),
storagedriver.vpool.name, mds_service.number)] = (
DiskPartition.ROLES.DB,
StorageDriverPartition.SUBROLE.MDS)
if storagedriver._data.get('mountpoint_temp'):
tasks['{0}/mds_{1}_{2}'.format(
storagedriver._data.get('mountpoint_temp'),
storagedriver.vpool.name, mds_service.number)] = (
DiskPartition.ROLES.SCRUB,
StorageDriverPartition.SUBROLE.MDS)
for disk in service.storagerouter.disks:
for partition in disk.partitions:
for directory, (role, subrole) in tasks.iteritems():
with remote(storagedriver.storagerouter.ip, [os],
username='******') as rem:
stat_dir = directory
while not rem.os.path.exists(
stat_dir) and stat_dir != '/':
stat_dir = stat_dir.rsplit('/', 1)[0]
if not stat_dir:
stat_dir = '/'
inode = rem.os.stat(stat_dir).st_dev
if partition.inode == inode:
if role not in partition.roles:
partition.roles.append(role)
partition.save()
number = 0
migrated = False
for sd_partition in storagedriver.partitions:
if sd_partition.role == role and sd_partition.sub_role == subrole:
if sd_partition.mds_service == mds_service:
migrated = True
break
if sd_partition.partition_guid == partition.guid:
number = max(
sd_partition.number, number)
if migrated is False:
sd_partition = StorageDriverPartition()
sd_partition.role = role
sd_partition.sub_role = subrole
sd_partition.partition = partition
sd_partition.storagedriver = storagedriver
sd_partition.mds_service = mds_service
sd_partition.size = None
sd_partition.number = number + 1
sd_partition.save()
client = SSHClient(
storagedriver.storagerouter,
username='******')
path = sd_partition.path.rsplit('/', 1)[0]
if path:
client.dir_create(path)
client.dir_chown(path, 'ovs', 'ovs')
client.dir_create(directory)
client.dir_chown(directory, 'ovs', 'ovs')
client.symlink(
{sd_partition.path: directory})
for storagedriver in StorageDriverList.get_storagedrivers():
migrated_objects = {}
for disk in storagedriver.storagerouter.disks:
for partition in disk.partitions:
# Process all mountpoints that are unique and don't have a specified size
for key, (role, sr_info) in {
'mountpoint_md': (DiskPartition.ROLES.DB, {
'metadata_{0}':
StorageDriverPartition.SUBROLE.MD,
'tlogs_{0}':
StorageDriverPartition.SUBROLE.TLOG
}),
'mountpoint_fragmentcache':
(DiskPartition.ROLES.WRITE, {
'fcache_{0}':
StorageDriverPartition.SUBROLE.FCACHE
}),
'mountpoint_foc': (DiskPartition.ROLES.WRITE, {
'fd_{0}':
StorageDriverPartition.SUBROLE.FD,
'dtl_{0}':
StorageDriverPartition.SUBROLE.DTL
}),
'mountpoint_dtl': (DiskPartition.ROLES.WRITE, {
'fd_{0}':
StorageDriverPartition.SUBROLE.FD,
'dtl_{0}':
StorageDriverPartition.SUBROLE.DTL
}),
'mountpoint_readcaches':
(DiskPartition.ROLES.READ, {
'': None
}),
'mountpoint_writecaches':
(DiskPartition.ROLES.WRITE, {
'sco_{0}': StorageDriverPartition.SUBROLE.SCO
})
}.iteritems():
if key in storagedriver._data:
is_list = isinstance(storagedriver._data[key],
list)
entries = storagedriver._data[
key][:] if is_list is True else [
storagedriver._data[key]
]
for entry in entries:
if not entry:
if is_list:
storagedriver._data[key].remove(
entry)
if len(storagedriver._data[key]
) == 0:
del storagedriver._data[key]
else:
del storagedriver._data[key]
else:
with remote(
storagedriver.storagerouter.ip,
[os],
username='******') as rem:
inode = rem.os.stat(entry).st_dev
if partition.inode == inode:
if role not in partition.roles:
partition.roles.append(role)
partition.save()
for folder, subrole in sr_info.iteritems(
):
number = 0
migrated = False
for sd_partition in storagedriver.partitions:
if sd_partition.role == role and sd_partition.sub_role == subrole:
if sd_partition.partition_guid == partition.guid:
number = max(
sd_partition.
number, number)
if migrated is False:
sd_partition = StorageDriverPartition(
)
sd_partition.role = role
sd_partition.sub_role = subrole
sd_partition.partition = partition
sd_partition.storagedriver = storagedriver
sd_partition.size = None
sd_partition.number = number + 1
sd_partition.save()
if folder:
source = '{0}/{1}'.format(
entry,
folder.format(
storagedriver.
vpool.name))
else:
source = entry
client = SSHClient(
storagedriver.
storagerouter,
username='******')
path = sd_partition.path.rsplit(
'/', 1)[0]
if path:
client.dir_create(path)
client.dir_chown(
path, 'ovs', 'ovs')
client.symlink({
sd_partition.path:
source
})
migrated_objects[
source] = sd_partition
if is_list:
storagedriver._data[
key].remove(entry)
if len(storagedriver._data[key]
) == 0:
del storagedriver._data[
key]
else:
del storagedriver._data[key]
storagedriver.save()
if 'mountpoint_bfs' in storagedriver._data:
storagedriver.mountpoint_dfs = storagedriver._data[
'mountpoint_bfs']
if not storagedriver.mountpoint_dfs:
storagedriver.mountpoint_dfs = None
del storagedriver._data['mountpoint_bfs']
storagedriver.save()
if 'mountpoint_temp' in storagedriver._data:
del storagedriver._data['mountpoint_temp']
storagedriver.save()
if migrated_objects:
print 'Loading sizes'
config = StorageDriverConfiguration(
'storagedriver', storagedriver.vpool_guid,
storagedriver.storagedriver_id)
config.load()
for readcache in config.configuration.get(
'content_addressed_cache',
{}).get('clustercache_mount_points', []):
path = readcache.get('path', '').rsplit('/', 1)[0]
size = int(readcache['size'].strip(
'KiB')) * 1024 if 'size' in readcache else None
if path in migrated_objects:
migrated_objects[path].size = long(size)
migrated_objects[path].save()
for writecache in config.configuration.get(
'scocache', {}).get('scocache_mount_points', []):
path = writecache.get('path', '')
size = int(writecache['size'].strip(
'KiB')) * 1024 if 'size' in writecache else None
if path in migrated_objects:
migrated_objects[path].size = long(size)
migrated_objects[path].save()
working_version = 4
# Version 5 introduced:
# - Failure Domains
if working_version < 5:
import os
from ovs.dal.hybrids.failuredomain import FailureDomain
from ovs.dal.lists.failuredomainlist import FailureDomainList
from ovs.dal.lists.storagerouterlist import StorageRouterList
from ovs.extensions.generic.remote import remote
from ovs.extensions.generic.sshclient import SSHClient
failure_domains = FailureDomainList.get_failure_domains()
if len(failure_domains) > 0:
failure_domain = failure_domains[0]
else:
failure_domain = FailureDomain()
failure_domain.name = 'Default'
failure_domain.save()
for storagerouter in StorageRouterList.get_storagerouters():
change = False
if storagerouter.primary_failure_domain is None:
storagerouter.primary_failure_domain = failure_domain
change = True
if storagerouter.rdma_capable is None:
client = SSHClient(storagerouter, username='******')
rdma_capable = False
with remote(client.ip, [os], username='******') as rem:
for root, dirs, files in rem.os.walk(
'/sys/class/infiniband'):
for directory in dirs:
ports_dir = '/'.join(
[root, directory, 'ports'])
if not rem.os.path.exists(ports_dir):
continue
for sub_root, sub_dirs, _ in rem.os.walk(
ports_dir):
if sub_root != ports_dir:
continue
for sub_directory in sub_dirs:
state_file = '/'.join(
[sub_root, sub_directory, 'state'])
if rem.os.path.exists(state_file):
if 'ACTIVE' in client.run(
'cat {0}'.format(
state_file)):
rdma_capable = True
storagerouter.rdma_capable = rdma_capable
change = True
if change is True:
storagerouter.save()
working_version = 5
# Version 6 introduced:
# - Distributed scrubbing
if working_version < 6:
from ovs.dal.hybrids.diskpartition import DiskPartition
from ovs.dal.lists.storagedriverlist import StorageDriverList
from ovs.extensions.generic.sshclient import SSHClient
for storage_driver in StorageDriverList.get_storagedrivers():
root_client = SSHClient(storage_driver.storagerouter,
username='******')
for partition in storage_driver.partitions:
if partition.role == DiskPartition.ROLES.SCRUB:
old_path = partition.path
partition.sub_role = None
partition.save()
partition.invalidate_dynamics(['folder', 'path'])
if root_client.dir_exists(partition.path):
continue # New directory already exists
if '_mds_' in old_path:
if root_client.dir_exists(old_path):
root_client.symlink({partition.path: old_path})
if not root_client.dir_exists(partition.path):
root_client.dir_create(partition.path)
root_client.dir_chmod(partition.path, 0777)
working_version = 6
# Version 7 introduced:
# - vPool status
if working_version < 7:
from ovs.dal.hybrids import vpool
reload(vpool)
from ovs.dal.hybrids.vpool import VPool
from ovs.dal.lists.vpoollist import VPoolList
for _vpool in VPoolList.get_vpools():
vpool = VPool(_vpool.guid)
if hasattr(vpool, 'status') and vpool.status is None:
vpool.status = VPool.STATUSES.RUNNING
vpool.save()
working_version = 7
# Version 10 introduced:
# - Reverse indexes are stored in persistent store
# - Store more non-changing metadata on disk iso using a dynamic property
if working_version < 10:
from ovs.dal.helpers import HybridRunner, Descriptor
from ovs.dal.datalist import DataList
from ovs.extensions.storage.persistentfactory import PersistentFactory
from ovs.extensions.storage.volatilefactory import VolatileFactory
persistent = PersistentFactory.get_client()
for prefix in ['ovs_listcache', 'ovs_reverseindex']:
for key in persistent.prefix(prefix):
persistent.delete(key)
for key in persistent.prefix('ovs_data_'):
persistent.set(key, persistent.get(key))
base_reverse_key = 'ovs_reverseindex_{0}_{1}|{2}|{3}'
hybrid_structure = HybridRunner.get_hybrids()
for class_descriptor in hybrid_structure.values():
cls = Descriptor().load(class_descriptor).get_object()
all_objects = DataList(cls, {
'type': DataList.where_operator.AND,
'items': []
})
for item in all_objects:
guid = item.guid
for relation in item._relations:
if relation.foreign_type is None:
rcls = cls
rclsname = rcls.__name__.lower()
else:
rcls = relation.foreign_type
rclsname = rcls.__name__.lower()
key = relation.name
rguid = item._data[key]['guid']
if rguid is not None:
reverse_key = base_reverse_key.format(
rclsname, rguid, relation.foreign_key, guid)
persistent.set(reverse_key, 0)
volatile = VolatileFactory.get_client()
try:
volatile._client.flush_all()
except:
pass
from ovs.dal.lists.vdisklist import VDiskList
for vdisk in VDiskList.get_vdisks():
try:
vdisk.metadata = {
'lba_size': vdisk.info['lba_size'],
'cluster_multiplier': vdisk.info['cluster_multiplier']
}
vdisk.save()
except:
pass
working_version = 10
# Version 11 introduced:
# - ALBA accelerated ALBA, meaning different vpool.metadata information
if working_version < 11:
from ovs.dal.lists.vpoollist import VPoolList
for vpool in VPoolList.get_vpools():
vpool.metadata = {'backend': vpool.metadata}
if 'metadata' in vpool.metadata['backend']:
vpool.metadata['backend'][
'arakoon_config'] = vpool.metadata['backend'].pop(
'metadata')
if 'backend_info' in vpool.metadata['backend']:
vpool.metadata['backend']['backend_info'][
'fragment_cache_on_read'] = True
vpool.metadata['backend']['backend_info'][
'fragment_cache_on_write'] = False
vpool.save()
working_version = 11
return working_version
def update_vmachine_config(vmachine, vm_object, pmachine=None):
"""
Update a vMachine configuration with a given vMachine configuration
"""
try:
vdisks_synced = 0
if vmachine.name is None:
MessageController.fire(
MessageController.Type.EVENT, {
'type': 'vmachine_created',
'metadata': {
'name': vm_object['name']
}
})
elif vmachine.name != vm_object['name']:
MessageController.fire(
MessageController.Type.EVENT, {
'type': 'vmachine_renamed',
'metadata': {
'old_name': vmachine.name,
'new_name': vm_object['name']
}
})
if pmachine is not None:
vmachine.pmachine = pmachine
vmachine.name = vm_object['name']
vmachine.hypervisor_id = vm_object['id']
vmachine.devicename = vm_object['backing']['filename']
vmachine.save()
# Updating and linking disks
storagedrivers = StorageDriverList.get_storagedrivers()
datastores = dict([('{}:{}'.format(storagedriver.storage_ip,
storagedriver.mountpoint),
storagedriver)
for storagedriver in storagedrivers])
vdisk_guids = []
for disk in vm_object['disks']:
if disk['datastore'] in vm_object['datastores']:
datastore = vm_object['datastores'][disk['datastore']]
if datastore in datastores:
vdisk = VDiskList.get_by_devicename_and_vpool(
disk['filename'], datastores[datastore].vpool)
if vdisk is None:
# The disk couldn't be located, but is in our datastore. We might be in a recovery scenario
vdisk = VDisk()
vdisk.vpool = datastores[datastore].vpool
vdisk.reload_client()
vdisk.devicename = disk['filename']
vdisk.volume_id = vdisk.storagedriver_client.get_volume_id(
str(disk['backingfilename']))
vdisk.size = vdisk.info['volume_size']
MDSServiceController.ensure_safety(vdisk)
# Update the disk with information from the hypervisor
if vdisk.vmachine is None:
MessageController.fire(
MessageController.Type.EVENT, {
'type': 'vdisk_attached',
'metadata': {
'vmachine_name': vmachine.name,
'vdisk_name': disk['name']
}
})
vdisk.vmachine = vmachine
vdisk.name = disk['name']
vdisk.order = disk['order']
vdisk.save()
vdisk_guids.append(vdisk.guid)
vdisks_synced += 1
for vdisk in vmachine.vdisks:
if vdisk.guid not in vdisk_guids:
MessageController.fire(
MessageController.Type.EVENT, {
'type': 'vdisk_detached',
'metadata': {
'vmachine_name': vmachine.name,
'vdisk_name': vdisk.name
}
})
vdisk.vmachine = None
vdisk.save()
logger.info(
'Updating vMachine finished (name {}, {} vdisks (re)linked)'.
format(vmachine.name, vdisks_synced))
except Exception as ex:
logger.info('Error during vMachine update: {0}'.format(str(ex)))
raise
def _bootstrap_dal_models(self):
"""
Load/hook dal models as snmp oids
"""
_guids = set()
enabled_key = "{0}_config_dal_enabled".format(STORAGE_PREFIX)
self.instance_oid = 0
try:
enabled = self.persistent.get(enabled_key)
except KeyNotFoundException:
enabled = True # Enabled by default, can be disabled by setting the key
if enabled:
from ovs.dal.lists.vdisklist import VDiskList
from ovs.dal.lists.storagerouterlist import StorageRouterList
from ovs.dal.lists.pmachinelist import PMachineList
from ovs.dal.lists.vmachinelist import VMachineList
from ovs.dal.lists.vpoollist import VPoolList
from ovs.dal.lists.storagedriverlist import StorageDriverList
for storagerouter in StorageRouterList.get_storagerouters():
_guids.add(storagerouter.guid)
if not self._check_added(storagerouter):
self._register_dal_model(10, storagerouter, 'guid', "0")
self._register_dal_model(10, storagerouter, 'name', "1")
self._register_dal_model(10, storagerouter, 'pmachine', "3", key = 'host_status')
self._register_dal_model(10, storagerouter, 'description', "4")
self._register_dal_model(10, storagerouter, 'devicename', "5")
self._register_dal_model(10, storagerouter, 'dtl_mode', "6")
self._register_dal_model(10, storagerouter, 'ip', "8")
self._register_dal_model(10, storagerouter, 'machineid', "9")
self._register_dal_model(10, storagerouter, 'status', "10")
self._register_dal_model(10, storagerouter, '#vdisks', "11",
func = lambda storagerouter: len([vdisk for vpool_vdisks in [storagedriver.vpool.vdisks for storagedriver in storagerouter.storagedrivers] for vdisk in vpool_vdisks if vdisk.storagedriver_id == storagedriver.storagedriver_id]),
atype = int)
self._register_dal_model(10, storagerouter, '#vmachines', "12",
func = lambda storagerouter: len(set([vdisk.vmachine.guid for vpool_vdisks in [storagedriver.vpool.vdisks for storagedriver in storagerouter.storagedrivers] for vdisk in vpool_vdisks if vdisk.storagedriver_id == storagedriver.storagedriver_id])),
atype = int)
self._register_dal_model(10, storagerouter, '#stored_data', "13",
func = lambda storagerouter: sum([vdisk.vmachine.stored_data for vpool_vdisks in [storagedriver.vpool.vdisks for storagedriver in storagerouter.storagedrivers] for vdisk in vpool_vdisks if vdisk.storagedriver_id == storagedriver.storagedriver_id]),
atype = int)
self.instance_oid += 1
for vm in VMachineList.get_vmachines():
_guids.add(vm.guid)
if not self._check_added(vm):
if vm.is_vtemplate:
self._register_dal_model(11, vm, 'guid', "0")
self._register_dal_model(11, vm, 'name', "1")
def _children(vmt):
children = 0
disks = [vd.guid for vd in vmt.vdisks]
for vdisk in [vdisk.parent_vdisk_guid for item in [vm.vdisks for vm in VMachineList.get_vmachines() if not vm.is_vtemplate] for vdisk in item]:
for disk in disks:
if vdisk == disk:
children += 1
return children
self._register_dal_model(11, vm, '#children', 2, func = _children, atype = int)
self.instance_oid += 1
for vm in VMachineList.get_vmachines():
_guids.add(vm.guid)
if not self._check_added(vm):
if not vm.is_vtemplate:
self._register_dal_model(0, vm, 'guid', "0")
self._register_dal_model(0, vm, 'name', "1")
self._register_dal_model(0, vm, 'statistics', "2.0", key = "operations", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.1", key = "cluster_cache_misses_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.2", key = "data_read", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.3", key = "sco_cache_misses", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.4", key = "sco_cache_hits_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.5", key = "sco_cache_hits", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.6", key = "write_operations", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.7", key = "cluster_cache_misses", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.8", key = "read_operations_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.9", key = "sco_cache_misses_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.10", key = "backend_write_operations", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.11", key = "backend_data_read", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.12", key = "cache_hits", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.13", key = "backend_write_operations_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.14", key = "metadata_store_hits_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.15", key = "metadata_store_misses", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.16", key = "backend_data_written", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.17", key = "data_read_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.18", key = "read_operations", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.19", key = "cluster_cache_hits", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.20", key = "data_written_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.21", key = "cluster_cache_hits_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.22", key = "cache_hits_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.23", key = "timestamp", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.24", key = "metadata_store_misses_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.25", key = "backend_data_written_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.26", key = "backend_read_operations", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.27", key = "data_written", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.28", key = "metadata_store_hits", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.29", key = "backend_data_read_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.30", key = "operations_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.31", key = "backend_read_operations_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.32", key = "data_transferred_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.33", key = "write_operations_ps", atype = int)
self._register_dal_model(0, vm, 'statistics', "2.34", key = "data_transferred", atype = int)
self._register_dal_model(0, vm, 'stored_data', "3", atype = int)
self._register_dal_model(0, vm, 'description', "4")
self._register_dal_model(0, vm, 'devicename', "5")
self._register_dal_model(0, vm, 'dtl_mode', "6")
self._register_dal_model(0, vm, 'hypervisorid', "7")
self._register_dal_model(0, vm, 'ip', "8")
self._register_dal_model(0, vm, 'status', "10")
self._register_dal_model(0, vm, 'stored_data', "10", atype = int)
self._register_dal_model(0, vm, 'snapshots', "11", atype = int)
self._register_dal_model(0, vm, 'vdisks', "12", atype = int)
self._register_dal_model(0, vm, 'DTL', '13',
func = lambda vm: 'DEGRADED' if all(item == 'DEGRADED' for item in [vd.info['failover_mode'] for vd in vm.vdisks]) else 'OK')
self.instance_oid += 1
for vd in VDiskList.get_vdisks():
_guids.add(vd.guid)
if not self._check_added(vd):
self._register_dal_model(1, vd, 'guid', "0")
self._register_dal_model(1, vd, 'name', "1")
self._register_dal_model(1, vd, 'statistics', "2.0", key = "operations", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.1", key = "data_written_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.2", key = "data_read", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.3", key = "sco_cache_misses", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.4", key = "sco_cache_hits_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.5", key = "sco_cache_hits", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.6", key = "write_operations", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.7", key = "cluster_cache_misses", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.8", key = "read_operations_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.9", key = "sco_cache_misses_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.10", key = "backend_write_operations", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.11", key = "backend_data_read", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.12", key = "cache_hits", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.13", key = "backend_write_operations_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.14", key = "metadata_store_hits_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.15", key = "metadata_store_misses", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.16", key = "backend_data_written", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.17", key = "data_read_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.18", key = "read_operations", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.19", key = "cluster_cache_hits", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.20", key = "cluster_cache_misses_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.21", key = "cluster_cache_hits_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.22", key = "cache_hits_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.23", key = "timestamp", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.24", key = "metadata_store_misses_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.25", key = "backend_data_written_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.26", key = "backend_read_operations", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.27", key = "data_written", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.28", key = "metadata_store_hits", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.29", key = "backend_data_read_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.30", key = "operations_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.31", key = "backend_read_operations_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.32", key = "data_transferred_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.33", key = "write_operations_ps", atype = int)
self._register_dal_model(1, vd, 'statistics', "2.34", key = "data_transferred", atype = int)
self._register_dal_model(1, vd, 'info', "3", key = 'stored', atype = int)
self._register_dal_model(1, vd, 'info', "4", key = 'failover_mode', atype = int)
self._register_dal_model(1, vd, 'snapshots', "5", atype = int)
self.instance_oid += 1
for pm in PMachineList.get_pmachines():
_guids.add(pm.guid)
if not self._check_added(pm):
self._register_dal_model(2, pm, 'guid', "0")
self._register_dal_model(2, pm, 'name', "1")
self._register_dal_model(2, pm, 'host_status', "2")
self.instance_oid += 1
for vp in VPoolList.get_vpools():
_guids.add(vp.guid)
if not self._check_added(vp):
self._register_dal_model(3, vp, 'guid', "0")
self._register_dal_model(3, vp, 'name', "1")
self._register_dal_model(3, vp, 'statistics', "2.0", key = "operations", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.1", key = "cluster_cache_misses_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.2", key = "data_read", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.3", key = "sco_cache_misses", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.4", key = "sco_cache_hits_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.5", key = "sco_cache_hits", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.6", key = "write_operations", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.7", key = "cluster_cache_misses", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.8", key = "read_operations_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.9", key = "sco_cache_misses_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.10", key = "backend_write_operations", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.11", key = "backend_data_read", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.12", key = "cache_hits", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.13", key = "backend_write_operations_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.14", key = "metadata_store_hits_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.15", key = "metadata_store_misses", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.16", key = "backend_data_written", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.17", key = "data_read_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.18", key = "read_operations", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.19", key = "cluster_cache_hits", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.20", key = "data_written_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.21", key = "cluster_cache_hits_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.22", key = "cache_hits_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.23", key = "timestamp", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.24", key = "metadata_store_misses_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.25", key = "backend_data_written_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.26", key = "backend_read_operations", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.27", key = "data_written", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.28", key = "metadata_store_hits", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.29", key = "backend_data_read_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.30", key = "operations_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.31", key = "backend_read_operations_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.32", key = "data_transferred_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.33", key = "write_operations_ps", atype = int)
self._register_dal_model(3, vp, 'statistics', "2.34", key = "data_transferred", atype = int)
self._register_dal_model(3, vp, 'status', "3")
self._register_dal_model(3, vp, 'description', "4")
self._register_dal_model(3, vp, 'vdisks', "5", atype = int)
self._register_dal_model(3, vp, '#vmachines', "6",
func = lambda vp: len(set([vd.vmachine.guid for vd in vp.vdisks])),
atype = int)
self.instance_oid += 1
for storagedriver in StorageDriverList.get_storagedrivers():
_guids.add(storagedriver.guid)
if not self._check_added(storagedriver):
self._register_dal_model(4, storagedriver, 'guid', "0")
self._register_dal_model(4, storagedriver, 'name', "1")
self._register_dal_model(4, storagedriver, 'stored_data', "2", atype = int)
self.instance_oid += 1
try:
# try to load OVS Backends
from ovs.dal.lists.albabackendlist import AlbaBackendList
for backend in AlbaBackendList.get_albabackends():
_guids.add(backend.guid)
if not self._check_added(backend):
self._register_dal_model(5, backend, 'guid', 0)
self._register_dal_model(5, backend, 'name', 1)
for disk_id in range(len((backend.all_disks))):
self._register_dal_model(5, backend, 'all_disks', '2.{0}.0'.format(disk_id), key = "name", index=disk_id)
self._register_dal_model(5, backend, 'all_disks', '2.{0}.1'.format(disk_id), key = "usage.size", atype = long, index=disk_id)
self._register_dal_model(5, backend, 'all_disks', '2.{0}.2'.format(disk_id), key = "usage.used", atype = long, index=disk_id)
self._register_dal_model(5, backend, 'all_disks', '2.{0}.3'.format(disk_id), key = "usage.available", atype = long, index=disk_id)
self._register_dal_model(5, backend, 'all_disks', '2.{0}.4'.format(disk_id), key = "state.state", index=disk_id)
self._register_dal_model(5, backend, 'all_disks', '2.{0}.5'.format(disk_id), key = "node_id", index=disk_id)
self.instance_oid += 1
except ImportError:
print('OVS Backend not present')
pass
reload = False
for object_guid in list(self.model_oids):
if object_guid not in _guids:
self.model_oids.remove(object_guid)
reload = True
if reload:
self._reload_snmp()
def update_vmachine_config(vmachine, vm_object, pmachine=None):
"""
Update a vMachine configuration with a given vMachine configuration
"""
try:
vdisks_synced = 0
if vmachine.name is None:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vmachine_created',
'metadata': {'name': vm_object['name']}})
elif vmachine.name != vm_object['name']:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vmachine_renamed',
'metadata': {'old_name': vmachine.name,
'new_name': vm_object['name']}})
if pmachine is not None:
vmachine.pmachine = pmachine
vmachine.name = vm_object['name']
vmachine.hypervisor_id = vm_object['id']
vmachine.devicename = vm_object['backing']['filename']
vmachine.save()
# Updating and linking disks
storagedrivers = StorageDriverList.get_storagedrivers()
datastores = dict([('{}:{}'.format(storagedriver.storage_ip, storagedriver.mountpoint), storagedriver) for storagedriver in storagedrivers])
vdisk_guids = []
for disk in vm_object['disks']:
if disk['datastore'] in vm_object['datastores']:
datastore = vm_object['datastores'][disk['datastore']]
if datastore in datastores:
vdisk = VDiskList.get_by_devicename_and_vpool(disk['filename'], datastores[datastore].vpool)
if vdisk is None:
# The disk couldn't be located, but is in our datastore. We might be in a recovery scenario
vdisk = VDisk()
vdisk.vpool = datastores[datastore].vpool
vdisk.reload_client()
vdisk.devicename = disk['filename']
vdisk.volume_id = vdisk.storagedriver_client.get_volume_id(str(disk['backingfilename']))
vdisk.size = vdisk.info['volume_size']
# Update the disk with information from the hypervisor
if vdisk.vmachine is None:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vdisk_attached',
'metadata': {'vmachine_name': vmachine.name,
'vdisk_name': disk['name']}})
vdisk.vmachine = vmachine
vdisk.name = disk['name']
vdisk.order = disk['order']
vdisk.save()
vdisk_guids.append(vdisk.guid)
vdisks_synced += 1
for vdisk in vmachine.vdisks:
if vdisk.guid not in vdisk_guids:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vdisk_detached',
'metadata': {'vmachine_name': vmachine.name,
'vdisk_name': vdisk.name}})
vdisk.vmachine = None
vdisk.save()
logger.info('Updating vMachine finished (name {}, {} vdisks (re)linked)'.format(
vmachine.name, vdisks_synced
))
except Exception as ex:
logger.info('Error during vMachine update: {0}'.format(str(ex)))
raise
def gather_scrub_work():
"""
Retrieve and execute scrub work
:return: None
"""
logger.info('Gather Scrub - Started')
scrub_locations = {}
for storage_driver in StorageDriverList.get_storagedrivers():
for partition in storage_driver.partitions:
if DiskPartition.ROLES.SCRUB == partition.role:
logger.info(
'Gather Scrub - Storage Router {0:<15} has SCRUB partition at {1}'
.format(storage_driver.storagerouter.ip,
partition.path))
if storage_driver.storagerouter not in scrub_locations:
try:
_ = SSHClient(storage_driver.storagerouter)
scrub_locations[
storage_driver.storagerouter] = str(
partition.path)
except UnableToConnectException:
logger.warning(
'Gather Scrub - Storage Router {0:<15} is not reachable'
.format(storage_driver.storagerouter.ip))
if len(scrub_locations) == 0:
raise RuntimeError('No scrub locations found')
vdisk_guids = set()
for vmachine in VMachineList.get_customer_vmachines():
for vdisk in vmachine.vdisks:
if vdisk.info['object_type'] == 'BASE':
vdisk_guids.add(vdisk.guid)
for vdisk in VDiskList.get_without_vmachine():
if vdisk.info['object_type'] == 'BASE':
vdisk_guids.add(vdisk.guid)
logger.info(
'Gather Scrub - Checking {0} volumes for scrub work'.format(
len(vdisk_guids)))
local_machineid = System.get_my_machine_id()
local_storage_router = None
local_scrub_location = None
local_vdisks_to_scrub = []
result_set = ResultSet([])
storage_router_list = []
for index, scrub_info in enumerate(scrub_locations.items()):
start_index = index * len(vdisk_guids) / len(scrub_locations)
end_index = (index + 1) * len(vdisk_guids) / len(scrub_locations)
storage_router = scrub_info[0]
vdisk_guids_to_scrub = list(vdisk_guids)[start_index:end_index]
local = storage_router.machine_id == local_machineid
logger.info(
'Gather Scrub - Storage Router {0:<15} ({1}) - Scrubbing {2} virtual disks'
.format(storage_router.ip,
'local' if local is True else 'remote',
len(vdisk_guids_to_scrub)))
if local is True:
local_storage_router = storage_router
local_scrub_location = scrub_info[1]
local_vdisks_to_scrub = vdisk_guids_to_scrub
else:
result_set.add(
ScheduledTaskController._execute_scrub_work.s(
scrub_location=scrub_info[1],
vdisk_guids=vdisk_guids_to_scrub).apply_async(
routing_key='sr.{0}'.format(
storage_router.machine_id)))
storage_router_list.append(storage_router)
# Remote tasks have been launched, now start the local task and then wait for remote tasks to finish
processed_guids = []
if local_scrub_location is not None and len(local_vdisks_to_scrub) > 0:
try:
processed_guids = ScheduledTaskController._execute_scrub_work(
scrub_location=local_scrub_location,
vdisk_guids=local_vdisks_to_scrub)
except Exception as ex:
logger.error(
'Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'
.format(local_storage_router.ip, ex))
all_results = result_set.join(
propagate=False
) # Propagate False makes sure all jobs are waited for even when 1 or more jobs fail
for index, result in enumerate(all_results):
if isinstance(result, list):
processed_guids.extend(result)
else:
logger.error(
'Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'
.format(storage_router_list[index].ip, result))
if len(processed_guids) != len(vdisk_guids) or set(
processed_guids).difference(vdisk_guids):
raise RuntimeError('Scrubbing failed for 1 or more storagerouters')
logger.info('Gather Scrub - Finished')
def migrate(previous_version):
"""
Migrates from any version to any version, running all migrations required
If previous_version is for example 0 and this script is at
verison 3 it will execute two steps:
- 1 > 2
- 2 > 3
@param previous_version: The previous version from which to start the migration.
"""
working_version = previous_version
# Version 1 introduced:
# - The datastore is still empty, add defaults
if working_version < 1:
from ovs.dal.hybrids.user import User
from ovs.dal.hybrids.group import Group
from ovs.dal.hybrids.role import Role
from ovs.dal.hybrids.client import Client
from ovs.dal.hybrids.j_rolegroup import RoleGroup
from ovs.dal.hybrids.j_roleclient import RoleClient
from ovs.dal.hybrids.backendtype import BackendType
from ovs.dal.hybrids.servicetype import ServiceType
from ovs.dal.hybrids.branding import Branding
from ovs.dal.lists.backendtypelist import BackendTypeList
# Create groups
admin_group = Group()
admin_group.name = 'administrators'
admin_group.description = 'Administrators'
admin_group.save()
viewers_group = Group()
viewers_group.name = 'viewers'
viewers_group.description = 'Viewers'
viewers_group.save()
# Create users
admin = User()
admin.username = '******'
admin.password = hashlib.sha256('admin').hexdigest()
admin.is_active = True
admin.group = admin_group
admin.save()
# Create internal OAuth 2 clients
admin_pw_client = Client()
admin_pw_client.ovs_type = 'INTERNAL'
admin_pw_client.grant_type = 'PASSWORD'
admin_pw_client.user = admin
admin_pw_client.save()
admin_cc_client = Client()
admin_cc_client.ovs_type = 'INTERNAL'
admin_cc_client.grant_type = 'CLIENT_CREDENTIALS'
admin_cc_client.client_secret = ''.join(random.choice(string.ascii_letters +
string.digits +
'|_=+*#@!/-[]{}<>.?,\'";:~')
for _ in range(128))
admin_cc_client.user = admin
admin_cc_client.save()
# Create roles
read_role = Role()
read_role.code = 'read'
read_role.name = 'Read'
read_role.description = 'Can read objects'
read_role.save()
write_role = Role()
write_role.code = 'write'
write_role.name = 'Write'
write_role.description = 'Can write objects'
write_role.save()
manage_role = Role()
manage_role.code = 'manage'
manage_role.name = 'Manage'
manage_role.description = 'Can manage the system'
manage_role.save()
# Attach groups to roles
mapping = [
(admin_group, [read_role, write_role, manage_role]),
(viewers_group, [read_role])
]
for setting in mapping:
for role in setting[1]:
rolegroup = RoleGroup()
rolegroup.group = setting[0]
rolegroup.role = role
rolegroup.save()
for user in setting[0].users:
for role in setting[1]:
for client in user.clients:
roleclient = RoleClient()
roleclient.client = client
roleclient.role = role
roleclient.save()
# Add backends
for backend_type_info in [('Ceph', 'ceph_s3'), ('Amazon', 'amazon_s3'), ('Swift', 'swift_s3'),
('Local', 'local'), ('Distributed', 'distributed'), ('ALBA', 'alba')]:
code = backend_type_info[1]
backend_type = BackendTypeList.get_backend_type_by_code(code)
if backend_type is None:
backend_type = BackendType()
backend_type.name = backend_type_info[0]
backend_type.code = code
backend_type.save()
# Add service types
for service_type_info in ['MetadataServer', 'AlbaProxy', 'Arakoon']:
service_type = ServiceType()
service_type.name = service_type_info
service_type.save()
# Brandings
branding = Branding()
branding.name = 'Default'
branding.description = 'Default bootstrap theme'
branding.css = 'bootstrap-default.min.css'
branding.productname = 'Open vStorage'
branding.is_default = True
branding.save()
slate = Branding()
slate.name = 'Slate'
slate.description = 'Dark bootstrap theme'
slate.css = 'bootstrap-slate.min.css'
slate.productname = 'Open vStorage'
slate.is_default = False
slate.save()
# We're now at version 1
working_version = 1
# Version 2 introduced:
# - new Descriptor format
if working_version < 2:
import imp
from ovs.dal.helpers import Descriptor
from ovs.extensions.storage.persistentfactory import PersistentFactory
client = PersistentFactory.get_client()
keys = client.prefix('ovs_data')
for key in keys:
data = client.get(key)
modified = False
for entry in data.keys():
if isinstance(data[entry], dict) and 'source' in data[entry] and 'hybrids' in data[entry]['source']:
filename = data[entry]['source']
if not filename.startswith('/'):
filename = '/opt/OpenvStorage/ovs/dal/{0}'.format(filename)
module = imp.load_source(data[entry]['name'], filename)
cls = getattr(module, data[entry]['type'])
new_data = Descriptor(cls, cached=False).descriptor
if 'guid' in data[entry]:
new_data['guid'] = data[entry]['guid']
data[entry] = new_data
modified = True
if modified is True:
data['_version'] += 1
client.set(key, data)
# We're now at version 2
working_version = 2
# Version 3 introduced:
# - new Descriptor format
if working_version < 3:
import imp
from ovs.dal.helpers import Descriptor
from ovs.extensions.storage.persistentfactory import PersistentFactory
client = PersistentFactory.get_client()
keys = client.prefix('ovs_data')
for key in keys:
data = client.get(key)
modified = False
for entry in data.keys():
if isinstance(data[entry], dict) and 'source' in data[entry]:
module = imp.load_source(data[entry]['name'], data[entry]['source'])
cls = getattr(module, data[entry]['type'])
new_data = Descriptor(cls, cached=False).descriptor
if 'guid' in data[entry]:
new_data['guid'] = data[entry]['guid']
data[entry] = new_data
modified = True
if modified is True:
data['_version'] += 1
client.set(key, data)
working_version = 3
# Version 4 introduced:
# - Flexible SSD layout
if working_version < 4:
import os
from ovs.dal.lists.storagedriverlist import StorageDriverList
from ovs.dal.hybrids.j_storagedriverpartition import StorageDriverPartition
from ovs.dal.hybrids.diskpartition import DiskPartition
from ovs.dal.lists.servicetypelist import ServiceTypeList
from ovs.extensions.generic.remote import Remote
from ovs.extensions.generic.sshclient import SSHClient
from ovs.extensions.storageserver.storagedriver import StorageDriverConfiguration
for service in ServiceTypeList.get_by_name('MetadataServer').services:
mds_service = service.mds_service
storagedriver = None
for current_storagedriver in service.storagerouter.storagedrivers:
if current_storagedriver.vpool_guid == mds_service.vpool_guid:
storagedriver = current_storagedriver
break
tasks = {}
if storagedriver._data.get('mountpoint_md'):
tasks['{0}/mds_{1}_{2}'.format(storagedriver._data.get('mountpoint_md'),
storagedriver.vpool.name,
mds_service.number)] = (DiskPartition.ROLES.DB, StorageDriverPartition.SUBROLE.MDS)
if storagedriver._data.get('mountpoint_temp'):
tasks['{0}/mds_{1}_{2}'.format(storagedriver._data.get('mountpoint_temp'),
storagedriver.vpool.name,
mds_service.number)] = (DiskPartition.ROLES.SCRUB, StorageDriverPartition.SUBROLE.MDS)
for disk in service.storagerouter.disks:
for partition in disk.partitions:
for directory, (role, subrole) in tasks.iteritems():
with Remote(storagedriver.storagerouter.ip, [os], username='******') as remote:
stat_dir = directory
while not remote.os.path.exists(stat_dir) and stat_dir != '/':
stat_dir = stat_dir.rsplit('/', 1)[0]
if not stat_dir:
stat_dir = '/'
inode = remote.os.stat(stat_dir).st_dev
if partition.inode == inode:
if role not in partition.roles:
partition.roles.append(role)
partition.save()
number = 0
migrated = False
for sd_partition in storagedriver.partitions:
if sd_partition.role == role and sd_partition.sub_role == subrole:
if sd_partition.mds_service == mds_service:
migrated = True
break
if sd_partition.partition_guid == partition.guid:
number = max(sd_partition.number, number)
if migrated is False:
sd_partition = StorageDriverPartition()
sd_partition.role = role
sd_partition.sub_role = subrole
sd_partition.partition = partition
sd_partition.storagedriver = storagedriver
sd_partition.mds_service = mds_service
sd_partition.size = None
sd_partition.number = number + 1
sd_partition.save()
client = SSHClient(storagedriver.storagerouter, username='******')
path = sd_partition.path.rsplit('/', 1)[0]
if path:
client.dir_create(path)
client.dir_chown(path, 'ovs', 'ovs')
client.dir_create(directory)
client.dir_chown(directory, 'ovs', 'ovs')
client.symlink({sd_partition.path: directory})
for storagedriver in StorageDriverList.get_storagedrivers():
migrated_objects = {}
for disk in storagedriver.storagerouter.disks:
for partition in disk.partitions:
# Process all mountpoints that are unique and don't have a specified size
for key, (role, sr_info) in {'mountpoint_md': (DiskPartition.ROLES.DB, {'metadata_{0}': StorageDriverPartition.SUBROLE.MD,
'tlogs_{0}': StorageDriverPartition.SUBROLE.TLOG}),
'mountpoint_fragmentcache': (DiskPartition.ROLES.WRITE, {'fcache_{0}': StorageDriverPartition.SUBROLE.FCACHE}),
'mountpoint_foc': (DiskPartition.ROLES.WRITE, {'fd_{0}': StorageDriverPartition.SUBROLE.FD,
'dtl_{0}': StorageDriverPartition.SUBROLE.DTL}),
'mountpoint_dtl': (DiskPartition.ROLES.WRITE, {'fd_{0}': StorageDriverPartition.SUBROLE.FD,
'dtl_{0}': StorageDriverPartition.SUBROLE.DTL}),
'mountpoint_readcaches': (DiskPartition.ROLES.READ, {'': None}),
'mountpoint_writecaches': (DiskPartition.ROLES.WRITE, {'sco_{0}': StorageDriverPartition.SUBROLE.SCO})}.iteritems():
if key in storagedriver._data:
is_list = isinstance(storagedriver._data[key], list)
entries = storagedriver._data[key][:] if is_list is True else [storagedriver._data[key]]
for entry in entries:
if not entry:
if is_list:
storagedriver._data[key].remove(entry)
if len(storagedriver._data[key]) == 0:
del storagedriver._data[key]
else:
del storagedriver._data[key]
else:
with Remote(storagedriver.storagerouter.ip, [os], username='******') as remote:
inode = remote.os.stat(entry).st_dev
if partition.inode == inode:
if role not in partition.roles:
partition.roles.append(role)
partition.save()
for folder, subrole in sr_info.iteritems():
number = 0
migrated = False
for sd_partition in storagedriver.partitions:
if sd_partition.role == role and sd_partition.sub_role == subrole:
if sd_partition.partition_guid == partition.guid:
number = max(sd_partition.number, number)
if migrated is False:
sd_partition = StorageDriverPartition()
sd_partition.role = role
sd_partition.sub_role = subrole
sd_partition.partition = partition
sd_partition.storagedriver = storagedriver
sd_partition.size = None
sd_partition.number = number + 1
sd_partition.save()
if folder:
source = '{0}/{1}'.format(entry, folder.format(storagedriver.vpool.name))
else:
source = entry
client = SSHClient(storagedriver.storagerouter, username='******')
path = sd_partition.path.rsplit('/', 1)[0]
if path:
client.dir_create(path)
client.dir_chown(path, 'ovs', 'ovs')
client.symlink({sd_partition.path: source})
migrated_objects[source] = sd_partition
if is_list:
storagedriver._data[key].remove(entry)
if len(storagedriver._data[key]) == 0:
del storagedriver._data[key]
else:
del storagedriver._data[key]
storagedriver.save()
if 'mountpoint_bfs' in storagedriver._data:
storagedriver.mountpoint_dfs = storagedriver._data['mountpoint_bfs']
if not storagedriver.mountpoint_dfs:
storagedriver.mountpoint_dfs = None
del storagedriver._data['mountpoint_bfs']
storagedriver.save()
if 'mountpoint_temp' in storagedriver._data:
del storagedriver._data['mountpoint_temp']
storagedriver.save()
if migrated_objects:
print 'Loading sizes'
config = StorageDriverConfiguration('storagedriver', storagedriver.vpool.name)
config.load(SSHClient(storagedriver.storagerouter, username='******'))
for readcache in config.configuration.get('content_addressed_cache', {}).get('clustercache_mount_points', []):
path = readcache.get('path', '').rsplit('/', 1)[0]
size = int(readcache['size'].strip('KiB')) * 1024 if 'size' in readcache else None
if path in migrated_objects:
migrated_objects[path].size = long(size)
migrated_objects[path].save()
for writecache in config.configuration.get('scocache', {}).get('scocache_mount_points', []):
path = writecache.get('path', '')
size = int(writecache['size'].strip('KiB')) * 1024 if 'size' in writecache else None
if path in migrated_objects:
migrated_objects[path].size = long(size)
migrated_objects[path].save()
working_version = 4
return working_version
def gather_scrub_work():
"""
Retrieve and execute scrub work
:return: None
"""
logger.info('Gather Scrub - Started')
scrub_locations = {}
for storage_driver in StorageDriverList.get_storagedrivers():
for partition in storage_driver.partitions:
if DiskPartition.ROLES.SCRUB == partition.role:
logger.info('Gather Scrub - Storage Router {0:<15} has SCRUB partition at {1}'.format(storage_driver.storagerouter.ip, partition.path))
if storage_driver.storagerouter not in scrub_locations:
try:
_ = SSHClient(storage_driver.storagerouter)
scrub_locations[storage_driver.storagerouter] = str(partition.path)
except UnableToConnectException:
logger.warning('Gather Scrub - Storage Router {0:<15} is not reachable'.format(storage_driver.storagerouter.ip))
if len(scrub_locations) == 0:
raise RuntimeError('No scrub locations found')
vdisk_guids = set()
for vmachine in VMachineList.get_customer_vmachines():
for vdisk in vmachine.vdisks:
if vdisk.info['object_type'] == 'BASE':
vdisk_guids.add(vdisk.guid)
for vdisk in VDiskList.get_without_vmachine():
if vdisk.info['object_type'] == 'BASE':
vdisk_guids.add(vdisk.guid)
logger.info('Gather Scrub - Checking {0} volumes for scrub work'.format(len(vdisk_guids)))
local_machineid = System.get_my_machine_id()
local_storage_router = None
local_scrub_location = None
local_vdisks_to_scrub = []
result_set = ResultSet([])
storage_router_list = []
for index, scrub_info in enumerate(scrub_locations.items()):
start_index = index * len(vdisk_guids) / len(scrub_locations)
end_index = (index + 1) * len(vdisk_guids) / len(scrub_locations)
storage_router = scrub_info[0]
vdisk_guids_to_scrub = list(vdisk_guids)[start_index:end_index]
local = storage_router.machine_id == local_machineid
logger.info('Gather Scrub - Storage Router {0:<15} ({1}) - Scrubbing {2} virtual disks'.format(storage_router.ip, 'local' if local is True else 'remote', len(vdisk_guids_to_scrub)))
if local is True:
local_storage_router = storage_router
local_scrub_location = scrub_info[1]
local_vdisks_to_scrub = vdisk_guids_to_scrub
else:
result_set.add(ScheduledTaskController._execute_scrub_work.s(scrub_location=scrub_info[1],
vdisk_guids=vdisk_guids_to_scrub).apply_async(
routing_key='sr.{0}'.format(storage_router.machine_id)
))
storage_router_list.append(storage_router)
# Remote tasks have been launched, now start the local task and then wait for remote tasks to finish
processed_guids = []
if local_scrub_location is not None and len(local_vdisks_to_scrub) > 0:
try:
processed_guids = ScheduledTaskController._execute_scrub_work(scrub_location=local_scrub_location,
vdisk_guids=local_vdisks_to_scrub)
except Exception as ex:
logger.error('Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'.format(local_storage_router.ip, ex))
all_results = result_set.join(propagate=False) # Propagate False makes sure all jobs are waited for even when 1 or more jobs fail
for index, result in enumerate(all_results):
if isinstance(result, list):
processed_guids.extend(result)
else:
logger.error('Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'.format(storage_router_list[index].ip, result))
if len(processed_guids) != len(vdisk_guids) or set(processed_guids).difference(vdisk_guids):
raise RuntimeError('Scrubbing failed for 1 or more storagerouters')
logger.info('Gather Scrub - Finished')
def migrate(previous_version):
"""
Migrates from any version to any version, running all migrations required
If previous_version is for example 0 and this script is at
version 3 it will execute two steps:
- 1 > 2
- 2 > 3
@param previous_version: The previous version from which to start the migration.
"""
working_version = previous_version
# Version 1 introduced:
# - The datastore is still empty, add defaults
if working_version < 1:
from ovs.dal.hybrids.user import User
from ovs.dal.hybrids.group import Group
from ovs.dal.hybrids.role import Role
from ovs.dal.hybrids.client import Client
from ovs.dal.hybrids.failuredomain import FailureDomain
from ovs.dal.hybrids.j_rolegroup import RoleGroup
from ovs.dal.hybrids.j_roleclient import RoleClient
from ovs.dal.hybrids.backendtype import BackendType
from ovs.dal.hybrids.servicetype import ServiceType
from ovs.dal.hybrids.branding import Branding
from ovs.dal.lists.backendtypelist import BackendTypeList
# Create groups
admin_group = Group()
admin_group.name = 'administrators'
admin_group.description = 'Administrators'
admin_group.save()
viewers_group = Group()
viewers_group.name = 'viewers'
viewers_group.description = 'Viewers'
viewers_group.save()
# Create users
admin = User()
admin.username = '******'
admin.password = hashlib.sha256('admin').hexdigest()
admin.is_active = True
admin.group = admin_group
admin.save()
# Create internal OAuth 2 clients
admin_pw_client = Client()
admin_pw_client.ovs_type = 'INTERNAL'
admin_pw_client.grant_type = 'PASSWORD'
admin_pw_client.user = admin
admin_pw_client.save()
admin_cc_client = Client()
admin_cc_client.ovs_type = 'INTERNAL'
admin_cc_client.grant_type = 'CLIENT_CREDENTIALS'
admin_cc_client.client_secret = ''.join(random.choice(string.ascii_letters +
string.digits +
'|_=+*#@!/-[]{}<>.?,\'";:~')
for _ in range(128))
admin_cc_client.user = admin
admin_cc_client.save()
# Create roles
read_role = Role()
read_role.code = 'read'
read_role.name = 'Read'
read_role.description = 'Can read objects'
read_role.save()
write_role = Role()
write_role.code = 'write'
write_role.name = 'Write'
write_role.description = 'Can write objects'
write_role.save()
manage_role = Role()
manage_role.code = 'manage'
manage_role.name = 'Manage'
manage_role.description = 'Can manage the system'
manage_role.save()
# Attach groups to roles
mapping = [
(admin_group, [read_role, write_role, manage_role]),
(viewers_group, [read_role])
]
for setting in mapping:
for role in setting[1]:
rolegroup = RoleGroup()
rolegroup.group = setting[0]
rolegroup.role = role
rolegroup.save()
for user in setting[0].users:
for role in setting[1]:
for client in user.clients:
roleclient = RoleClient()
roleclient.client = client
roleclient.role = role
roleclient.save()
# Add backends
for backend_type_info in [('Ceph', 'ceph_s3'), ('Amazon', 'amazon_s3'), ('Swift', 'swift_s3'),
('Local', 'local'), ('Distributed', 'distributed'), ('ALBA', 'alba')]:
code = backend_type_info[1]
backend_type = BackendTypeList.get_backend_type_by_code(code)
if backend_type is None:
backend_type = BackendType()
backend_type.name = backend_type_info[0]
backend_type.code = code
backend_type.save()
# Add service types
for service_type_info in [ServiceType.SERVICE_TYPES.MD_SERVER, ServiceType.SERVICE_TYPES.ALBA_PROXY, ServiceType.SERVICE_TYPES.ARAKOON]:
service_type = ServiceType()
service_type.name = service_type_info
service_type.save()
# Branding
branding = Branding()
branding.name = 'Default'
branding.description = 'Default bootstrap theme'
branding.css = 'bootstrap-default.min.css'
branding.productname = 'Open vStorage'
branding.is_default = True
branding.save()
slate = Branding()
slate.name = 'Slate'
slate.description = 'Dark bootstrap theme'
slate.css = 'bootstrap-slate.min.css'
slate.productname = 'Open vStorage'
slate.is_default = False
slate.save()
# Failure Domain
failure_domain = FailureDomain()
failure_domain.name = 'Default'
failure_domain.save()
# We're now at version 1
working_version = 1
# Version 2 introduced:
# - new Descriptor format
if working_version < 2:
import imp
from ovs.dal.helpers import Descriptor
from ovs.extensions.storage.persistentfactory import PersistentFactory
client = PersistentFactory.get_client()
keys = client.prefix('ovs_data')
for key in keys:
data = client.get(key)
modified = False
for entry in data.keys():
if isinstance(data[entry], dict) and 'source' in data[entry] and 'hybrids' in data[entry]['source']:
filename = data[entry]['source']
if not filename.startswith('/'):
filename = '/opt/OpenvStorage/ovs/dal/{0}'.format(filename)
module = imp.load_source(data[entry]['name'], filename)
cls = getattr(module, data[entry]['type'])
new_data = Descriptor(cls, cached=False).descriptor
if 'guid' in data[entry]:
new_data['guid'] = data[entry]['guid']
data[entry] = new_data
modified = True
if modified is True:
data['_version'] += 1
client.set(key, data)
# We're now at version 2
working_version = 2
# Version 3 introduced:
# - new Descriptor format
if working_version < 3:
import imp
from ovs.dal.helpers import Descriptor
from ovs.extensions.storage.persistentfactory import PersistentFactory
client = PersistentFactory.get_client()
keys = client.prefix('ovs_data')
for key in keys:
data = client.get(key)
modified = False
for entry in data.keys():
if isinstance(data[entry], dict) and 'source' in data[entry]:
module = imp.load_source(data[entry]['name'], data[entry]['source'])
cls = getattr(module, data[entry]['type'])
new_data = Descriptor(cls, cached=False).descriptor
if 'guid' in data[entry]:
new_data['guid'] = data[entry]['guid']
data[entry] = new_data
modified = True
if modified is True:
data['_version'] += 1
client.set(key, data)
working_version = 3
# Version 4 introduced:
# - Flexible SSD layout
if working_version < 4:
import os
from ovs.dal.hybrids.diskpartition import DiskPartition
from ovs.dal.hybrids.j_storagedriverpartition import StorageDriverPartition
from ovs.dal.hybrids.servicetype import ServiceType
from ovs.dal.lists.servicetypelist import ServiceTypeList
from ovs.dal.lists.storagedriverlist import StorageDriverList
from ovs.extensions.generic.remote import remote
from ovs.extensions.generic.sshclient import SSHClient
from ovs.extensions.storageserver.storagedriver import StorageDriverConfiguration
for service in ServiceTypeList.get_by_name(ServiceType.SERVICE_TYPES.MD_SERVER).services:
mds_service = service.mds_service
storagedriver = None
for current_storagedriver in service.storagerouter.storagedrivers:
if current_storagedriver.vpool_guid == mds_service.vpool_guid:
storagedriver = current_storagedriver
break
tasks = {}
if storagedriver._data.get('mountpoint_md'):
tasks['{0}/mds_{1}_{2}'.format(storagedriver._data.get('mountpoint_md'),
storagedriver.vpool.name,
mds_service.number)] = (DiskPartition.ROLES.DB, StorageDriverPartition.SUBROLE.MDS)
if storagedriver._data.get('mountpoint_temp'):
tasks['{0}/mds_{1}_{2}'.format(storagedriver._data.get('mountpoint_temp'),
storagedriver.vpool.name,
mds_service.number)] = (DiskPartition.ROLES.SCRUB, StorageDriverPartition.SUBROLE.MDS)
for disk in service.storagerouter.disks:
for partition in disk.partitions:
for directory, (role, subrole) in tasks.iteritems():
with remote(storagedriver.storagerouter.ip, [os], username='******') as rem:
stat_dir = directory
while not rem.os.path.exists(stat_dir) and stat_dir != '/':
stat_dir = stat_dir.rsplit('/', 1)[0]
if not stat_dir:
stat_dir = '/'
inode = rem.os.stat(stat_dir).st_dev
if partition.inode == inode:
if role not in partition.roles:
partition.roles.append(role)
partition.save()
number = 0
migrated = False
for sd_partition in storagedriver.partitions:
if sd_partition.role == role and sd_partition.sub_role == subrole:
if sd_partition.mds_service == mds_service:
migrated = True
break
if sd_partition.partition_guid == partition.guid:
number = max(sd_partition.number, number)
if migrated is False:
sd_partition = StorageDriverPartition()
sd_partition.role = role
sd_partition.sub_role = subrole
sd_partition.partition = partition
sd_partition.storagedriver = storagedriver
sd_partition.mds_service = mds_service
sd_partition.size = None
sd_partition.number = number + 1
sd_partition.save()
client = SSHClient(storagedriver.storagerouter, username='******')
path = sd_partition.path.rsplit('/', 1)[0]
if path:
client.dir_create(path)
client.dir_chown(path, 'ovs', 'ovs')
client.dir_create(directory)
client.dir_chown(directory, 'ovs', 'ovs')
client.symlink({sd_partition.path: directory})
for storagedriver in StorageDriverList.get_storagedrivers():
migrated_objects = {}
for disk in storagedriver.storagerouter.disks:
for partition in disk.partitions:
# Process all mountpoints that are unique and don't have a specified size
for key, (role, sr_info) in {'mountpoint_md': (DiskPartition.ROLES.DB, {'metadata_{0}': StorageDriverPartition.SUBROLE.MD,
'tlogs_{0}': StorageDriverPartition.SUBROLE.TLOG}),
'mountpoint_fragmentcache': (DiskPartition.ROLES.WRITE, {'fcache_{0}': StorageDriverPartition.SUBROLE.FCACHE}),
'mountpoint_foc': (DiskPartition.ROLES.WRITE, {'fd_{0}': StorageDriverPartition.SUBROLE.FD,
'dtl_{0}': StorageDriverPartition.SUBROLE.DTL}),
'mountpoint_dtl': (DiskPartition.ROLES.WRITE, {'fd_{0}': StorageDriverPartition.SUBROLE.FD,
'dtl_{0}': StorageDriverPartition.SUBROLE.DTL}),
'mountpoint_readcaches': (DiskPartition.ROLES.READ, {'': None}),
'mountpoint_writecaches': (DiskPartition.ROLES.WRITE, {'sco_{0}': StorageDriverPartition.SUBROLE.SCO})}.iteritems():
if key in storagedriver._data:
is_list = isinstance(storagedriver._data[key], list)
entries = storagedriver._data[key][:] if is_list is True else [storagedriver._data[key]]
for entry in entries:
if not entry:
if is_list:
storagedriver._data[key].remove(entry)
if len(storagedriver._data[key]) == 0:
del storagedriver._data[key]
else:
del storagedriver._data[key]
else:
with remote(storagedriver.storagerouter.ip, [os], username='******') as rem:
inode = rem.os.stat(entry).st_dev
if partition.inode == inode:
if role not in partition.roles:
partition.roles.append(role)
partition.save()
for folder, subrole in sr_info.iteritems():
number = 0
migrated = False
for sd_partition in storagedriver.partitions:
if sd_partition.role == role and sd_partition.sub_role == subrole:
if sd_partition.partition_guid == partition.guid:
number = max(sd_partition.number, number)
if migrated is False:
sd_partition = StorageDriverPartition()
sd_partition.role = role
sd_partition.sub_role = subrole
sd_partition.partition = partition
sd_partition.storagedriver = storagedriver
sd_partition.size = None
sd_partition.number = number + 1
sd_partition.save()
if folder:
source = '{0}/{1}'.format(entry, folder.format(storagedriver.vpool.name))
else:
source = entry
client = SSHClient(storagedriver.storagerouter, username='******')
path = sd_partition.path.rsplit('/', 1)[0]
if path:
client.dir_create(path)
client.dir_chown(path, 'ovs', 'ovs')
client.symlink({sd_partition.path: source})
migrated_objects[source] = sd_partition
if is_list:
storagedriver._data[key].remove(entry)
if len(storagedriver._data[key]) == 0:
del storagedriver._data[key]
else:
del storagedriver._data[key]
storagedriver.save()
if 'mountpoint_bfs' in storagedriver._data:
storagedriver.mountpoint_dfs = storagedriver._data['mountpoint_bfs']
if not storagedriver.mountpoint_dfs:
storagedriver.mountpoint_dfs = None
del storagedriver._data['mountpoint_bfs']
storagedriver.save()
if 'mountpoint_temp' in storagedriver._data:
del storagedriver._data['mountpoint_temp']
storagedriver.save()
if migrated_objects:
print 'Loading sizes'
config = StorageDriverConfiguration('storagedriver', storagedriver.vpool_guid, storagedriver.storagedriver_id)
config.load()
for readcache in config.configuration.get('content_addressed_cache', {}).get('clustercache_mount_points', []):
path = readcache.get('path', '').rsplit('/', 1)[0]
size = int(readcache['size'].strip('KiB')) * 1024 if 'size' in readcache else None
if path in migrated_objects:
migrated_objects[path].size = long(size)
migrated_objects[path].save()
for writecache in config.configuration.get('scocache', {}).get('scocache_mount_points', []):
path = writecache.get('path', '')
size = int(writecache['size'].strip('KiB')) * 1024 if 'size' in writecache else None
if path in migrated_objects:
migrated_objects[path].size = long(size)
migrated_objects[path].save()
working_version = 4
# Version 5 introduced:
# - Failure Domains
if working_version < 5:
import os
from ovs.dal.hybrids.failuredomain import FailureDomain
from ovs.dal.lists.failuredomainlist import FailureDomainList
from ovs.dal.lists.storagerouterlist import StorageRouterList
from ovs.extensions.generic.remote import remote
from ovs.extensions.generic.sshclient import SSHClient
failure_domains = FailureDomainList.get_failure_domains()
if len(failure_domains) > 0:
failure_domain = failure_domains[0]
else:
failure_domain = FailureDomain()
failure_domain.name = 'Default'
failure_domain.save()
for storagerouter in StorageRouterList.get_storagerouters():
change = False
if storagerouter.primary_failure_domain is None:
storagerouter.primary_failure_domain = failure_domain
change = True
if storagerouter.rdma_capable is None:
client = SSHClient(storagerouter, username='******')
rdma_capable = False
with remote(client.ip, [os], username='******') as rem:
for root, dirs, files in rem.os.walk('/sys/class/infiniband'):
for directory in dirs:
ports_dir = '/'.join([root, directory, 'ports'])
if not rem.os.path.exists(ports_dir):
continue
for sub_root, sub_dirs, _ in rem.os.walk(ports_dir):
if sub_root != ports_dir:
continue
for sub_directory in sub_dirs:
state_file = '/'.join([sub_root, sub_directory, 'state'])
if rem.os.path.exists(state_file):
if 'ACTIVE' in client.run('cat {0}'.format(state_file)):
rdma_capable = True
storagerouter.rdma_capable = rdma_capable
change = True
if change is True:
storagerouter.save()
working_version = 5
# Version 6 introduced:
# - Distributed scrubbing
if working_version < 6:
from ovs.dal.hybrids.diskpartition import DiskPartition
from ovs.dal.lists.storagedriverlist import StorageDriverList
from ovs.extensions.generic.sshclient import SSHClient
for storage_driver in StorageDriverList.get_storagedrivers():
root_client = SSHClient(storage_driver.storagerouter, username='******')
for partition in storage_driver.partitions:
if partition.role == DiskPartition.ROLES.SCRUB:
old_path = partition.path
partition.sub_role = None
partition.save()
partition.invalidate_dynamics(['folder', 'path'])
if root_client.dir_exists(partition.path):
continue # New directory already exists
if '_mds_' in old_path:
if root_client.dir_exists(old_path):
root_client.symlink({partition.path: old_path})
if not root_client.dir_exists(partition.path):
root_client.dir_create(partition.path)
root_client.dir_chmod(partition.path, 0777)
working_version = 6
# Version 7 introduced:
# - vPool status
if working_version < 7:
from ovs.dal.hybrids import vpool
reload(vpool)
from ovs.dal.hybrids.vpool import VPool
from ovs.dal.lists.vpoollist import VPoolList
for _vpool in VPoolList.get_vpools():
vpool = VPool(_vpool.guid)
if hasattr(vpool, 'status') and vpool.status is None:
vpool.status = VPool.STATUSES.RUNNING
vpool.save()
working_version = 7
# Version 10 introduced:
# - Reverse indexes are stored in persistent store
# - Store more non-changing metadata on disk iso using a dynamic property
if working_version < 10:
from ovs.dal.helpers import HybridRunner, Descriptor
from ovs.dal.datalist import DataList
from ovs.extensions.storage.persistentfactory import PersistentFactory
from ovs.extensions.storage.volatilefactory import VolatileFactory
persistent = PersistentFactory.get_client()
for prefix in ['ovs_listcache', 'ovs_reverseindex']:
for key in persistent.prefix(prefix):
persistent.delete(key)
for key in persistent.prefix('ovs_data_'):
persistent.set(key, persistent.get(key))
base_reverse_key = 'ovs_reverseindex_{0}_{1}|{2}|{3}'
hybrid_structure = HybridRunner.get_hybrids()
for class_descriptor in hybrid_structure.values():
cls = Descriptor().load(class_descriptor).get_object()
all_objects = DataList(cls, {'type': DataList.where_operator.AND,
'items': []})
for item in all_objects:
guid = item.guid
for relation in item._relations:
if relation.foreign_type is None:
rcls = cls
rclsname = rcls.__name__.lower()
else:
rcls = relation.foreign_type
rclsname = rcls.__name__.lower()
key = relation.name
rguid = item._data[key]['guid']
if rguid is not None:
reverse_key = base_reverse_key.format(rclsname, rguid, relation.foreign_key, guid)
persistent.set(reverse_key, 0)
volatile = VolatileFactory.get_client()
try:
volatile._client.flush_all()
except:
pass
from ovs.dal.lists.vdisklist import VDiskList
for vdisk in VDiskList.get_vdisks():
try:
vdisk.metadata = {'lba_size': vdisk.info['lba_size'],
'cluster_multiplier': vdisk.info['cluster_multiplier']}
vdisk.save()
except:
pass
working_version = 10
# Version 11 introduced:
# - ALBA accelerated ALBA, meaning different vpool.metadata information
if working_version < 11:
from ovs.dal.lists.vpoollist import VPoolList
for vpool in VPoolList.get_vpools():
vpool.metadata = {'backend': vpool.metadata}
if 'metadata' in vpool.metadata['backend']:
vpool.metadata['backend']['arakoon_config'] = vpool.metadata['backend'].pop('metadata')
if 'backend_info' in vpool.metadata['backend']:
vpool.metadata['backend']['backend_info']['fragment_cache_on_read'] = True
vpool.metadata['backend']['backend_info']['fragment_cache_on_write'] = False
vpool.save()
working_version = 11
return working_version
def update_vmachine_config(vmachine, vm_object, pmachine=None):
"""
Update a vMachine configuration with a given vMachine configuration
:param vmachine: Virtual Machine to update
:param vm_object: New virtual machine info
:param pmachine: Physical machine of the virtual machine
"""
try:
vdisks_synced = 0
if vmachine.name is None:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vmachine_created',
'metadata': {'name': vm_object['name']}})
elif vmachine.name != vm_object['name']:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vmachine_renamed',
'metadata': {'old_name': vmachine.name,
'new_name': vm_object['name']}})
if pmachine is not None:
vmachine.pmachine = pmachine
vmachine.name = vm_object['name']
vmachine.hypervisor_id = vm_object['id']
vmachine.devicename = vm_object['backing']['filename']
vmachine.save()
# Updating and linking disks
storagedrivers = StorageDriverList.get_storagedrivers()
datastores = dict([('{0}:{1}'.format(storagedriver.storage_ip, storagedriver.mountpoint), storagedriver) for storagedriver in storagedrivers])
vdisk_guids = []
mutex = volatile_mutex('{0}_{1}'.format(vmachine.name, vmachine.devicename))
for disk in vm_object['disks']:
ensure_safety = False
if disk['datastore'] in vm_object['datastores']:
datastore = vm_object['datastores'][disk['datastore']]
if datastore in datastores:
try:
mutex.acquire(wait=10)
vdisk = VDiskList.get_by_devicename_and_vpool(disk['filename'], datastores[datastore].vpool)
if vdisk is None:
# The disk couldn't be located, but is in our datastore. We might be in a recovery scenario
vdisk = VDisk()
vdisk.vpool = datastores[datastore].vpool
vdisk.reload_client()
vdisk.devicename = disk['filename']
vdisk.volume_id = vdisk.storagedriver_client.get_volume_id(str(disk['backingfilename']))
vdisk.size = vdisk.info['volume_size']
vdisk.metadata = {'lba_size': vdisk.info['lba_size'],
'cluster_multiplier': vdisk.info['cluster_multiplier']}
# Create the disk in a locked context, but don't execute long running-task in same context
vdisk.save()
ensure_safety = True
finally:
mutex.release()
if ensure_safety:
MDSServiceController.ensure_safety(vdisk)
VDiskController.dtl_checkup(vdisk_guid=vdisk.guid)
# Update the disk with information from the hypervisor
if vdisk.vmachine is None:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vdisk_attached',
'metadata': {'vmachine_name': vmachine.name,
'vdisk_name': disk['name']}})
vdisk.vmachine = vmachine
vdisk.name = disk['name']
vdisk.order = disk['order']
vdisk.save()
vdisk_guids.append(vdisk.guid)
vdisks_synced += 1
for vdisk in vmachine.vdisks:
if vdisk.guid not in vdisk_guids:
MessageController.fire(MessageController.Type.EVENT,
{'type': 'vdisk_detached',
'metadata': {'vmachine_name': vmachine.name,
'vdisk_name': vdisk.name}})
vdisk.vmachine = None
vdisk.save()
VMachineController._logger.info('Updating vMachine finished (name {0}, {1} vdisks (re)linked)'.format(
vmachine.name, vdisks_synced
))
except Exception as ex:
VMachineController._logger.info('Error during vMachine update: {0}'.format(str(ex)))
raise
def verify_vdisk_cache_quota():
"""
Validates whether the caching quota is reaching its limits or has surpassed it
Each vDisk can consume a part of the total fragment caching capacity
"""
MonitoringController._logger.info(
'Starting vDisk caching quota verification...')
alba_guid_size_map = {}
for storagedriver in StorageDriverList.get_storagedrivers():
storagedriver.invalidate_dynamics(
['vpool_backend_info', 'vdisks_guids'])
for cache_type in [[
'cache_quota_fc', 'backend_info', 'connection_info',
'fragment'
],
[
'cache_quota_bc',
'block_cache_backend_info',
'block_cache_connection_info', 'block'
]]:
cache_quota = storagedriver.vpool_backend_info[cache_type[0]]
backend_info = storagedriver.vpool_backend_info[cache_type[1]]
connection_info = storagedriver.vpool_backend_info[
cache_type[2]]
if backend_info is None or connection_info is None:
continue
alba_backend_name = backend_info['name']
alba_backend_host = connection_info['host']
alba_backend_guid = backend_info['alba_backend_guid']
if alba_backend_guid not in alba_guid_size_map:
ovs_client = OVSClient(
ip=alba_backend_host,
port=connection_info['port'],
credentials=(connection_info['client_id'],
connection_info['client_secret']),
version=2,
cache_store=VolatileFactory.get_client())
try:
alba_guid_size_map[alba_backend_guid] = {
'name':
alba_backend_name,
'backend_ip':
alba_backend_host,
'total_size':
ovs_client.get('/alba/backends/{0}/'.format(
alba_backend_guid),
params={'contents': 'usages'
})['usages']['size'],
'requested_size':
0
}
except Exception:
MonitoringController._logger.exception(
'Failed to retrieve ALBA Backend info for {0} on host {1}'
.format(alba_backend_name, alba_backend_host))
continue
for vdisk_guid in storagedriver.vdisks_guids:
vdisk = VDisk(vdisk_guid)
vdisk_cq = vdisk.cache_quota.get(
cache_type[3]
) if vdisk.cache_quota is not None else None
if vdisk_cq is None:
alba_guid_size_map[alba_backend_guid][
'requested_size'] += cache_quota if cache_quota is not None else 0
else:
alba_guid_size_map[alba_backend_guid][
'requested_size'] += vdisk_cq
local_ips = [sr.ip for sr in StorageRouterList.get_storagerouters()]
for alba_backend_info in alba_guid_size_map.itervalues():
name = alba_backend_info['name']
backend_ip = alba_backend_info['backend_ip']
location = 'local'
remote_msg = ''
if backend_ip not in local_ips:
location = 'remote'
remote_msg = ' (on remote IP {0})'.format(backend_ip)
percentage = alba_backend_info[
'requested_size'] / alba_backend_info['total_size'] * 100
if percentage > 100:
MonitoringController._logger.error(
'OVS_WARNING: Over-allocation for vDisk caching quota on {0} ALBA Backend {1}{2}. Unexpected behavior might occur'
.format(location, name, remote_msg))
elif percentage > 70:
MonitoringController._logger.warning(
'OVS_WARNING: vDisk caching quota on {0} ALBA Backend {1} is at {2:.1f}%{3}'
.format(location, name, percentage, remote_msg))
MonitoringController._logger.info(
'Finished vDisk cache quota verification')
def gather_scrub_work():
"""
Retrieve and execute scrub work
:return: None
"""
ScheduledTaskController._logger.info('Gather Scrub - Started')
scrub_locations = {}
for storage_driver in StorageDriverList.get_storagedrivers():
for partition in storage_driver.partitions:
if DiskPartition.ROLES.SCRUB == partition.role:
ScheduledTaskController._logger.info('Gather Scrub - Storage Router {0:<15} has SCRUB partition at {1}'.format(storage_driver.storagerouter.ip, partition.path))
if storage_driver.storagerouter not in scrub_locations:
try:
sshclient = SSHClient(storage_driver.storagerouter)
# Use ServiceManager(sshclient) to make sure ovs-workers are actually running
if ServiceManager.get_service_status('workers', sshclient) is False:
ScheduledTaskController._logger.warning('Gather Scrub - Storage Router {0:<15} - workers are not running'.format(storage_driver.storagerouter.ip))
else:
scrub_locations[storage_driver.storagerouter] = str(partition.path)
except UnableToConnectException:
ScheduledTaskController._logger.warning('Gather Scrub - Storage Router {0:<15} is not reachable'.format(storage_driver.storagerouter.ip))
if len(scrub_locations) == 0:
raise RuntimeError('No scrub locations found')
vdisk_guids = set()
for vmachine in VMachineList.get_customer_vmachines():
for vdisk in vmachine.vdisks:
if vdisk.info['object_type'] == 'BASE':
vdisk_guids.add(vdisk.guid)
for vdisk in VDiskList.get_without_vmachine():
if vdisk.info['object_type'] == 'BASE':
vdisk_guids.add(vdisk.guid)
if len(vdisk_guids) == 0:
ScheduledTaskController._logger.info('Gather Scrub - No scrub work needed'.format(len(vdisk_guids)))
return
ScheduledTaskController._logger.info('Gather Scrub - Checking {0} volumes for scrub work'.format(len(vdisk_guids)))
local_machineid = System.get_my_machine_id()
local_storage_router = None
local_scrub_location = None
local_vdisks_to_scrub = []
result_set = {}
storage_router_list = []
scrub_map = {}
for index, scrub_info in enumerate(scrub_locations.items()):
start_index = index * len(vdisk_guids) / len(scrub_locations)
end_index = (index + 1) * len(vdisk_guids) / len(scrub_locations)
storage_router = scrub_info[0]
vdisk_guids_to_scrub = list(vdisk_guids)[start_index:end_index]
local = storage_router.machine_id == local_machineid
ScheduledTaskController._logger.info('Gather Scrub - Storage Router {0:<15} ({1}) - Scrubbing {2} virtual disks'.format(storage_router.ip, 'local' if local is True else 'remote', len(vdisk_guids_to_scrub)))
if local is True:
local_storage_router = storage_router
local_scrub_location = scrub_info[1]
local_vdisks_to_scrub = vdisk_guids_to_scrub
else:
result_set[storage_router.ip] = ScheduledTaskController._execute_scrub_work.s(scrub_location=scrub_info[1],
vdisk_guids=vdisk_guids_to_scrub).apply_async(routing_key='sr.{0}'.format(storage_router.machine_id))
storage_router_list.append(storage_router)
scrub_map[storage_router.ip] = vdisk_guids_to_scrub
# Remote tasks have been launched, now start the local task and then wait for remote tasks to finish
processed_guids = []
if local_scrub_location is not None and len(local_vdisks_to_scrub) > 0:
try:
processed_guids = ScheduledTaskController._execute_scrub_work(scrub_location=local_scrub_location,
vdisk_guids=local_vdisks_to_scrub)
except Exception as ex:
ScheduledTaskController._logger.error('Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'.format(local_storage_router.ip, ex))
all_results, failed_nodes = CeleryToolbox.manage_running_tasks(result_set,
timesleep=60) # Check every 60 seconds if tasks are still running
for ip, result in all_results.iteritems():
if isinstance(result, list):
processed_guids.extend(result)
else:
ScheduledTaskController._logger.error('Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'.format(ip, result))
result_set = {}
for failed_node in failed_nodes:
ScheduledTaskController._logger.warning('Scrubbing failed on node {0}. Will reschedule on another node.'.format(failed_node))
vdisk_guids_to_scrub = scrub_map[failed_node]
rescheduled_work = False
for storage_router, scrub_location in scrub_locations.items():
if storage_router.ip not in failed_nodes:
if storage_router.machine_id != local_machineid:
ScheduledTaskController._logger.info('Rescheduled scrub work from node {0} to node {1}.'.format(failed_node, storage_router.ip))
result_set[storage_router.ip] = ScheduledTaskController._execute_scrub_work.s(scrub_location=scrub_location,
vdisk_guids=vdisk_guids_to_scrub).apply_async(
routing_key='sr.{0}'.format(storage_router.machine_id))
storage_router_list.append(storage_router)
rescheduled_work = True
break
if rescheduled_work is False:
if local_scrub_location is not None:
try:
processed_guids.extend(ScheduledTaskController._execute_scrub_work(scrub_location=local_scrub_location,
vdisk_guids=vdisk_guids_to_scrub))
except Exception as ex:
ScheduledTaskController._logger.error(
'Gather Scrub - Storage Router Local - Scrubbing failed with error:\n - {0}'.format(ex))
else:
ScheduledTaskController._logger.warning('No nodes left to reschedule work from node {0}'.format(failed_node))
if len(result_set) > 0:
all_results2, failed_nodes = CeleryToolbox.manage_running_tasks(result_set,
timesleep=60) # Check every 60 seconds if tasks are still running
for ip, result in all_results2.iteritems():
if isinstance(result, list):
processed_guids.extend(result)
else:
ScheduledTaskController._logger.error('Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'.format(ip, result))
if len(set(processed_guids)) != len(vdisk_guids) or set(processed_guids).difference(vdisk_guids):
raise RuntimeError('Scrubbing failed for 1 or more storagerouters')
ScheduledTaskController._logger.info('Gather Scrub - Finished')
def list(self):
"""
Overview of all StorageDrivers
"""
return StorageDriverList.get_storagedrivers()
def add_vpool(parameters):
"""
Add a vPool to the machine this task is running on
"""
parameters = {} if parameters is None else parameters
ip = parameters['storagerouter_ip']
vpool_name = parameters['vpool_name']
if StorageRouterController._validate_ip(ip) is False:
raise ValueError('The entered ip address is invalid')
if not re.match('^[0-9a-z]+(\-+[0-9a-z]+)*$', vpool_name):
raise ValueError('Invalid vpool_name given')
client = SSHClient.load(ip) # Make sure to ALWAYS reload the client, as Fabric seems to be singleton-ish
unique_id = System.get_my_machine_id(client)
storagerouter = None
for current_storagerouter in StorageRouterList.get_storagerouters():
if current_storagerouter.ip == ip and current_storagerouter.machine_id == unique_id:
storagerouter = current_storagerouter
break
if storagerouter is None:
raise RuntimeError('Could not find Storage Router with given ip address')
vpool = VPoolList.get_vpool_by_name(vpool_name)
storagedriver = None
if vpool is not None:
if vpool.backend_type.code == 'local':
# Might be an issue, investigating whether it's on the same not or not
if len(vpool.storagedrivers) == 1 and vpool.storagedrivers[0].storagerouter.machine_id != unique_id:
raise RuntimeError('A local vPool with name {0} already exists'.format(vpool_name))
for vpool_storagedriver in vpool.storagedrivers:
if vpool_storagedriver.storagerouter_guid == storagerouter.guid:
storagedriver = vpool_storagedriver # The vPool is already added to this Storage Router and this might be a cleanup/recovery
# Check whether there are running machines on this vPool
machine_guids = []
for vdisk in vpool.vdisks:
if vdisk.vmachine_guid not in machine_guids:
machine_guids.append(vdisk.vmachine_guid)
if vdisk.vmachine.hypervisor_status in ['RUNNING', 'PAUSED']:
raise RuntimeError(
'At least one vMachine using this vPool is still running or paused. Make sure there are no active vMachines'
)
nodes = {ip} # Set comprehension
if vpool is not None:
for vpool_storagedriver in vpool.storagedrivers:
nodes.add(vpool_storagedriver.storagerouter.ip)
nodes = list(nodes)
services = ['volumedriver_{0}'.format(vpool_name),
'failovercache_{0}'.format(vpool_name)]
# Stop services
for node in nodes:
node_client = SSHClient.load(node)
for service in services:
System.exec_remote_python(node_client, """
from ovs.plugin.provider.service import Service
if Service.has_service('{0}'):
Service.disable_service('{0}')
""".format(service))
System.exec_remote_python(node_client, """
from ovs.plugin.provider.service import Service
if Service.has_service('{0}'):
Service.stop_service('{0}')
""".format(service))
# Keep in mind that if the Storage Driver exists, the vPool does as well
client = SSHClient.load(ip)
mountpoint_bfs = ''
directories_to_create = []
if vpool is None:
vpool = VPool()
supported_backends = System.read_remote_config(client, 'volumedriver.supported.backends').split(',')
if 'rest' in supported_backends:
supported_backends.remove('rest') # REST is not supported for now
backend_type = BackendTypeList.get_backend_type_by_code(parameters['type'])
vpool.backend_type = backend_type
connection_host = connection_port = connection_username = connection_password = None
if vpool.backend_type.code in ['local', 'distributed']:
vpool.metadata = {'backend_type': 'LOCAL'}
mountpoint_bfs = parameters['mountpoint_bfs']
directories_to_create.append(mountpoint_bfs)
vpool.metadata['local_connection_path'] = mountpoint_bfs
if vpool.backend_type.code == 'rest':
connection_host = parameters['connection_host']
connection_port = parameters['connection_port']
rest_connection_timeout_secs = parameters['connection_timeout']
vpool.metadata = {'rest_connection_host': connection_host,
'rest_connection_port': connection_port,
'buchla_connection_log_level': "0",
'rest_connection_verbose_logging': rest_connection_timeout_secs,
'rest_connection_metadata_format': "JSON",
'backend_type': 'REST'}
elif vpool.backend_type.code in ('ceph_s3', 'amazon_s3', 'swift_s3'):
connection_host = parameters['connection_host']
connection_port = parameters['connection_port']
connection_username = parameters['connection_username']
connection_password = parameters['connection_password']
if vpool.backend_type.code in ['swift_s3']:
strict_consistency = 'false'
s3_connection_flavour = 'SWIFT'
else:
strict_consistency = 'true'
s3_connection_flavour = 'S3'
vpool.metadata = {'s3_connection_host': connection_host,
's3_connection_port': connection_port,
's3_connection_username': connection_username,
's3_connection_password': connection_password,
's3_connection_flavour': s3_connection_flavour,
's3_connection_strict_consistency': strict_consistency,
's3_connection_verbose_logging': 1,
'backend_type': 'S3'}
vpool.name = vpool_name
vpool.description = "{} {}".format(vpool.backend_type.code, vpool_name)
vpool.login = connection_username
vpool.password = connection_password
if not connection_host:
vpool.connection = None
else:
vpool.connection = '{}:{}'.format(connection_host, connection_port)
vpool.save()
# Connection information is Storage Driver related information
new_storagedriver = False
if storagedriver is None:
storagedriver = StorageDriver()
new_storagedriver = True
mountpoint_temp = parameters['mountpoint_temp']
mountpoint_md = parameters['mountpoint_md']
mountpoint_readcache1 = parameters['mountpoint_readcache1']
mountpoint_readcache2 = parameters.get('mountpoint_readcache2', '')
mountpoint_writecache = parameters['mountpoint_writecache']
mountpoint_foc = parameters['mountpoint_foc']
directories_to_create.append(mountpoint_temp)
directories_to_create.append(mountpoint_md)
directories_to_create.append(mountpoint_readcache1)
if mountpoint_readcache2:
directories_to_create.append(mountpoint_readcache2)
directories_to_create.append(mountpoint_writecache)
directories_to_create.append(mountpoint_foc)
client = SSHClient.load(ip)
dir_create_script = """
import os
for directory in {0}:
if not os.path.exists(directory):
os.makedirs(directory)
""".format(directories_to_create)
System.exec_remote_python(client, dir_create_script)
read_cache1_fs = os.statvfs(mountpoint_readcache1)
read_cache2_fs = None
if mountpoint_readcache2:
read_cache2_fs = os.statvfs(mountpoint_readcache2)
write_cache_fs = os.statvfs(mountpoint_writecache)
fdcache = '{}/fd_{}'.format(mountpoint_writecache, vpool_name)
scocache = '{}/sco_{}'.format(mountpoint_writecache, vpool_name)
readcache1 = '{}/read1_{}'.format(mountpoint_readcache1, vpool_name)
files2create = [readcache1]
if mountpoint_readcache2 and mountpoint_readcache1 != mountpoint_readcache2:
readcache2 = '{}/read2_{}'.format(mountpoint_readcache2, vpool_name)
files2create.append(readcache2)
else:
readcache2 = ''
failovercache = '{}/foc_{}'.format(mountpoint_foc, vpool_name)
metadatapath = '{}/metadata_{}'.format(mountpoint_md, vpool_name)
tlogpath = '{}/tlogs_{}'.format(mountpoint_md, vpool_name)
rsppath = '/var/rsp/{}'.format(vpool_name)
dirs2create = [scocache, failovercache, metadatapath, tlogpath, rsppath,
System.read_remote_config(client, 'volumedriver.readcache.serialization.path')]
cmd = "cat /etc/mtab | grep ^/dev/ | cut -d ' ' -f 2"
mountpoints = [device.strip() for device in client.run(cmd).strip().split('\n')]
mountpoints.remove('/')
def is_partition(directory):
for mountpoint in mountpoints:
if directory == mountpoint:
return True
return False
# Cache sizes
# 20% = scocache
# 20% = failovercache (@TODO: check if this can possibly consume more than 20%)
# 60% = readcache
# safety values:
readcache1_factor = 0.2
readcache2_factor = 0.2
writecache_factor = 0.1
if (mountpoint_readcache1 == mountpoint_readcache2) or not mountpoint_readcache2:
delta = set()
delta.add(mountpoint_readcache1 if is_partition(mountpoint_readcache1) else '/dummy')
delta.add(mountpoint_writecache if is_partition(mountpoint_writecache) else '/dummy')
delta.add(mountpoint_foc if is_partition(mountpoint_foc) else '/dummy')
if len(delta) == 1:
readcache1_factor = 0.49
writecache_factor = 0.2
elif len(delta) == 2:
if mountpoint_writecache == mountpoint_foc:
readcache1_factor = 0.98
writecache_factor = 0.49
else:
readcache1_factor = 0.49
if mountpoint_readcache1 == mountpoint_writecache:
writecache_factor = 0.49
else:
writecache_factor = 0.98
elif len(delta) == 3:
readcache1_factor = 0.98
writecache_factor = 0.98
else:
delta = set()
delta.add(mountpoint_readcache1 if is_partition(mountpoint_readcache1) else '/dummy')
delta.add(mountpoint_readcache2 if is_partition(mountpoint_readcache2) else '/dummy')
delta.add(mountpoint_writecache if is_partition(mountpoint_writecache) else '/dummy')
delta.add(mountpoint_foc if is_partition(mountpoint_foc) else '/dummy')
if len(delta) == 1:
# consider them all to be directories
readcache1_factor = 0.24
readcache2_factor = 0.24
writecache_factor = 0.24
elif len(delta) == 2:
if mountpoint_writecache == mountpoint_foc:
writecache_factor = 0.24
if mountpoint_readcache1 == mountpoint_writecache:
readcache1_factor = 0.49
readcache2_factor = 0.98
else:
readcache1_factor = 0.98
readcache2_factor = 0.49
else:
readcache1_factor = readcache2_factor = 0.49
writecache_factor = 0.49
elif len(delta) == 3:
if mountpoint_writecache == mountpoint_foc:
readcache1_factor = 0.98
readcache2_factor = 0.98
writecache_factor = 0.49
elif mountpoint_readcache1 == mountpoint_writecache:
readcache1_factor = 0.49
readcache2_factor = 0.98
writecache_factor = 0.49
elif mountpoint_readcache1 == mountpoint_foc:
readcache1_factor = 0.49
readcache2_factor = 0.98
writecache_factor = 0.98
elif mountpoint_readcache2 == mountpoint_writecache:
readcache1_factor = 0.98
readcache2_factor = 0.49
writecache_factor = 0.49
elif mountpoint_readcache2 == mountpoint_foc:
readcache1_factor = 0.98
readcache2_factor = 0.49
writecache_factor = 0.98
elif len(delta) == 4:
readcache1_factor = 0.98
readcache2_factor = 0.98
writecache_factor = 0.98
# summarize caching on root partition (directory only)
root_assigned = dict()
if not is_partition(mountpoint_readcache1):
root_assigned['readcache1_factor'] = readcache1_factor
if not is_partition(mountpoint_readcache2):
root_assigned['readcache2_factor'] = readcache2_factor
if not is_partition(mountpoint_writecache):
root_assigned['writecache_factor'] = writecache_factor
if not is_partition(mountpoint_foc):
root_assigned['foc_factor'] = min(readcache1_factor, readcache2_factor, writecache_factor)
# always leave at least 20% of free space
division_factor = 1.0
total_size = sum(root_assigned.values()) + .02 * len(root_assigned)
if 0.8 < total_size < 1.6:
division_factor = 2.0
elif 1.6 < total_size < 3.2:
division_factor = 4.0
elif total_size >= 3.2:
division_factor = 8.0
if 'readcache1_factor' in root_assigned.keys():
readcache1_factor /= division_factor
if 'readcache2_factor' in root_assigned.keys():
readcache2_factor /= division_factor
if 'writecache_factor' in root_assigned.keys():
writecache_factor /= division_factor
scocache_size = '{0}KiB'.format((int(write_cache_fs.f_bavail * writecache_factor / 4096) * 4096) * 4)
if (mountpoint_readcache1 and not mountpoint_readcache2) or (mountpoint_readcache1 == mountpoint_readcache2):
mountpoint_readcache2 = ''
readcache1_size = '{0}KiB'.format((int(read_cache1_fs.f_bavail * readcache1_factor / 4096) * 4096) * 4)
readcache2 = ''
readcache2_size = '0KiB'
else:
readcache1_size = '{0}KiB'.format((int(read_cache1_fs.f_bavail * readcache1_factor / 4096) * 4096) * 4)
readcache2_size = '{0}KiB'.format((int(read_cache2_fs.f_bavail * readcache2_factor / 4096) * 4096) * 4)
if new_storagedriver:
ports_in_use = System.ports_in_use(client)
ports_reserved = []
ports_in_use_model = {}
for port_storagedriver in StorageDriverList.get_storagedrivers():
if port_storagedriver.vpool_guid not in ports_in_use_model:
ports_in_use_model[port_storagedriver.vpool_guid] = port_storagedriver.ports
ports_reserved += port_storagedriver.ports
if vpool.guid in ports_in_use_model: # The vPool is extended to another StorageRouter. We need to use these ports.
ports = ports_in_use_model[vpool.guid]
if any(port in ports_in_use for port in ports):
raise RuntimeError('The required ports are in use')
else: # First StorageDriver for this vPool, so generating new ports
ports = []
for port_range in System.read_remote_config(client, 'volumedriver.filesystem.ports').split(','):
port_range = port_range.strip()
if '-' in port_range:
current_range = (int(port_range.split('-')[0]), int(port_range.split('-')[1]))
else:
current_range = (int(port_range), 65536)
current_port = current_range[0]
while len(ports) < 3:
if current_port not in ports_in_use and current_port not in ports_reserved:
ports.append(current_port)
current_port += 1
if current_port > current_range[1]:
break
if len(ports) != 3:
raise RuntimeError('Could not find enough free ports')
else:
ports = storagedriver.ports
ip_path = Configuration.get('ovs.core.ip.path')
if ip_path is None:
ip_path = "`which ip`"
cmd = "{0} a | grep 'inet ' | sed 's/\s\s*/ /g' | cut -d ' ' -f 3 | cut -d '/' -f 1".format(ip_path)
ipaddresses = client.run(cmd).strip().split('\n')
ipaddresses = [ipaddr.strip() for ipaddr in ipaddresses]
grid_ip = System.read_remote_config(client, 'ovs.grid.ip')
if grid_ip in ipaddresses:
ipaddresses.remove(grid_ip)
if not ipaddresses:
raise RuntimeError('No available ip addresses found suitable for Storage Router storage ip')
if storagerouter.pmachine.hvtype == 'KVM':
volumedriver_storageip = '127.0.0.1'
else:
volumedriver_storageip = parameters['storage_ip']
vrouter_id = '{0}{1}'.format(vpool_name, unique_id)
vrouter_config = {'vrouter_id': vrouter_id,
'vrouter_redirect_timeout_ms': '5000',
'vrouter_routing_retries': 10,
'vrouter_volume_read_threshold': 1024,
'vrouter_volume_write_threshold': 1024,
'vrouter_file_read_threshold': 1024,
'vrouter_file_write_threshold': 1024,
'vrouter_min_workers': 4,
'vrouter_max_workers': 16}
voldrv_arakoon_cluster_id = str(System.read_remote_config(client, 'volumedriver.arakoon.clusterid'))
voldrv_arakoon_cluster = ArakoonManagementEx().getCluster(voldrv_arakoon_cluster_id)
voldrv_arakoon_client_config = voldrv_arakoon_cluster.getClientConfig()
arakoon_node_configs = []
for arakoon_node in voldrv_arakoon_client_config.keys():
arakoon_node_configs.append(ArakoonNodeConfig(arakoon_node,
voldrv_arakoon_client_config[arakoon_node][0][0],
voldrv_arakoon_client_config[arakoon_node][1]))
vrouter_clusterregistry = ClusterRegistry(str(vpool.guid), voldrv_arakoon_cluster_id, arakoon_node_configs)
node_configs = []
for existing_storagedriver in StorageDriverList.get_storagedrivers():
if existing_storagedriver.vpool_guid == vpool.guid:
node_configs.append(ClusterNodeConfig(str(existing_storagedriver.storagedriver_id),
str(existing_storagedriver.cluster_ip),
existing_storagedriver.ports[0],
existing_storagedriver.ports[1],
existing_storagedriver.ports[2]))
if new_storagedriver:
node_configs.append(ClusterNodeConfig(vrouter_id, grid_ip, ports[0], ports[1], ports[2]))
vrouter_clusterregistry.set_node_configs(node_configs)
readcaches = [{'path': readcache1, 'size': readcache1_size}]
if readcache2:
readcaches.append({'path': readcache2, 'size': readcache2_size})
scocaches = [{'path': scocache, 'size': scocache_size}]
filesystem_config = {'fs_backend_path': mountpoint_bfs}
volumemanager_config = {'metadata_path': metadatapath, 'tlog_path': tlogpath}
storagedriver_config_script = """
from ovs.plugin.provider.configuration import Configuration
from ovs.extensions.storageserver.storagedriver import StorageDriverConfiguration
fd_config = {{'fd_cache_path': '{11}',
'fd_extent_cache_capacity': '1024',
'fd_namespace' : 'fd-{0}-{12}'}}
storagedriver_configuration = StorageDriverConfiguration('{0}')
storagedriver_configuration.configure_backend({1})
storagedriver_configuration.configure_readcache({2}, Configuration.get('volumedriver.readcache.serialization.path') + '/{0}')
storagedriver_configuration.configure_scocache({3}, '1GB', '2GB')
storagedriver_configuration.configure_failovercache('{4}')
storagedriver_configuration.configure_filesystem({5})
storagedriver_configuration.configure_volumemanager({6})
storagedriver_configuration.configure_volumerouter('{12}', {7})
storagedriver_configuration.configure_arakoon_cluster('{8}', {9})
storagedriver_configuration.configure_hypervisor('{10}')
storagedriver_configuration.configure_filedriver(fd_config)
""".format(vpool_name, vpool.metadata, readcaches, scocaches, failovercache, filesystem_config,
volumemanager_config, vrouter_config, voldrv_arakoon_cluster_id, voldrv_arakoon_client_config,
storagerouter.pmachine.hvtype, fdcache, vpool.guid)
System.exec_remote_python(client, storagedriver_config_script)
remote_script = """
import os
from configobj import ConfigObj
from ovs.plugin.provider.configuration import Configuration
protocol = Configuration.get('ovs.core.broker.protocol')
login = Configuration.get('ovs.core.broker.login')
password = Configuration.get('ovs.core.broker.password')
vpool_name = {0}
uris = []
cfg = ConfigObj('/opt/OpenvStorage/config/rabbitmqclient.cfg')
main_section = cfg.get('main')
nodes = main_section['nodes'] if type(main_section['nodes']) == list else [main_section['nodes']]
for node in nodes:
uris.append({{'amqp_uri': '{{0}}://{{1}}:{{2}}@{{3}}'.format(protocol, login, password, cfg.get(node)['location'])}})
from ovs.extensions.storageserver.storagedriver import StorageDriverConfiguration
queue_config = {{'events_amqp_routing_key': Configuration.get('ovs.core.broker.volumerouter.queue'),
'events_amqp_uris': uris}}
for config_file in os.listdir('/opt/OpenvStorage/config/voldrv_vpools'):
this_vpool_name = config_file.replace('.json', '')
if config_file.endswith('.json') and (vpool_name is None or vpool_name == this_vpool_name):
storagedriver_configuration = StorageDriverConfiguration(this_vpool_name)
storagedriver_configuration.configure_event_publisher(queue_config)
""".format(vpool_name if vpool_name is None else "'{0}'".format(vpool_name))
System.exec_remote_python(client, remote_script)
# Updating the model
storagedriver.storagedriver_id = vrouter_id
storagedriver.name = vrouter_id.replace('_', ' ')
storagedriver.description = storagedriver.name
storagedriver.storage_ip = volumedriver_storageip
storagedriver.cluster_ip = grid_ip
storagedriver.ports = ports
storagedriver.mountpoint = '/mnt/{0}'.format(vpool_name)
storagedriver.mountpoint_temp = mountpoint_temp
storagedriver.mountpoint_readcache1 = mountpoint_readcache1
storagedriver.mountpoint_readcache2 = mountpoint_readcache2
storagedriver.mountpoint_writecache = mountpoint_writecache
storagedriver.mountpoint_foc = mountpoint_foc
storagedriver.mountpoint_bfs = mountpoint_bfs
storagedriver.mountpoint_md = mountpoint_md
storagedriver.storagerouter = storagerouter
storagedriver.vpool = vpool
storagedriver.save()
dirs2create.append(storagedriver.mountpoint)
dirs2create.append(mountpoint_writecache + '/' + '/fd_' + vpool_name)
dirs2create.append('{0}/fd_{1}'.format(mountpoint_writecache, vpool_name))
file_create_script = """
import os
for directory in {0}:
if not os.path.exists(directory):
os.makedirs(directory)
for filename in {1}:
if not os.path.exists(filename):
open(filename, 'a').close()
""".format(dirs2create, files2create)
System.exec_remote_python(client, file_create_script)
voldrv_config_file = '{0}/voldrv_vpools/{1}.json'.format(System.read_remote_config(client, 'ovs.core.cfgdir'),
vpool_name)
log_file = '/var/log/ovs/volumedriver/{0}.log'.format(vpool_name)
vd_cmd = '/usr/bin/volumedriver_fs -f --config-file={0} --mountpoint {1} --logrotation --logfile {2} -o big_writes -o sync_read -o allow_other'.format(
voldrv_config_file, storagedriver.mountpoint, log_file)
if storagerouter.pmachine.hvtype == 'KVM':
vd_stopcmd = 'umount {0}'.format(storagedriver.mountpoint)
else:
vd_stopcmd = 'exportfs -u *:{0}; umount {0}'.format(storagedriver.mountpoint)
vd_name = 'volumedriver_{}'.format(vpool_name)
log_file = '/var/log/ovs/volumedriver/foc_{0}.log'.format(vpool_name)
fc_cmd = '/usr/bin/failovercachehelper --config-file={0} --logfile={1}'.format(voldrv_config_file, log_file)
fc_name = 'failovercache_{0}'.format(vpool_name)
params = {'<VPOOL_MOUNTPOINT>': storagedriver.mountpoint,
'<HYPERVISOR_TYPE>': storagerouter.pmachine.hvtype,
'<VPOOL_NAME>': vpool_name,
'<UUID>': str(uuid.uuid4())}
if Osdist.is_ubuntu(client):
if client.file_exists('/opt/OpenvStorage/config/templates/upstart/ovs-volumedriver.conf'):
client.run('cp -f /opt/OpenvStorage/config/templates/upstart/ovs-volumedriver.conf /opt/OpenvStorage/config/templates/upstart/ovs-volumedriver_{0}.conf'.format(vpool_name))
client.run('cp -f /opt/OpenvStorage/config/templates/upstart/ovs-failovercache.conf /opt/OpenvStorage/config/templates/upstart/ovs-failovercache_{0}.conf'.format(vpool_name))
else:
if client.file_exists('/opt/OpenvStorage/config/templates/systemd/ovs-volumedriver.service'):
client.run('cp -f /opt/OpenvStorage/config/templates/systemd/ovs-volumedriver.service /opt/OpenvStorage/config/templates/systemd/ovs-volumedriver_{0}.service'.format(vpool_name))
client.run('cp -f /opt/OpenvStorage/config/templates/systemd/ovs-failovercache.service /opt/OpenvStorage/config/templates/systemd/ovs-failovercache_{0}.service'.format(vpool_name))
service_script = """
from ovs.plugin.provider.service import Service
Service.add_service(package=('openvstorage', 'volumedriver'), name='{0}', command='{1}', stop_command='{2}', params={5})
Service.add_service(package=('openvstorage', 'failovercache'), name='{3}', command='{4}', stop_command=None, params={5})
""".format(
vd_name, vd_cmd, vd_stopcmd,
fc_name, fc_cmd, params
)
System.exec_remote_python(client, service_script)
if storagerouter.pmachine.hvtype == 'VMWARE':
client.run("grep -q '/tmp localhost(ro,no_subtree_check)' /etc/exports || echo '/tmp localhost(ro,no_subtree_check)' >> /etc/exports")
if Osdist.is_ubuntu(client):
client.run('service nfs-kernel-server start')
else:
client.run('service nfs start')
if storagerouter.pmachine.hvtype == 'KVM':
client.run('virsh pool-define-as {0} dir - - - - {1}'.format(vpool_name, storagedriver.mountpoint))
client.run('virsh pool-build {0}'.format(vpool_name))
client.run('virsh pool-start {0}'.format(vpool_name))
client.run('virsh pool-autostart {0}'.format(vpool_name))
# Start services
for node in nodes:
node_client = SSHClient.load(node)
for service in services:
System.exec_remote_python(node_client, """
from ovs.plugin.provider.service import Service
Service.enable_service('{0}')
""".format(service))
System.exec_remote_python(node_client, """
from ovs.plugin.provider.service import Service
Service.start_service('{0}')
""".format(service))
# Fill vPool size
vfs_info = os.statvfs('/mnt/{0}'.format(vpool_name))
vpool.size = vfs_info.f_blocks * vfs_info.f_bsize
vpool.save()
# Configure Cinder
ovsdb = PersistentFactory.get_client()
vpool_config_key = str('ovs_openstack_cinder_%s' % storagedriver.vpool_guid)
if ovsdb.exists(vpool_config_key):
# Second node gets values saved by first node
cinder_password, cinder_user, tenant_name, controller_ip, config_cinder = ovsdb.get(vpool_config_key)
else:
config_cinder = parameters.get('config_cinder', False)
cinder_password = ''
cinder_user = ''
tenant_name = ''
controller_ip = ''
if config_cinder:
cinder_password = parameters.get('cinder_pass', cinder_password)
cinder_user = parameters.get('cinder_user', cinder_user)
tenant_name = parameters.get('cinder_tenant', tenant_name)
controller_ip = parameters.get('cinder_controller', controller_ip) # Keystone host
if cinder_password:
osc = OpenStackCinder(cinder_password = cinder_password,
cinder_user = cinder_user,
tenant_name = tenant_name,
controller_ip = controller_ip)
osc.configure_vpool(vpool_name, storagedriver.mountpoint)
# Save values for first node to use
ovsdb.set(vpool_config_key,
[cinder_password, cinder_user, tenant_name, controller_ip, config_cinder])
def migrate():
"""
Executes async migrations. It doesn't matter too much when they are executed, as long as they get eventually
executed. This code will typically contain:
* "dangerous" migration code (it needs certain running services)
* Migration code depending on a cluster-wide state
* ...
* Successfully finishing a piece of migration code, should create an entry in /ovs/framework/migration in case it should not be executed again
* Eg: /ovs/framework/migration|stats_monkey_integration: True
"""
MigrationController._logger.info('Preparing out of band migrations...')
from ovs.dal.lists.servicetypelist import ServiceTypeList
from ovs.dal.lists.storagedriverlist import StorageDriverList
from ovs.dal.lists.storagerouterlist import StorageRouterList
from ovs.dal.lists.vpoollist import VPoolList
from ovs.extensions.db.arakooninstaller import ArakoonInstaller
from ovs.extensions.generic.configuration import Configuration
from ovs.extensions.generic.sshclient import SSHClient
from ovs_extensions.generic.toolbox import ExtensionsToolbox
from ovs.extensions.migration.migration.ovsmigrator import ExtensionMigrator
from ovs.extensions.packages.packagefactory import PackageFactory
from ovs_extensions.services.interfaces.systemd import Systemd
from ovs.extensions.services.servicefactory import ServiceFactory
from ovs.extensions.storageserver.storagedriver import StorageDriverConfiguration
from ovs.lib.helpers.storagedriver.installer import StorageDriverInstaller
MigrationController._logger.info('Start out of band migrations...')
service_manager = ServiceFactory.get_manager()
sr_client_map = {}
for storagerouter in StorageRouterList.get_storagerouters():
sr_client_map[storagerouter.guid] = SSHClient(endpoint=storagerouter.ip, # Is triggered during post-update code too during which the ovs-watcher-framework service is still down and thus not refreshing the heartbeat --> use IP i/o StorageRouter
username='******')
#########################################################
# Addition of 'ExecReload' for AlbaProxy SystemD services
if ServiceFactory.get_service_type() == 'systemd':
changed_clients = set()
for storagedriver in StorageDriverList.get_storagedrivers():
root_client = sr_client_map[storagedriver.storagerouter_guid]
for alba_proxy in storagedriver.alba_proxies:
service = alba_proxy.service
service_name = 'ovs-{0}'.format(service.name)
if not service_manager.has_service(name=service_name, client=root_client):
continue
if 'ExecReload=' in root_client.file_read(filename='/lib/systemd/system/{0}.service'.format(service_name)):
continue
try:
service_manager.regenerate_service(name=StorageDriverInstaller.SERVICE_TEMPLATE_PROXY, client=root_client, target_name=service_name)
changed_clients.add(root_client)
except:
MigrationController._logger.exception('Error rebuilding service {0}'.format(service_name))
for root_client in changed_clients:
root_client.run(['systemctl', 'daemon-reload'])
##################################################################
# Adjustment of open file descriptors for Arakoon services to 8192
changed_clients = set()
for storagerouter in StorageRouterList.get_storagerouters():
root_client = sr_client_map[storagerouter.guid]
for service_name in service_manager.list_services(client=root_client):
if not service_name.startswith('ovs-arakoon-'):
continue
if ServiceFactory.get_service_type() == 'systemd':
path = '/lib/systemd/system/{0}.service'.format(service_name)
check = 'LimitNOFILE=8192'
else:
path = '/etc/init/{0}.conf'.format(service_name)
check = 'limit nofile 8192 8192'
if not root_client.file_exists(path):
continue
if check in root_client.file_read(path):
continue
try:
service_manager.regenerate_service(name='ovs-arakoon', client=root_client, target_name=service_name)
changed_clients.add(root_client)
ExtensionsToolbox.edit_version_file(client=root_client,
package_name='arakoon',
old_run_file='{0}/{1}.version'.format(ServiceFactory.RUN_FILE_DIR, service_name))
except:
MigrationController._logger.exception('Error rebuilding service {0}'.format(service_name))
for root_client in changed_clients:
root_client.run(['systemctl', 'daemon-reload'])
#############################
# Migrate to multiple proxies
for storagedriver in StorageDriverList.get_storagedrivers():
vpool = storagedriver.vpool
root_client = sr_client_map[storagedriver.storagerouter_guid]
for alba_proxy in storagedriver.alba_proxies:
# Rename alba_proxy service in model
service = alba_proxy.service
old_service_name = 'albaproxy_{0}'.format(vpool.name)
new_service_name = 'albaproxy_{0}_0'.format(vpool.name)
if old_service_name != service.name:
continue
service.name = new_service_name
service.save()
if not service_manager.has_service(name=old_service_name, client=root_client):
continue
old_configuration_key = '/ovs/framework/hosts/{0}/services/{1}'.format(storagedriver.storagerouter.machine_id, old_service_name)
if not Configuration.exists(key=old_configuration_key):
continue
# Add '-reboot' to alba_proxy services (because of newly created services and removal of old service)
ExtensionsToolbox.edit_version_file(client=root_client,
package_name='alba',
old_run_file='{0}/{1}.version'.format(ServiceFactory.RUN_FILE_DIR, old_service_name),
new_run_file='{0}/{1}.version'.format(ServiceFactory.RUN_FILE_DIR, new_service_name))
# Register new service and remove old service
service_manager.add_service(name=StorageDriverInstaller.SERVICE_TEMPLATE_PROXY,
client=root_client,
params=Configuration.get(old_configuration_key),
target_name='ovs-{0}'.format(new_service_name))
# Update scrub proxy config
proxy_config_key = '/ovs/vpools/{0}/proxies/{1}/config/main'.format(vpool.guid, alba_proxy.guid)
proxy_config = None if Configuration.exists(key=proxy_config_key) is False else Configuration.get(proxy_config_key)
if proxy_config is not None:
fragment_cache = proxy_config.get(StorageDriverConfiguration.CACHE_FRAGMENT, ['none', {}])
if fragment_cache[0] == 'alba' and fragment_cache[1].get('cache_on_write') is True: # Accelerated ALBA configured
fragment_cache_scrub_info = copy.deepcopy(fragment_cache)
fragment_cache_scrub_info[1]['cache_on_read'] = False
proxy_scrub_config_key = '/ovs/vpools/{0}/proxies/scrub/generic_scrub'.format(vpool.guid)
proxy_scrub_config = None if Configuration.exists(key=proxy_scrub_config_key) is False else Configuration.get(proxy_scrub_config_key)
if proxy_scrub_config is not None and proxy_scrub_config[StorageDriverConfiguration.CACHE_FRAGMENT] == ['none']:
proxy_scrub_config[StorageDriverConfiguration.CACHE_FRAGMENT] = fragment_cache_scrub_info
Configuration.set(key=proxy_scrub_config_key, value=proxy_scrub_config)
# Update 'backend_connection_manager' section
changes = False
storagedriver_config = StorageDriverConfiguration(vpool.guid, storagedriver.storagedriver_id)
if 'backend_connection_manager' not in storagedriver_config.configuration:
continue
current_config = storagedriver_config.configuration['backend_connection_manager']
if current_config.get('backend_type') != 'MULTI':
changes = True
backend_connection_manager = {'backend_type': 'MULTI'}
for index, proxy in enumerate(sorted(storagedriver.alba_proxies, key=lambda pr: pr.service.ports[0])):
backend_connection_manager[str(index)] = copy.deepcopy(current_config)
# noinspection PyUnresolvedReferences
backend_connection_manager[str(index)]['alba_connection_use_rora'] = True
# noinspection PyUnresolvedReferences
backend_connection_manager[str(index)]['alba_connection_rora_manifest_cache_capacity'] = 5000
# noinspection PyUnresolvedReferences
for key, value in backend_connection_manager[str(index)].items():
if key.startswith('backend_interface'):
backend_connection_manager[key] = value
# noinspection PyUnresolvedReferences
del backend_connection_manager[str(index)][key]
for key, value in {'backend_interface_retries_on_error': 5,
'backend_interface_retry_interval_secs': 1,
'backend_interface_retry_backoff_multiplier': 2.0}.iteritems():
if key not in backend_connection_manager:
backend_connection_manager[key] = value
else:
backend_connection_manager = current_config
for value in backend_connection_manager.values():
if isinstance(value, dict):
for key, val in value.items():
if key.startswith('backend_interface'):
backend_connection_manager[key] = val
changes = True
del value[key]
for key, value in {'backend_interface_retries_on_error': 5,
'backend_interface_retry_interval_secs': 1,
'backend_interface_retry_backoff_multiplier': 2.0}.iteritems():
if key not in backend_connection_manager:
changes = True
backend_connection_manager[key] = value
if changes is True:
storagedriver_config.clear_backend_connection_manager()
storagedriver_config.configure_backend_connection_manager(**backend_connection_manager)
storagedriver_config.save(root_client)
# Add '-reboot' to volumedriver services (because of updated 'backend_connection_manager' section)
ExtensionsToolbox.edit_version_file(client=root_client,
package_name='volumedriver',
old_run_file='{0}/{1}.version'.format(ServiceFactory.RUN_FILE_DIR, 'volumedriver_{0}'.format(vpool.name)))
if service_manager.__class__ == Systemd:
root_client.run(['systemctl', 'daemon-reload'])
########################################
# Update metadata_store_bits information
vpools = VPoolList.get_vpools()
for vpool in vpools:
bits = None
for storagedriver in vpool.storagedrivers:
key = '/ovs/framework/hosts/{0}/services/volumedriver_{1}'.format(storagedriver.storagerouter.machine_id, vpool.name)
if Configuration.exists(key=key) and 'METADATASTORE_BITS' not in Configuration.get(key=key):
if bits is None:
entries = service_manager.extract_from_service_file(name='ovs-volumedriver_{0}'.format(vpool.name),
client=sr_client_map[storagedriver.storagerouter_guid],
entries=['METADATASTORE_BITS='])
if len(entries) == 1:
bits = entries[0].split('=')[-1]
bits = int(bits) if bits.isdigit() else 5
if bits is not None:
try:
content = Configuration.get(key=key)
content['METADATASTORE_BITS'] = bits
Configuration.set(key=key, value=content)
except:
MigrationController._logger.exception('Error updating volumedriver info for vPool {0} on StorageRouter {1}'.format(vpool.name, storagedriver.storagerouter.name))
if bits is not None:
vpool.metadata_store_bits = bits
vpool.save()
#####################################
# Update the vPool metadata structure
def _update_metadata_structure(metadata):
metadata = copy.deepcopy(metadata)
cache_structure = {'read': False,
'write': False,
'is_backend': False,
'quota': None,
'backend_info': {'name': None, # Will be filled in when is_backend is true
'backend_guid': None,
'alba_backend_guid': None,
'policies': None,
'preset': None,
'arakoon_config': None,
'connection_info': {'client_id': None,
'client_secret': None,
'host': None,
'port': None,
'local': None}}
}
structure_map = {StorageDriverConfiguration.CACHE_BLOCK: {'read': 'block_cache_on_read',
'write': 'block_cache_on_write',
'quota': 'quota_bc',
'backend_prefix': 'backend_bc_{0}'},
StorageDriverConfiguration.CACHE_FRAGMENT: {'read': 'fragment_cache_on_read',
'write': 'fragment_cache_on_write',
'quota': 'quota_fc',
'backend_prefix': 'backend_aa_{0}'}}
if 'arakoon_config' in metadata['backend']: # Arakoon config should be placed under the backend info
metadata['backend']['backend_info']['arakoon_config'] = metadata['backend'].pop('arakoon_config')
if 'connection_info' in metadata['backend']: # Connection info sohuld be placed under the backend info
metadata['backend']['backend_info']['connection_info'] = metadata['backend'].pop('connection_info')
if 'caching_info' not in metadata: # Caching info is the new key
would_be_caching_info = {}
metadata['caching_info'] = would_be_caching_info
# Extract all caching data for every storagerouter
current_caching_info = metadata['backend'].pop('caching_info') # Pop to mutate metadata
for storagerouter_guid in current_caching_info.iterkeys():
current_cache_data = current_caching_info[storagerouter_guid]
storagerouter_caching_info = {}
would_be_caching_info[storagerouter_guid] = storagerouter_caching_info
for cache_type, cache_type_mapping in structure_map.iteritems():
new_cache_structure = copy.deepcopy(cache_structure)
storagerouter_caching_info[cache_type] = new_cache_structure
for new_structure_key, old_structure_key in cache_type_mapping.iteritems():
if new_structure_key == 'backend_prefix':
# Get possible backend related info
metadata_key = old_structure_key.format(storagerouter_guid)
if metadata_key not in metadata:
continue
backend_data = metadata.pop(metadata_key) # Pop to mutate metadata
new_cache_structure['is_backend'] = True
# Copy over the old data
new_cache_structure['backend_info']['arakoon_config'] = backend_data['arakoon_config']
new_cache_structure['backend_info'].update(backend_data['backend_info'])
new_cache_structure['backend_info']['connection_info'].update(backend_data['connection_info'])
else:
new_cache_structure[new_structure_key] = current_cache_data.get(old_structure_key)
return metadata
vpools = VPoolList.get_vpools()
for vpool in vpools:
try:
new_metadata = _update_metadata_structure(vpool.metadata)
vpool.metadata = new_metadata
vpool.save()
except KeyError:
MigrationController._logger.exception('Exceptions occurred when updating the metadata for vPool {0}'.format(vpool.name))
##############################################
# Always use indent=4 during Configuration set
def _resave_all_config_entries(config_path='/ovs'):
"""
Recursive functions which checks every config management key if its a directory or not.
If not a directory, we retrieve the config and just save it again using the new indentation logic
"""
for item in Configuration.list(config_path):
new_path = config_path + '/' + item
print new_path
if Configuration.dir_exists(new_path) is True:
_resave_all_config_entries(config_path=new_path)
else:
try:
_config = Configuration.get(new_path)
Configuration.set(new_path, _config)
except:
_config = Configuration.get(new_path, raw=True)
Configuration.set(new_path, _config, raw=True)
if ExtensionMigrator.THIS_VERSION <= 13: # There is no way of checking whether this new indentation logic has been applied, so we only perform this for version 13 and lower
MigrationController._logger.info('Re-saving every configuration setting with new indentation rules')
_resave_all_config_entries()
############################
# Update some default values
def _update_manifest_cache_size(_proxy_config_key):
updated = False
manifest_cache_size = 500 * 1024 * 1024
if Configuration.exists(key=_proxy_config_key):
_proxy_config = Configuration.get(key=_proxy_config_key)
for cache_type in [StorageDriverConfiguration.CACHE_BLOCK, StorageDriverConfiguration.CACHE_FRAGMENT]:
if cache_type in _proxy_config and _proxy_config[cache_type][0] == 'alba':
if _proxy_config[cache_type][1]['manifest_cache_size'] != manifest_cache_size:
updated = True
_proxy_config[cache_type][1]['manifest_cache_size'] = manifest_cache_size
if _proxy_config['manifest_cache_size'] != manifest_cache_size:
updated = True
_proxy_config['manifest_cache_size'] = manifest_cache_size
if updated is True:
Configuration.set(key=_proxy_config_key, value=_proxy_config)
return updated
for storagedriver in StorageDriverList.get_storagedrivers():
try:
vpool = storagedriver.vpool
root_client = sr_client_map[storagedriver.storagerouter_guid]
_update_manifest_cache_size('/ovs/vpools/{0}/proxies/scrub/generic_scrub'.format(vpool.guid)) # Generic scrub proxy is deployed every time scrubbing kicks in, so no need to restart these services
for alba_proxy in storagedriver.alba_proxies:
if _update_manifest_cache_size('/ovs/vpools/{0}/proxies/{1}/config/main'.format(vpool.guid, alba_proxy.guid)) is True:
# Add '-reboot' to alba_proxy services (because of newly created services and removal of old service)
ExtensionsToolbox.edit_version_file(client=root_client,
package_name='alba',
old_run_file='{0}/{1}.version'.format(ServiceFactory.RUN_FILE_DIR, alba_proxy.service.name))
# Update 'backend_connection_manager' section
changes = False
storagedriver_config = StorageDriverConfiguration(vpool.guid, storagedriver.storagedriver_id)
if 'backend_connection_manager' not in storagedriver_config.configuration:
continue
current_config = storagedriver_config.configuration['backend_connection_manager']
for key, value in current_config.iteritems():
if key.isdigit() is True:
if value.get('alba_connection_asd_connection_pool_capacity') != 10:
changes = True
value['alba_connection_asd_connection_pool_capacity'] = 10
if value.get('alba_connection_timeout') != 30:
changes = True
value['alba_connection_timeout'] = 30
if value.get('alba_connection_rora_manifest_cache_capacity') != 25000:
changes = True
value['alba_connection_rora_manifest_cache_capacity'] = 25000
if changes is True:
storagedriver_config.clear_backend_connection_manager()
storagedriver_config.configure_backend_connection_manager(**current_config)
storagedriver_config.save(root_client)
# Add '-reboot' to volumedriver services (because of updated 'backend_connection_manager' section)
ExtensionsToolbox.edit_version_file(client=root_client,
package_name='volumedriver',
old_run_file='{0}/{1}.version'.format(ServiceFactory.RUN_FILE_DIR, 'volumedriver_{0}'.format(vpool.name)))
except Exception:
MigrationController._logger.exception('Updating default configuration values failed for StorageDriver {0}'.format(storagedriver.storagedriver_id))
####################################################
# Adding proxy fail fast as env variable for proxies
changed_clients = set()
for storagerouter in StorageRouterList.get_storagerouters():
root_client = sr_client_map[storagerouter.guid]
for service_name in service_manager.list_services(client=root_client):
if not service_name.startswith('ovs-albaproxy_'):
continue
if ServiceFactory.get_service_type() == 'systemd':
path = '/lib/systemd/system/{0}.service'.format(service_name)
check = 'Environment=ALBA_FAIL_FAST=true'
else:
path = '/etc/init/{0}.conf'.format(service_name)
check = 'env ALBA_FAIL_FAST=true'
if not root_client.file_exists(path):
continue
if check in root_client.file_read(path):
continue
try:
service_manager.regenerate_service(name=StorageDriverInstaller.SERVICE_TEMPLATE_PROXY, client=root_client, target_name=service_name)
changed_clients.add(root_client)
ExtensionsToolbox.edit_version_file(client=root_client,
package_name='alba',
old_run_file='{0}/{1}.version'.format(ServiceFactory.RUN_FILE_DIR, service_name))
except:
MigrationController._logger.exception('Error rebuilding service {0}'.format(service_name))
for root_client in changed_clients:
root_client.run(['systemctl', 'daemon-reload'])
######################################
# Integration of stats monkey (2.10.2)
if Configuration.get(key='/ovs/framework/migration|stats_monkey_integration', default=False) is False:
try:
# Get content of old key into new key
old_stats_monkey_key = '/statsmonkey/statsmonkey'
if Configuration.exists(key=old_stats_monkey_key) is True:
Configuration.set(key='/ovs/framework/monitoring/stats_monkey', value=Configuration.get(key=old_stats_monkey_key))
Configuration.delete(key=old_stats_monkey_key)
# Make sure to disable the stats monkey by default or take over the current schedule if it was configured manually before
celery_key = '/ovs/framework/scheduling/celery'
current_value = None
scheduling_config = Configuration.get(key=celery_key, default={})
if 'statsmonkey.run_all_stats' in scheduling_config: # Old celery task name of the stats monkey
current_value = scheduling_config.pop('statsmonkey.run_all_stats')
scheduling_config['ovs.stats_monkey.run_all'] = current_value
scheduling_config['alba.stats_monkey.run_all'] = current_value
Configuration.set(key=celery_key, value=scheduling_config)
support_key = '/ovs/framework/support'
support_config = Configuration.get(key=support_key)
support_config['support_agent'] = support_config.pop('enabled', True)
support_config['remote_access'] = support_config.pop('enablesupport', False)
Configuration.set(key=support_key, value=support_config)
# Make sure once this finished, it never runs again by setting this key to True
Configuration.set(key='/ovs/framework/migration|stats_monkey_integration', value=True)
except Exception:
MigrationController._logger.exception('Integration of stats monkey failed')
######################################################
# Write away cluster ID to a file for back-up purposes
try:
cluster_id = Configuration.get(key='/ovs/framework/cluster_id', default=None)
with open(Configuration.CONFIG_STORE_LOCATION, 'r') as config_file:
config = json.load(config_file)
if cluster_id is not None and config.get('cluster_id', None) is None:
config['cluster_id'] = cluster_id
with open(Configuration.CONFIG_STORE_LOCATION, 'w') as config_file:
json.dump(config, config_file, indent=4)
except Exception:
MigrationController._logger.exception('Writing cluster id to a file failed.')
#########################################################
# Additional string formatting in Arakoon services (2.11)
try:
if Configuration.get(key='/ovs/framework/migration|arakoon_service_update', default=False) is False:
arakoon_service_names = [ArakoonInstaller.get_service_name_for_cluster(cluster_name=cluster_name) for cluster_name in Configuration.list(key='ovs/arakoon')]
for storagerouter in StorageRouterList.get_masters():
for service_name in arakoon_service_names:
config_key = ServiceFactory.SERVICE_CONFIG_KEY.format(storagerouter.machine_id, service_name)
if Configuration.exists(key=config_key):
config = Configuration.get(key=config_key)
config['RUN_FILE_DIR'] = ServiceFactory.RUN_FILE_DIR
config['ARAKOON_PKG_NAME'] = PackageFactory.PKG_ARAKOON
config['ARAKOON_VERSION_CMD'] = PackageFactory.VERSION_CMD_ARAKOON
Configuration.set(key=config_key, value=config)
# Make sure once this finished, it never runs again by setting this key to True
Configuration.set(key='/ovs/framework/migration|arakoon_service_update', value=True)
except Exception:
MigrationController._logger.exception('Updating the string formatting for the Arakoon services failed')
############################################################
# Additional string formatting in ALBA proxy services (2.11)
changed_clients = set()
try:
if Configuration.get(key='/ovs/framework/migration|alba_proxy_service_update', default=False) is False:
alba_pkg_name, alba_version_cmd = PackageFactory.get_package_and_version_cmd_for(component=PackageFactory.COMP_ALBA)
for service in ServiceTypeList.get_by_name('AlbaProxy').services:
root_client = sr_client_map[service.storagerouter_guid]
config_key = ServiceFactory.SERVICE_CONFIG_KEY.format(service.storagerouter.machine_id, service.name)
if Configuration.exists(key=config_key):
config = Configuration.get(key=config_key)
config['RUN_FILE_DIR'] = ServiceFactory.RUN_FILE_DIR
config['ALBA_PKG_NAME'] = alba_pkg_name
config['ALBA_VERSION_CMD'] = alba_version_cmd
Configuration.set(key=config_key, value=config)
service_manager.regenerate_service(name=StorageDriverInstaller.SERVICE_TEMPLATE_PROXY,
client=root_client,
target_name='ovs-{0}'.format(service.name))
changed_clients.add(root_client)
# Make sure once this finished, it never runs again by setting this key to True
Configuration.set(key='/ovs/framework/migration|alba_proxy_service_update', value=True)
except Exception:
MigrationController._logger.exception('Updating the string formatting for the Arakoon services failed')
############################################################
# Additional string formatting in DTL/VOLDRV services (2.11)
try:
if Configuration.get(key='/ovs/framework/migration|voldrv_service_update', default=False) is False:
sd_pkg_name, sd_version_cmd = PackageFactory.get_package_and_version_cmd_for(component=PackageFactory.COMP_SD)
for vpool in VPoolList.get_vpools():
for storagedriver in vpool.storagedrivers:
root_client = sr_client_map[storagedriver.storagerouter_guid]
for entry in ['dtl', 'volumedriver']:
service_name = '{0}_{1}'.format(entry, vpool.name)
service_template = StorageDriverInstaller.SERVICE_TEMPLATE_DTL if entry == 'dtl' else StorageDriverInstaller.SERVICE_TEMPLATE_SD
config_key = ServiceFactory.SERVICE_CONFIG_KEY.format(storagedriver.storagerouter.machine_id, service_name)
if Configuration.exists(key=config_key):
config = Configuration.get(key=config_key)
config['RUN_FILE_DIR'] = ServiceFactory.RUN_FILE_DIR
config['VOLDRV_PKG_NAME'] = sd_pkg_name
config['VOLDRV_VERSION_CMD'] = sd_version_cmd
Configuration.set(key=config_key, value=config)
service_manager.regenerate_service(name=service_template,
client=root_client,
target_name='ovs-{0}'.format(service_name))
changed_clients.add(root_client)
# Make sure once this finished, it never runs again by setting this key to True
Configuration.set(key='/ovs/framework/migration|voldrv_service_update', value=True)
except Exception:
MigrationController._logger.exception('Updating the string formatting for the Arakoon services failed')
#######################################################
# Storing actual package name in version files (2.11.0) (https://github.com/openvstorage/framework/issues/1876)
if Configuration.get(key='/ovs/framework/migration|actual_package_name_in_version_file', default=False) is False:
try:
voldrv_pkg_name, _ = PackageFactory.get_package_and_version_cmd_for(component=PackageFactory.COMP_SD)
for storagerouter in StorageRouterList.get_storagerouters():
root_client = sr_client_map.get(storagerouter.guid)
if root_client is None:
continue
for file_name in root_client.file_list(directory=ServiceFactory.RUN_FILE_DIR):
if not file_name.endswith('.version'):
continue
file_path = '{0}/{1}'.format(ServiceFactory.RUN_FILE_DIR, file_name)
contents = root_client.file_read(filename=file_path)
regenerate = False
if voldrv_pkg_name == PackageFactory.PKG_VOLDRV_SERVER:
if 'volumedriver-server' in contents:
regenerate = True
contents = contents.replace('volumedriver-server', PackageFactory.PKG_VOLDRV_SERVER)
root_client.file_write(filename=file_path, contents=contents)
elif voldrv_pkg_name == PackageFactory.PKG_VOLDRV_SERVER_EE:
if 'volumedriver-server' in contents or PackageFactory.PKG_VOLDRV_SERVER in contents:
regenerate = True
contents = contents.replace('volumedriver-server', PackageFactory.PKG_VOLDRV_SERVER_EE)
contents = contents.replace(PackageFactory.PKG_VOLDRV_SERVER, PackageFactory.PKG_VOLDRV_SERVER_EE)
root_client.file_write(filename=file_path, contents=contents)
if regenerate is True:
service_manager.regenerate_service(name=StorageDriverInstaller.SERVICE_TEMPLATE_DTL if file_name.startswith('dtl') else StorageDriverInstaller.SERVICE_TEMPLATE_SD,
client=root_client,
target_name='ovs-{0}'.format(file_name.split('.')[0])) # Leave out .version
changed_clients.add(root_client)
Configuration.set(key='/ovs/framework/migration|actual_package_name_in_version_file', value=True)
except Exception:
MigrationController._logger.exception('Updating actual package name for version files failed')
for root_client in changed_clients:
try:
root_client.run(['systemctl', 'daemon-reload'])
except Exception:
MigrationController._logger.exception('Executing command "systemctl daemon-reload" failed')
#########################################################
# Addition of 'Environment=OCAMLRUNPARAM='b,a=1,s=4096k,O=50' for AlbaProxy SystemD services
if ServiceFactory.get_service_type() == 'systemd':
changed_clients = set()
for storagedriver in StorageDriverList.get_storagedrivers():
root_client = sr_client_map[storagedriver.storagerouter_guid]
for alba_proxy in storagedriver.alba_proxies:
service = alba_proxy.service
service_name = 'ovs-{0}'.format(service.name)
if not service_manager.has_service(name=service_name, client=root_client):
continue
if "Environment=OCAMLRUNPARAM='b,a=1,s=4096k,O=50" in root_client.file_read(filename='/lib/systemd/system/{0}.service'.format(service_name)):
continue
try:
service_manager.regenerate_service(name='ovs-albaproxy', client=root_client, target_name=service_name)
changed_clients.add(root_client)
except:
MigrationController._logger.exception('Error rebuilding service {0}'.format(service_name))
for root_client in changed_clients:
root_client.run(['systemctl', 'daemon-reload'])
#########################################################
# Addition of 'Environment=OCAMLRUNPARAM='b,a=1,s=4096k,O=50' for Arakoon SystemD services
if ServiceFactory.get_service_type() == 'systemd':
changed_clients = set()
for storagerouter in StorageRouterList.get_storagerouters():
root_client = sr_client_map[storagerouter.guid]
for service_name in service_manager.list_services(client=root_client):
if not service_name.startswith('ovs-arakoon-'):
continue
if not service_manager.has_service(name=service_name, client=root_client):
continue
if "Environment=OCAMLRUNPARAM='b,a=1,s=4096k,O=50" in root_client.file_read(filename='/lib/systemd/system/{0}.service'.format(service_name)):
continue
try:
service_manager.regenerate_service(name='ovs-arakoon', client=root_client, target_name=service_name)
changed_clients.add(root_client)
except:
MigrationController._logger.exception('Error rebuilding service {0}'.format(service_name))
for root_client in changed_clients:
root_client.run(['systemctl', 'daemon-reload'])
MigrationController._logger.info('Finished out of band migrations')
def gather_scrub_work():
"""
Retrieve and execute scrub work
:return: None
"""
ScheduledTaskController._logger.info('Gather Scrub - Started')
scrub_locations = {}
for storage_driver in StorageDriverList.get_storagedrivers():
for partition in storage_driver.partitions:
if DiskPartition.ROLES.SCRUB == partition.role:
ScheduledTaskController._logger.info(
'Gather Scrub - Storage Router {0:<15} has SCRUB partition at {1}'
.format(storage_driver.storagerouter.ip,
partition.path))
if storage_driver.storagerouter not in scrub_locations:
try:
sshclient = SSHClient(storage_driver.storagerouter)
# Use ServiceManager(sshclient) to make sure ovs-workers are actually running
if ServiceManager.get_service_status(
'workers', sshclient) is False:
ScheduledTaskController._logger.warning(
'Gather Scrub - Storage Router {0:<15} - workers are not running'
.format(storage_driver.storagerouter.ip))
else:
scrub_locations[
storage_driver.storagerouter] = str(
partition.path)
except UnableToConnectException:
ScheduledTaskController._logger.warning(
'Gather Scrub - Storage Router {0:<15} is not reachable'
.format(storage_driver.storagerouter.ip))
if len(scrub_locations) == 0:
raise RuntimeError('No scrub locations found')
vdisk_guids = set()
for vmachine in VMachineList.get_customer_vmachines():
for vdisk in vmachine.vdisks:
if vdisk.info['object_type'] == 'BASE':
vdisk_guids.add(vdisk.guid)
for vdisk in VDiskList.get_without_vmachine():
if vdisk.info['object_type'] == 'BASE':
vdisk_guids.add(vdisk.guid)
if len(vdisk_guids) == 0:
ScheduledTaskController._logger.info(
'Gather Scrub - No scrub work needed'.format(len(vdisk_guids)))
return
ScheduledTaskController._logger.info(
'Gather Scrub - Checking {0} volumes for scrub work'.format(
len(vdisk_guids)))
local_machineid = System.get_my_machine_id()
local_storage_router = None
local_scrub_location = None
local_vdisks_to_scrub = []
result_set = {}
storage_router_list = []
scrub_map = {}
for index, scrub_info in enumerate(scrub_locations.items()):
start_index = index * len(vdisk_guids) / len(scrub_locations)
end_index = (index + 1) * len(vdisk_guids) / len(scrub_locations)
storage_router = scrub_info[0]
vdisk_guids_to_scrub = list(vdisk_guids)[start_index:end_index]
local = storage_router.machine_id == local_machineid
ScheduledTaskController._logger.info(
'Gather Scrub - Storage Router {0:<15} ({1}) - Scrubbing {2} virtual disks'
.format(storage_router.ip,
'local' if local is True else 'remote',
len(vdisk_guids_to_scrub)))
if local is True:
local_storage_router = storage_router
local_scrub_location = scrub_info[1]
local_vdisks_to_scrub = vdisk_guids_to_scrub
else:
result_set[storage_router.
ip] = ScheduledTaskController._execute_scrub_work.s(
scrub_location=scrub_info[1],
vdisk_guids=vdisk_guids_to_scrub).apply_async(
routing_key='sr.{0}'.format(
storage_router.machine_id))
storage_router_list.append(storage_router)
scrub_map[storage_router.ip] = vdisk_guids_to_scrub
# Remote tasks have been launched, now start the local task and then wait for remote tasks to finish
processed_guids = []
if local_scrub_location is not None and len(local_vdisks_to_scrub) > 0:
try:
processed_guids = ScheduledTaskController._execute_scrub_work(
scrub_location=local_scrub_location,
vdisk_guids=local_vdisks_to_scrub)
except Exception as ex:
ScheduledTaskController._logger.error(
'Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'
.format(local_storage_router.ip, ex))
all_results, failed_nodes = CeleryToolbox.manage_running_tasks(
result_set,
timesleep=60) # Check every 60 seconds if tasks are still running
for ip, result in all_results.iteritems():
if isinstance(result, list):
processed_guids.extend(result)
else:
ScheduledTaskController._logger.error(
'Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'
.format(ip, result))
result_set = {}
for failed_node in failed_nodes:
ScheduledTaskController._logger.warning(
'Scrubbing failed on node {0}. Will reschedule on another node.'
.format(failed_node))
vdisk_guids_to_scrub = scrub_map[failed_node]
rescheduled_work = False
for storage_router, scrub_location in scrub_locations.items():
if storage_router.ip not in failed_nodes:
if storage_router.machine_id != local_machineid:
ScheduledTaskController._logger.info(
'Rescheduled scrub work from node {0} to node {1}.'
.format(failed_node, storage_router.ip))
result_set[
storage_router.
ip] = ScheduledTaskController._execute_scrub_work.s(
scrub_location=scrub_location,
vdisk_guids=vdisk_guids_to_scrub).apply_async(
routing_key='sr.{0}'.format(
storage_router.machine_id))
storage_router_list.append(storage_router)
rescheduled_work = True
break
if rescheduled_work is False:
if local_scrub_location is not None:
try:
processed_guids.extend(
ScheduledTaskController._execute_scrub_work(
scrub_location=local_scrub_location,
vdisk_guids=vdisk_guids_to_scrub))
except Exception as ex:
ScheduledTaskController._logger.error(
'Gather Scrub - Storage Router Local - Scrubbing failed with error:\n - {0}'
.format(ex))
else:
ScheduledTaskController._logger.warning(
'No nodes left to reschedule work from node {0}'.
format(failed_node))
if len(result_set) > 0:
all_results2, failed_nodes = CeleryToolbox.manage_running_tasks(
result_set, timesleep=60
) # Check every 60 seconds if tasks are still running
for ip, result in all_results2.iteritems():
if isinstance(result, list):
processed_guids.extend(result)
else:
ScheduledTaskController._logger.error(
'Gather Scrub - Storage Router {0:<15} - Scrubbing failed with error:\n - {1}'
.format(ip, result))
if len(set(processed_guids)) != len(vdisk_guids) or set(
processed_guids).difference(vdisk_guids):
raise RuntimeError('Scrubbing failed for 1 or more storagerouters')
ScheduledTaskController._logger.info('Gather Scrub - Finished')
