def extend_cluster(master_ip, new_ip, cluster_name, exclude_ports, base_dir):
"""
Extends a cluster to a given new node
"""
logger.debug("Extending cluster {0} from {1} to {2}".format(cluster_name, master_ip, new_ip))
client = SSHClient(master_ip)
config = ArakoonClusterConfig(cluster_name)
config.load_config(client)
client = SSHClient(new_ip)
base_dir = base_dir.rstrip("/")
port_range = client.config_read("ovs.ports.arakoon")
ports = System.get_free_ports(port_range, exclude_ports, 2, client)
node_name = System.get_my_machine_id(client)
if not [node.name for node in config.nodes if node.name == node_name]:
config.nodes.append(
ArakoonNodeConfig(
name=node_name,
ip=new_ip,
client_port=ports[0],
messaging_port=ports[1],
log_dir=ArakoonInstaller.ARAKOON_LOG_DIR.format(cluster_name),
home=ArakoonInstaller.ARAKOON_HOME_DIR.format(base_dir, cluster_name),
tlog_dir=ArakoonInstaller.ARAKOON_TLOG_DIR.format(base_dir, cluster_name),
)
)
ArakoonInstaller._deploy(config)
logger.debug("Extending cluster {0} from {1} to {2} completed".format(cluster_name, master_ip, new_ip))
return {"client_port": ports[0], "messaging_port": ports[1]}
def create_cluster(cluster_name, ip, exclude_ports, plugins=None):
"""
Creates a cluster
"""
logger.debug('Creating cluster {0} on {1}'.format(cluster_name, ip))
client = SSHClient(ip)
base_dir = client.config_read('ovs.arakoon.location').rstrip('/')
port_range = client.config_read('ovs.ports.arakoon')
ports = System.get_free_ports(port_range, exclude_ports, 2, client)
node_name = System.get_my_machine_id(client)
config = ArakoonClusterConfig(cluster_name, plugins)
if not [node.name for node in config.nodes if node.name == node_name]:
config.nodes.append(
ArakoonNodeConfig(
name=node_name,
ip=ip,
client_port=ports[0],
messaging_port=ports[1],
log_dir=ArakoonInstaller.ARAKOON_LOG_DIR.format(
cluster_name),
home=ArakoonInstaller.ARAKOON_HOME_DIR.format(
base_dir, cluster_name),
tlog_dir=ArakoonInstaller.ARAKOON_TLOG_DIR.format(
base_dir, cluster_name)))
ArakoonInstaller._deploy(config)
logger.debug('Creating cluster {0} on {1} completed'.format(
cluster_name, ip))
return {'client_port': ports[0], 'messaging_port': ports[1]}
def extend_cluster(master_ip, new_ip, cluster_name, exclude_ports):
"""
Extends a cluster to a given new node
"""
logger.debug('Extending cluster {0} from {1} to {2}'.format(
cluster_name, master_ip, new_ip))
client = SSHClient(master_ip)
config = ArakoonClusterConfig(cluster_name)
config.load_config(client)
client = SSHClient(new_ip)
base_dir = client.config_read('ovs.arakoon.location').rstrip('/')
port_range = client.config_read('ovs.ports.arakoon')
ports = System.get_free_ports(port_range, exclude_ports, 2, client)
node_name = System.get_my_machine_id(client)
if not [node.name for node in config.nodes if node.name == node_name]:
config.nodes.append(
ArakoonNodeConfig(
name=node_name,
ip=new_ip,
client_port=ports[0],
messaging_port=ports[1],
log_dir=ArakoonInstaller.ARAKOON_LOG_DIR.format(
cluster_name),
home=ArakoonInstaller.ARAKOON_HOME_DIR.format(
base_dir, cluster_name),
tlog_dir=ArakoonInstaller.ARAKOON_TLOG_DIR.format(
base_dir, cluster_name)))
ArakoonInstaller._deploy(config)
logger.debug('Extending cluster {0} from {1} to {2} completed'.format(
cluster_name, master_ip, new_ip))
return {'client_port': ports[0], 'messaging_port': ports[1]}
def create_cluster(cluster_name, ip, exclude_ports, base_dir, plugins=None):
"""
Creates a cluster
"""
logger.debug("Creating cluster {0} on {1}".format(cluster_name, ip))
client = SSHClient(ip)
base_dir = base_dir.rstrip("/")
port_range = client.config_read("ovs.ports.arakoon")
ports = System.get_free_ports(port_range, exclude_ports, 2, client)
node_name = System.get_my_machine_id(client)
config = ArakoonClusterConfig(cluster_name, plugins)
if not [node.name for node in config.nodes if node.name == node_name]:
config.nodes.append(
ArakoonNodeConfig(
name=node_name,
ip=ip,
client_port=ports[0],
messaging_port=ports[1],
log_dir=ArakoonInstaller.ARAKOON_LOG_DIR.format(cluster_name),
home=ArakoonInstaller.ARAKOON_HOME_DIR.format(base_dir, cluster_name),
tlog_dir=ArakoonInstaller.ARAKOON_TLOG_DIR.format(base_dir, cluster_name),
)
)
ArakoonInstaller._deploy(config)
logger.debug("Creating cluster {0} on {1} completed".format(cluster_name, ip))
return {"client_port": ports[0], "messaging_port": ports[1]}
def _get_free_ports(client):
node_name = System.get_my_machine_id(client)
clusters = []
exclude_ports = []
if EtcdConfiguration.dir_exists(ArakoonInstaller.ETCD_CONFIG_ROOT):
for cluster_name in EtcdConfiguration.list(
ArakoonInstaller.ETCD_CONFIG_ROOT):
try:
config = ArakoonClusterConfig(cluster_name)
config.load_config()
for node in config.nodes:
if node.name == node_name:
clusters.append(cluster_name)
exclude_ports.append(node.client_port)
exclude_ports.append(node.messaging_port)
except:
logger.error(
' Could not load port information of cluster {0}'.
format(cluster_name))
ports = System.get_free_ports(
EtcdConfiguration.get(
'/ovs/framework/hosts/{0}/ports|arakoon'.format(node_name)),
exclude_ports, 2, client)
logger.debug(
' Loaded free ports {0} based on existing clusters {1}'.format(
ports, clusters))
return ports
def create_cluster(cluster_name, ip, exclude_ports, plugins=None):
ai = ArakoonInstaller()
ai.clear_config()
client = SSHClient.load(ip)
port_range = System.read_remote_config(client, 'ovs.ports.arakoon')
free_ports = System.get_free_ports(port_range, exclude_ports, 2, client)
ai.create_config(cluster_name, ip, free_ports[0], free_ports[1], plugins)
ai.generate_configs(client)
ai.create_dir_structure(client)
return {'client_port': free_ports[0],
'messaging_port': free_ports[1]}
def extend_cluster(src_ip, tgt_ip, cluster_name, exclude_ports):
ai = ArakoonInstaller()
ai.load_config_from(cluster_name, src_ip)
client = SSHClient.load(tgt_ip)
tgt_id = System.get_my_machine_id(client)
port_range = System.read_remote_config(client, 'ovs.ports.arakoon')
free_ports = System.get_free_ports(port_range, exclude_ports, 2, client)
ai.create_dir_structure(client)
ai.add_node_to_config(tgt_id, tgt_ip, free_ports[0], free_ports[1])
ai.upload_config_for(cluster_name)
return {'client_port': free_ports[0],
'messaging_port': free_ports[1]}
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_free_ports(client):
node_name = System.get_my_machine_id(client)
clusters = []
exclude_ports = []
if Configuration.dir_exists(ArakoonInstaller.CONFIG_ROOT):
for cluster_name in Configuration.list(ArakoonInstaller.CONFIG_ROOT):
config = ArakoonClusterConfig(cluster_name, False)
config.load_config()
for node in config.nodes:
if node.name == node_name:
clusters.append(cluster_name)
exclude_ports.append(node.client_port)
exclude_ports.append(node.messaging_port)
ports = System.get_free_ports(Configuration.get('/ovs/framework/hosts/{0}/ports|arakoon'.format(node_name)), exclude_ports, 2, client)
ArakoonInstaller._logger.debug(' Loaded free ports {0} based on existing clusters {1}'.format(ports, clusters))
return ports
def check_port_ranges(result_handler, minimal_port_amount=20):
"""
Checks whether the expected amount of ports is available for the requested amount of ports
:param result_handler: logging object
:type result_handler: ovs.extensions.healthcheck.result.HCResults
:param minimal_port_amount: minimal number of ports without warning
:type minimal_port_amount: int
:return: None
:rtype: NoneType
"""
# @todo: check other port ranges too
port_range = Configuration.get('/ovs/framework/hosts/{0}/ports|storagedriver'.format(OpenvStorageHealthCheck.LOCAL_ID))
expected_ports = System.get_free_ports(selected_range=port_range, amount=0)
result_handler.info('Checking if enough ports are still available for OpenvStorage')
if len(expected_ports) >= minimal_port_amount:
result_handler.success('{} ports free'.format(len(expected_ports)))
else:
result_handler.warning('{} ports found, less than {}'.format(len(expected_ports), minimal_port_amount))
def _get_free_ports(client):
node_name = System.get_my_machine_id(client)
clusters = []
exclude_ports = []
if EtcdConfiguration.dir_exists(ArakoonInstaller.ETCD_CONFIG_ROOT):
for cluster_name in EtcdConfiguration.list(ArakoonInstaller.ETCD_CONFIG_ROOT):
try:
config = ArakoonClusterConfig(cluster_name)
config.load_config()
for node in config.nodes:
if node.name == node_name:
clusters.append(cluster_name)
exclude_ports.append(node.client_port)
exclude_ports.append(node.messaging_port)
except:
logger.error(' Could not load port information of cluster {0}'.format(cluster_name))
ports = System.get_free_ports(EtcdConfiguration.get('/ovs/framework/hosts/{0}/ports|arakoon'.format(node_name)), exclude_ports, 2, client)
logger.debug(' Loaded free ports {0} based on existing clusters {1}'.format(ports, clusters))
return ports
def _get_free_ports(client):
node_name = System.get_my_machine_id(client)
clusters = []
exclude_ports = []
if Configuration.dir_exists(ArakoonInstaller.CONFIG_ROOT):
for cluster_name in Configuration.list(
ArakoonInstaller.CONFIG_ROOT):
config = ArakoonClusterConfig(cluster_name, False)
config.load_config()
for node in config.nodes:
if node.name == node_name:
clusters.append(cluster_name)
exclude_ports.append(node.client_port)
exclude_ports.append(node.messaging_port)
ports = System.get_free_ports(
Configuration.get(
'/ovs/framework/hosts/{0}/ports|arakoon'.format(node_name)),
exclude_ports, 2, client)
ArakoonInstaller._logger.debug(
' Loaded free ports {0} based on existing clusters {1}'.format(
ports, clusters))
return ports
def prepare_mds_service(client, storagerouter, vpool, fresh_only=True, start=False):
"""
Prepares an MDS service:
* Creates the required configuration
* Sets up the service files
Assumes the StorageRouter and VPool are already configured with a StorageDriver and that all model-wise
configuration regarding both is completed.
"""
mdsservice_type = ServiceTypeList.get_by_name('MetadataServer')
storagedriver = [sd for sd in vpool.storagedrivers if sd.storagerouter_guid == storagerouter.guid][0]
# Fetch service sequence number
service_number = -1
for mds_service in vpool.mds_services:
if mds_service.service.storagerouter_guid == storagerouter.guid:
service_number = max(mds_service.number, service_number)
if fresh_only is True and service_number >= 0:
return None # There are already one or more MDS services running, aborting
service_number += 1
# Find free port
occupied_ports = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
occupied_ports.append(service.ports[0])
port = System.get_free_ports(Configuration.get('ovs.ports.mds'),
exclude=occupied_ports, nr=1, client=client)[0]
# Add service to the model
service = Service()
service.name = 'metadataserver_{0}_{1}'.format(vpool.name, service_number)
service.type = mdsservice_type
service.storagerouter = storagerouter
service.ports = [port]
service.save()
mds_service = MDSService()
mds_service.service = service
mds_service.vpool = vpool
mds_service.number = service_number
mds_service.save()
# Prepare some directores
scratch_dir = '{0}/mds_{1}_{2}'.format(storagedriver.mountpoint_temp, vpool.name, service_number)
rocksdb_dir = '{0}/mds_{1}_{2}'.format(storagedriver.mountpoint_md, vpool.name, service_number)
client.run('mkdir -p {0}'.format(scratch_dir))
client.run('mkdir -p {0}'.format(rocksdb_dir))
# Generate the configuration file
metadataserver_config = StorageDriverConfiguration('metadataserver', vpool.name, number=service_number)
metadataserver_config.load(client)
metadataserver_config.clean() # Clean out obsolete values
if vpool.backend_type.code == 'alba':
metadataserver_config.configure_backend_connection_manager(alba_connection_host='127.0.0.1',
alba_connection_port=storagedriver.alba_proxy.service.ports[0],
backend_type='ALBA')
else:
metadataserver_config.configure_backend_connection_manager(**vpool.metadata)
metadataserver_config.configure_metadata_server(mds_address=storagerouter.ip,
mds_port=service.ports[0],
mds_scratch_dir=scratch_dir,
mds_rocksdb_path=rocksdb_dir)
metadataserver_config.save(client)
# Create system services
params = {'<VPOOL_NAME>': vpool.name,
'<SERVICE_NUMBER>': str(service_number)}
template_dir = '/opt/OpenvStorage/config/templates/upstart'
client.run('cp -f {0}/ovs-metadataserver.conf {0}/ovs-metadataserver_{1}_{2}.conf'.format(template_dir, vpool.name, service_number))
service_script = """
from ovs.plugin.provider.service import Service
Service.add_service(package=('openvstorage', 'metadataserver'), name='metadataserver_{0}_{1}', command=None, stop_command=None, params={2})
""".format(vpool.name, service_number, params)
System.exec_remote_python(client, service_script)
if start is True:
System.exec_remote_python(client, """
from ovs.plugin.provider.service import Service
Service.enable_service('{0}')
""".format(service.name))
System.exec_remote_python(client, """
from ovs.plugin.provider.service import Service
Service.start_service('{0}')
""".format(service.name))
return mds_service
def prepare_mds_service(storagerouter, vpool, fresh_only, reload_config):
"""
Prepares an MDS service:
* Creates the required configuration
* Sets up the service files
Assumes the StorageRouter and VPool are already configured with a StorageDriver and that all model-wise
configuration regarding both is completed.
:param storagerouter: Storagerouter on which MDS service will be created
:param vpool: The vPool for which the MDS service will be created
:param fresh_only: If True and no current mds services exist for this vpool on this storagerouter, a new 1 will be created
:param reload_config: If True, the volumedriver's updated configuration will be reloaded
"""
# Fetch service sequence number based on MDS services for current vPool and current storage router
service_number = -1
for mds_service in vpool.mds_services:
if mds_service.service.storagerouter_guid == storagerouter.guid:
service_number = max(mds_service.number, service_number)
if fresh_only is True and service_number >= 0:
return # There is already 1 or more MDS services running, aborting
# VALIDATIONS
# 1. Find free port based on MDS services for all vPools on current storage router
client = SSHClient(storagerouter)
mdsservice_type = ServiceTypeList.get_by_name('MetadataServer')
occupied_ports = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
occupied_ports.extend(service.ports)
mds_port_range = client.config_read('ovs.ports.mds')
free_ports = System.get_free_ports(selected_range=mds_port_range,
exclude=occupied_ports,
nr=1,
client=client)
if not free_ports:
raise RuntimeError('Failed to find an available port on storage router {0} within range {1}'.format(storagerouter.name, mds_port_range))
# 2. Partition check
db_partition = None
for disk in storagerouter.disks:
for partition in disk.partitions:
if DiskPartition.ROLES.DB in partition.roles:
db_partition = partition
break
if db_partition is None:
raise RuntimeError('Could not find DB partition on storage router {0}'.format(storagerouter.name))
# 3. Verify storage driver configured
storagedrivers = [sd for sd in vpool.storagedrivers if sd.storagerouter_guid == storagerouter.guid]
if not storagedrivers:
raise RuntimeError('Expected to find a configured storagedriver for vpool {0} on storage router {1}'.format(vpool.name, storagerouter.name))
# MODEL UPDATES
# 1. Service
service_number += 1
service = Service()
service.name = 'metadataserver_{0}_{1}'.format(vpool.name, service_number)
service.type = mdsservice_type
service.ports = [free_ports[0]]
service.storagerouter = storagerouter
service.save()
mds_service = MDSService()
mds_service.vpool = vpool
mds_service.number = service_number
mds_service.service = service
mds_service.save()
# 2. Storage driver partitions
from ovs.lib.storagedriver import StorageDriverController
sdp = StorageDriverController.add_storagedriverpartition(storagedrivers[0], {'size': None,
'role': DiskPartition.ROLES.DB,
'sub_role': StorageDriverPartition.SUBROLE.MDS,
'partition': db_partition,
'mds_service': mds_service})
# CONFIGURATIONS
# 1. Volumedriver
mds_nodes = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
mds_service = service.mds_service
if mds_service.vpool_guid == vpool.guid:
mds_nodes.append({'host': service.storagerouter.ip,
'port': service.ports[0],
'db_directory': sdp.path,
'scratch_directory': sdp.path})
# Generate the correct section in the Storage Driver's configuration
storagedriver_config = StorageDriverConfiguration('storagedriver', vpool.name)
storagedriver_config.load(client)
storagedriver_config.clean() # Clean out obsolete values
storagedriver_config.configure_metadata_server(mds_nodes=mds_nodes)
storagedriver_config.save(client, reload_config=reload_config)
return mds_service
def test_local_remote_check(self):
local_result = System.get_free_ports(self.single_port_range)
client = SSHClient(self.ip)
remote_result = System.get_free_ports(self.single_port_range,
client=client)
self.assertEqual(local_result, remote_result)
def test_check_if_single_port_is_free(self):
result = System.get_free_ports(self.single_port)
expected = [1500]
self.assertTrue(result == expected,
'Expected: {0}, got: {1}'.format(expected, result))
def test_support_for_multiple_port_ranges(self):
result = System.get_free_ports(self.multiple_port_range,
range(3000, 8874), 2)
expected = [8874, 8875]
self.assertTrue(result == expected,
'Expected: {0} got: {1}'.format(expected, result))
def prepare_mds_service(client, storagerouter, vpool, fresh_only=True, reload_config=False):
"""
Prepares an MDS service:
* Creates the required configuration
* Sets up the service files
Assumes the StorageRouter and VPool are already configured with a StorageDriver and that all model-wise
configuration regarding both is completed.
"""
from ovs.lib.storagedriver import StorageDriverController
mdsservice_type = ServiceTypeList.get_by_name('MetadataServer')
storagedriver = [sd for sd in vpool.storagedrivers if sd.storagerouter_guid == storagerouter.guid][0]
# Fetch service sequence number
service_number = -1
for mds_service in vpool.mds_services:
if mds_service.service.storagerouter_guid == storagerouter.guid:
service_number = max(mds_service.number, service_number)
if fresh_only is True and service_number >= 0:
return None # There are already one or more MDS services running, aborting
service_number += 1
# Find free port
occupied_ports = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
occupied_ports.append(service.ports[0])
port = System.get_free_ports(Configuration.get('ovs.ports.mds'),
exclude=occupied_ports, nr=1, client=client)[0]
# Add service to the model
service = DalService()
service.name = 'metadataserver_{0}_{1}'.format(vpool.name, service_number)
service.type = mdsservice_type
service.storagerouter = storagerouter
service.ports = [port]
service.save()
mds_service = MDSService()
mds_service.service = service
mds_service.vpool = vpool
mds_service.number = service_number
mds_service.save()
scrub_partition = None
db_partition = None
for disk in storagerouter.disks:
for partition in disk.partitions:
if DiskPartition.ROLES.DB in partition.roles:
db_partition = partition
if DiskPartition.ROLES.SCRUB in partition.roles:
scrub_partition = partition
if scrub_partition is None or db_partition is None:
raise RuntimeError('Could not find DB or SCRUB partition on StorageRouter {0}'.format(storagerouter.name))
StorageDriverController.add_storagedriverpartition(storagedriver, {'size': None,
'role': DiskPartition.ROLES.DB,
'sub_role': StorageDriverPartition.SUBROLE.MDS,
'partition': db_partition,
'mds_service': mds_service})
StorageDriverController.add_storagedriverpartition(storagedriver, {'size': None,
'role': DiskPartition.ROLES.SCRUB,
'sub_role': StorageDriverPartition.SUBROLE.MDS,
'partition': scrub_partition,
'mds_service': mds_service})
mds_nodes = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
mds_service = service.mds_service
if mds_service.vpool_guid == vpool.guid:
mds_nodes.append({'host': service.storagerouter.ip,
'port': service.ports[0],
'db_directory': [sd_partition.path for sd_partition in mds_service.storagedriver_partitions
if sd_partition.role == DiskPartition.ROLES.DB and sd_partition.sub_role == StorageDriverPartition.SUBROLE.MDS][0],
'scratch_directory': [sd_partition.path for sd_partition in mds_service.storagedriver_partitions
if sd_partition.role == DiskPartition.ROLES.SCRUB and sd_partition.sub_role == StorageDriverPartition.SUBROLE.MDS][0]})
# Generate the correct section in the Storage Driver's configuration
storagedriver_config = StorageDriverConfiguration('storagedriver', vpool.name)
storagedriver_config.load(client)
storagedriver_config.clean() # Clean out obsolete values
storagedriver_config.configure_metadata_server(mds_nodes=mds_nodes)
storagedriver_config.save(client, reload_config=reload_config)
return mds_service
def prepare_mds_service(storagerouter, vpool, fresh_only, reload_config):
"""
Prepares an MDS service:
* Creates the required configuration
* Sets up the service files
Assumes the StorageRouter and VPool are already configured with a StorageDriver and that all model-wise
configuration regarding both is completed.
:param storagerouter: Storagerouter on which MDS service will be created
:type storagerouter: StorageRouter
:param vpool: The vPool for which the MDS service will be created
:type vpool: VPool
:param fresh_only: If True and no current mds services exist for this vpool on this storagerouter, a new 1 will be created
:type fresh_only: bool
:param reload_config: If True, the volumedriver's updated configuration will be reloaded
:type reload_config: bool
:return: Newly created service
:rtype: MDSService
"""
# Fetch service sequence number based on MDS services for current vPool and current storage router
service_number = -1
for mds_service in vpool.mds_services:
if mds_service.service.storagerouter_guid == storagerouter.guid:
service_number = max(mds_service.number, service_number)
if fresh_only is True and service_number >= 0:
return # There is already 1 or more MDS services running, aborting
# VALIDATIONS
# 1. Find free port based on MDS services for all vPools on current storage router
client = SSHClient(storagerouter)
mdsservice_type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.MD_SERVER)
occupied_ports = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
occupied_ports.extend(service.ports)
mds_port_range = Configuration.get(
'/ovs/framework/hosts/{0}/ports|mds'.format(
System.get_my_machine_id(client)))
free_ports = System.get_free_ports(selected_range=mds_port_range,
exclude=occupied_ports,
nr=1,
client=client)
if not free_ports:
raise RuntimeError(
'Failed to find an available port on storage router {0} within range {1}'
.format(storagerouter.name, mds_port_range))
# 2. Partition check
db_partition = None
for disk in storagerouter.disks:
for partition in disk.partitions:
if DiskPartition.ROLES.DB in partition.roles:
db_partition = partition
break
if db_partition is None:
raise RuntimeError(
'Could not find DB partition on storage router {0}'.format(
storagerouter.name))
# 3. Verify storage driver configured
storagedrivers = [
sd for sd in vpool.storagedrivers
if sd.storagerouter_guid == storagerouter.guid
]
if not storagedrivers:
raise RuntimeError(
'Expected to find a configured storagedriver for vpool {0} on storage router {1}'
.format(vpool.name, storagerouter.name))
storagedriver = storagedrivers[0]
# MODEL UPDATES
# 1. Service
service_number += 1
service = Service()
service.name = 'metadataserver_{0}_{1}'.format(vpool.name,
service_number)
service.type = mdsservice_type
service.ports = [free_ports[0]]
service.storagerouter = storagerouter
service.save()
mds_service = MDSService()
mds_service.vpool = vpool
mds_service.number = service_number
mds_service.service = service
mds_service.save()
# 2. Storage driver partitions
from ovs.lib.storagedriver import StorageDriverController
StorageDriverController.add_storagedriverpartition(
storagedriver, {
'size': None,
'role': DiskPartition.ROLES.DB,
'sub_role': StorageDriverPartition.SUBROLE.MDS,
'partition': db_partition,
'mds_service': mds_service
})
# CONFIGURATIONS
# 1. Volumedriver
mds_nodes = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
mds_service = service.mds_service
if mds_service is not None:
if mds_service.vpool_guid == vpool.guid:
sdp = [
sd_partition for sd_partition in
mds_service.storagedriver_partitions
if sd_partition.role == DiskPartition.ROLES.DB
and sd_partition.sub_role ==
StorageDriverPartition.SUBROLE.MDS
][0]
mds_nodes.append({
'host': service.storagerouter.ip,
'port': service.ports[0],
'db_directory': sdp.path,
'scratch_directory': sdp.path
})
# Generate the correct section in the Storage Driver's configuration
storagedriver_config = StorageDriverConfiguration(
'storagedriver', vpool.guid, storagedriver.storagedriver_id)
storagedriver_config.load()
storagedriver_config.configure_metadata_server(mds_nodes=mds_nodes)
storagedriver_config.save(client, reload_config=reload_config)
return mds_service
def test_get_2_free_ports(self):
result = System.get_free_ports(self.single_port_range, range(8870, 8874), 2)
self.assertTrue(result == [8874, 8875], 'Expected 8874 got: {0}'.format(result))
def prepare_mds_service(client, storagerouter, vpool, fresh_only=True, reload_config=False):
"""
Prepares an MDS service:
* Creates the required configuration
* Sets up the service files
Assumes the StorageRouter and VPool are already configured with a StorageDriver and that all model-wise
configuration regarding both is completed.
"""
mdsservice_type = ServiceTypeList.get_by_name('MetadataServer')
storagedriver = [sd for sd in vpool.storagedrivers if sd.storagerouter_guid == storagerouter.guid][0]
# Fetch service sequence number
service_number = -1
for mds_service in vpool.mds_services:
if mds_service.service.storagerouter_guid == storagerouter.guid:
service_number = max(mds_service.number, service_number)
if fresh_only is True and service_number >= 0:
return None # There are already one or more MDS services running, aborting
service_number += 1
# Find free port
occupied_ports = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
occupied_ports.append(service.ports[0])
port = System.get_free_ports(Configuration.get('ovs.ports.mds'),
exclude=occupied_ports, nr=1, client=client)[0]
# Add service to the model
service = DalService()
service.name = 'metadataserver_{0}_{1}'.format(vpool.name, service_number)
service.type = mdsservice_type
service.storagerouter = storagerouter
service.ports = [port]
service.save()
mds_service = MDSService()
mds_service.service = service
mds_service.vpool = vpool
mds_service.number = service_number
mds_service.save()
mds_nodes = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
mds_service = service.mds_service
if mds_service.vpool_guid == vpool.guid:
mds_nodes.append({'host': service.storagerouter.ip,
'port': service.ports[0],
'db_directory': '{0}/mds_{1}_{2}'.format(storagedriver.mountpoint_md,
vpool.name,
mds_service.number),
'scratch_directory': '{0}/mds_{1}_{2}'.format(storagedriver.mountpoint_temp,
vpool.name,
mds_service.number)})
# Generate the correct section in the Storage Driver's configuration
storagedriver_config = StorageDriverConfiguration('storagedriver', vpool.name)
storagedriver_config.load(client)
storagedriver_config.clean() # Clean out obsolete values
storagedriver_config.configure_metadata_server(mds_nodes=mds_nodes)
storagedriver_config.save(client, reload_config=reload_config)
return mds_service
def test_check_if_single_port_is_free(self):
result = System.get_free_ports(self.single_port)
expected = [1500]
self.assertTrue(result == expected, 'Expected: {0}, got: {1}'.format(expected, result))
def test_get_2_free_ports(self):
result = System.get_free_ports(self.single_port_range, range(8870, 8874), 2)
expected = [8874, 8875]
self.assertTrue(result == expected, 'Expected {0} got: {1}'.format(expected, result))
def prepare_mds_service(storagerouter, vpool, fresh_only, reload_config):
"""
Prepares an MDS service:
* Creates the required configuration
* Sets up the service files
Assumes the StorageRouter and VPool are already configured with a StorageDriver and that all model-wise
configuration regarding both is completed.
:param storagerouter: Storagerouter on which MDS service will be created
:type storagerouter: StorageRouter
:param vpool: The vPool for which the MDS service will be created
:type vpool: VPool
:param fresh_only: If True and no current mds services exist for this vpool on this storagerouter, a new 1 will be created
:type fresh_only: bool
:param reload_config: If True, the volumedriver's updated configuration will be reloaded
:type reload_config: bool
:return: Newly created service
:rtype: MDSService
"""
# Fetch service sequence number based on MDS services for current vPool and current storage router
service_number = -1
for mds_service in vpool.mds_services:
if mds_service.service.storagerouter_guid == storagerouter.guid:
service_number = max(mds_service.number, service_number)
if fresh_only is True and service_number >= 0:
return # There is already 1 or more MDS services running, aborting
# VALIDATIONS
# 1. Find free port based on MDS services for all vPools on current storage router
client = SSHClient(storagerouter)
mdsservice_type = ServiceTypeList.get_by_name(ServiceType.SERVICE_TYPES.MD_SERVER)
occupied_ports = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
occupied_ports.extend(service.ports)
mds_port_range = Configuration.get(
"/ovs/framework/hosts/{0}/ports|mds".format(System.get_my_machine_id(client))
)
free_ports = System.get_free_ports(selected_range=mds_port_range, exclude=occupied_ports, nr=1, client=client)
if not free_ports:
raise RuntimeError(
"Failed to find an available port on storage router {0} within range {1}".format(
storagerouter.name, mds_port_range
)
)
# 2. Partition check
db_partition = None
for disk in storagerouter.disks:
for partition in disk.partitions:
if DiskPartition.ROLES.DB in partition.roles:
db_partition = partition
break
if db_partition is None:
raise RuntimeError("Could not find DB partition on storage router {0}".format(storagerouter.name))
# 3. Verify storage driver configured
storagedrivers = [sd for sd in vpool.storagedrivers if sd.storagerouter_guid == storagerouter.guid]
if not storagedrivers:
raise RuntimeError(
"Expected to find a configured storagedriver for vpool {0} on storage router {1}".format(
vpool.name, storagerouter.name
)
)
storagedriver = storagedrivers[0]
# MODEL UPDATES
# 1. Service
service_number += 1
service = Service()
service.name = "metadataserver_{0}_{1}".format(vpool.name, service_number)
service.type = mdsservice_type
service.ports = [free_ports[0]]
service.storagerouter = storagerouter
service.save()
mds_service = MDSService()
mds_service.vpool = vpool
mds_service.number = service_number
mds_service.service = service
mds_service.save()
# 2. Storage driver partitions
from ovs.lib.storagedriver import StorageDriverController
StorageDriverController.add_storagedriverpartition(
storagedriver,
{
"size": None,
"role": DiskPartition.ROLES.DB,
"sub_role": StorageDriverPartition.SUBROLE.MDS,
"partition": db_partition,
"mds_service": mds_service,
},
)
# CONFIGURATIONS
# 1. Volumedriver
mds_nodes = []
for service in mdsservice_type.services:
if service.storagerouter_guid == storagerouter.guid:
mds_service = service.mds_service
if mds_service is not None:
if mds_service.vpool_guid == vpool.guid:
sdp = [
sd_partition
for sd_partition in mds_service.storagedriver_partitions
if sd_partition.role == DiskPartition.ROLES.DB
and sd_partition.sub_role == StorageDriverPartition.SUBROLE.MDS
][0]
mds_nodes.append(
{
"host": service.storagerouter.ip,
"port": service.ports[0],
"db_directory": sdp.path,
"scratch_directory": sdp.path,
}
)
# Generate the correct section in the Storage Driver's configuration
storagedriver_config = StorageDriverConfiguration("storagedriver", vpool.guid, storagedriver.storagedriver_id)
storagedriver_config.load()
storagedriver_config.configure_metadata_server(mds_nodes=mds_nodes)
storagedriver_config.save(client, reload_config=reload_config)
return mds_service
def execute_scrub_work(queue, vpool, scrub_info, error_messages):
"""
Executes scrub work for a given vDisk queue and vPool, based on scrub_info
:param queue: a Queue with vDisk guids that need to be scrubbed (they should only be member of a single vPool)
:type queue: Queue
:param vpool: the vPool object of the vDisks
:type vpool: VPool
:param scrub_info: A dict containing scrub information: `scrub_path` with the path where to scrub and `storage_router` with the StorageRouter
that needs to do the work
:type scrub_info: dict
:param error_messages: A list of error messages to be filled
:type error_messages: list
:return: a list of error messages
:rtype: list
"""
def _verify_mds_config(current_vdisk):
current_vdisk.invalidate_dynamics('info')
vdisk_configs = current_vdisk.info['metadata_backend_config']
if len(vdisk_configs) == 0:
raise RuntimeError('Could not load MDS configuration')
return vdisk_configs
client = None
lock_time = 5 * 60
storagerouter = scrub_info['storage_router']
scrub_directory = '{0}/scrub_work_{1}_{2}'.format(scrub_info['scrub_path'], vpool.name, storagerouter.name)
scrub_config_key = 'ovs/vpools/{0}/proxies/scrub/scrub_config_{1}'.format(vpool.guid, storagerouter.guid)
backend_config_key = 'ovs/vpools/{0}/proxies/scrub/backend_config_{1}'.format(vpool.guid, storagerouter.guid)
alba_proxy_service = 'ovs-albaproxy_{0}_{1}_scrub'.format(vpool.name, storagerouter.name)
# Deploy a proxy
try:
with file_mutex(name='ovs_albaproxy_scrub', wait=lock_time):
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - Deploying ALBA proxy {2}'.format(vpool.name, storagerouter.name, alba_proxy_service))
client = SSHClient(storagerouter, 'root')
client.dir_create(scrub_directory)
client.dir_chmod(scrub_directory, 0777) # Celery task executed by 'ovs' user and should be able to write in it
if ServiceManager.has_service(name=alba_proxy_service, client=client) is True and ServiceManager.get_service_status(name=alba_proxy_service, client=client) is True:
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - Re-using existing proxy service {2}'.format(vpool.name, storagerouter.name, alba_proxy_service))
scrub_config = Configuration.get(scrub_config_key)
else:
machine_id = System.get_my_machine_id(client)
port_range = Configuration.get('/ovs/framework/hosts/{0}/ports|storagedriver'.format(machine_id))
port = System.get_free_ports(selected_range=port_range, nr=1, client=client)[0]
# Scrub config
# {u'albamgr_cfg_url': u'arakoon://config/ovs/vpools/71e2f717-f270-4a41-bbb0-d4c8c084d43e/proxies/64759516-3471-4321-b912-fb424568fc5b/config/abm?ini=%2Fopt%2FOpenvStorage%2Fconfig%2Farakoon_cacc.ini',
# u'fragment_cache': [u'none'],
# u'ips': [u'127.0.0.1'],
# u'log_level': u'info',
# u'manifest_cache_size': 17179869184,
# u'port': 0,
# u'transport': u'tcp'}
# Backend config
# {u'alba_connection_host': u'10.100.193.155',
# u'alba_connection_port': 26204,
# u'alba_connection_preset': u'preset',
# u'alba_connection_timeout': 15,
# u'alba_connection_transport': u'TCP',
# u'backend_interface_retries_on_error': 5,
# u'backend_interface_retry_backoff_multiplier': 2.0,
# u'backend_interface_retry_interval_secs': 1,
# u'backend_type': u'ALBA'}
scrub_config = Configuration.get('ovs/vpools/{0}/proxies/scrub/generic_scrub'.format(vpool.guid))
scrub_config['port'] = port
scrub_config['transport'] = 'tcp'
Configuration.set(scrub_config_key, json.dumps(scrub_config, indent=4), raw=True)
params = {'VPOOL_NAME': vpool.name,
'LOG_SINK': LogHandler.get_sink_path('alba_proxy'),
'CONFIG_PATH': Configuration.get_configuration_path(scrub_config_key)}
ServiceManager.add_service(name='ovs-albaproxy', params=params, client=client, target_name=alba_proxy_service)
ServiceManager.start_service(name=alba_proxy_service, client=client)
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - Deployed ALBA proxy {2}'.format(vpool.name, storagerouter.name, alba_proxy_service))
backend_config = Configuration.get('ovs/vpools/{0}/hosts/{1}/config'.format(vpool.guid, vpool.storagedrivers[0].storagedriver_id))['backend_connection_manager']
backend_config['alba_connection_host'] = '127.0.0.1'
backend_config['alba_connection_port'] = scrub_config['port']
Configuration.set(backend_config_key, json.dumps({"backend_connection_manager": backend_config}, indent=4), raw=True)
except Exception:
message = 'Scrubber - vPool {0} - StorageRouter {1} - An error occurred deploying ALBA proxy {2}'.format(vpool.name, storagerouter.name, alba_proxy_service)
error_messages.append(message)
ScheduledTaskController._logger.exception(message)
if client is not None and ServiceManager.has_service(name=alba_proxy_service, client=client) is True:
if ServiceManager.get_service_status(name=alba_proxy_service, client=client) is True:
ServiceManager.stop_service(name=alba_proxy_service, client=client)
ServiceManager.remove_service(name=alba_proxy_service, client=client)
if Configuration.exists(scrub_config_key):
Configuration.delete(scrub_config_key)
try:
# Empty the queue with vDisks to scrub
with remote(storagerouter.ip, [VDisk]) as rem:
while True:
vdisk = None
vdisk_guid = queue.get(False)
try:
# Check MDS master is local. Trigger MDS handover if necessary
vdisk = rem.VDisk(vdisk_guid)
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - Started scrubbing at location {3}'.format(vpool.name, storagerouter.name, vdisk.name, scrub_directory))
configs = _verify_mds_config(current_vdisk=vdisk)
storagedriver = StorageDriverList.get_by_storagedriver_id(vdisk.storagedriver_id)
if configs[0].get('ip') != storagedriver.storagerouter.ip:
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - MDS master is not local, trigger handover'.format(vpool.name, storagerouter.name, vdisk.name))
MDSServiceController.ensure_safety(VDisk(vdisk_guid)) # Do not use a remote VDisk instance here
configs = _verify_mds_config(current_vdisk=vdisk)
if configs[0].get('ip') != storagedriver.storagerouter.ip:
ScheduledTaskController._logger.warning('Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - Skipping because master MDS still not local'.format(vpool.name, storagerouter.name, vdisk.name))
continue
# Do the actual scrubbing
with vdisk.storagedriver_client.make_locked_client(str(vdisk.volume_id)) as locked_client:
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - Retrieve and apply scrub work'.format(vpool.name, storagerouter.name, vdisk.name))
work_units = locked_client.get_scrubbing_workunits()
for work_unit in work_units:
res = locked_client.scrub(work_unit=work_unit,
scratch_dir=scrub_directory,
log_sinks=[LogHandler.get_sink_path('scrubber', allow_override=True)],
backend_config=Configuration.get_configuration_path(backend_config_key))
locked_client.apply_scrubbing_result(scrubbing_work_result=res)
if work_units:
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - {3} work units successfully applied'.format(vpool.name, storagerouter.name, vdisk.name, len(work_units)))
else:
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - No scrubbing required'.format(vpool.name, storagerouter.name, vdisk.name))
except Exception:
if vdisk is None:
message = 'Scrubber - vPool {0} - StorageRouter {1} - vDisk with guid {2} could not be found'.format(vpool.name, storagerouter.name, vdisk_guid)
else:
message = 'Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - Scrubbing failed'.format(vpool.name, storagerouter.name, vdisk.name)
error_messages.append(message)
ScheduledTaskController._logger.exception(message)
except Empty: # Raised when all items have been fetched from the queue
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - Queue completely processed'.format(vpool.name, storagerouter.name))
except Exception:
message = 'Scrubber - vPool {0} - StorageRouter {1} - Scrubbing failed'.format(vpool.name, storagerouter.name)
error_messages.append(message)
ScheduledTaskController._logger.exception(message)
# Delete the proxy again
try:
with file_mutex(name='ovs_albaproxy_scrub', wait=lock_time):
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - Removing service {2}'.format(vpool.name, storagerouter.name, alba_proxy_service))
client = SSHClient(storagerouter, 'root')
client.dir_delete(scrub_directory)
if ServiceManager.has_service(alba_proxy_service, client=client):
ServiceManager.stop_service(alba_proxy_service, client=client)
ServiceManager.remove_service(alba_proxy_service, client=client)
if Configuration.exists(scrub_config_key):
Configuration.delete(scrub_config_key)
ScheduledTaskController._logger.info('Scrubber - vPool {0} - StorageRouter {1} - Removed service {2}'.format(vpool.name, storagerouter.name, alba_proxy_service))
except Exception:
message = 'Scrubber - vPool {0} - StorageRouter {1} - Removing service {2} failed'.format(vpool.name, storagerouter.name, alba_proxy_service)
error_messages.append(message)
ScheduledTaskController._logger.exception(message)
def prepare_mds_service(cls, storagerouter, vpool):
"""
Prepares an MDS service:
* Creates the required configuration
* Sets up the service files
Assumes the StorageRouter and vPool are already configured with a StorageDriver and that all model-wise configurations regarding both have been completed.
:param storagerouter: StorageRouter on which the MDS service will be created
:type storagerouter: ovs.dal.hybrids.storagerouter.StorageRouter
:param vpool: The vPool for which the MDS service will be created
:type vpool: ovs.dal.hybrids.vpool.VPool
:raises RuntimeError: vPool is not extended on StorageRouter
No ServiceType found for 'MetadataServer'
No free port is found for the new MDSService
No partition found on StorageRouter with DB role
:return: Newly created junction service
:rtype: ovs.dal.hybrids.j_mdsservice.MDSService
"""
from ovs.lib.storagedriver import StorageDriverController # Import here to prevent from circular imports
cls._logger.info('StorageRouter {0} - vPool {1}: Preparing MDS junction service'.format(storagerouter.name, vpool.name))
mds_service = MDSService()
with volatile_mutex(name='prepare_mds_{0}'.format(storagerouter.guid), wait=30):
# VALIDATIONS
# Verify passed StorageRouter is part of the vPool
storagerouter.invalidate_dynamics(['vpools_guids'])
if vpool.guid not in storagerouter.vpools_guids:
raise RuntimeError('StorageRouter {0} is not part of vPool {1}'.format(storagerouter.name, vpool.name))
# Verify ServiceType existence
mds_service_type = ServiceTypeList.get_by_name(ServiceType.SERVICE_TYPES.MD_SERVER)
if mds_service_type is None:
raise RuntimeError('No ServiceType found with name {0}'.format(ServiceType.SERVICE_TYPES.MD_SERVER))
# Retrieve occupied ports for current StorageRouter and max MDSService number for current vPool/StorageRouter combo
service_number = -1
occupied_ports = []
for service in mds_service_type.services:
if service.storagerouter_guid == storagerouter.guid:
occupied_ports.extend(service.ports)
if service.mds_service.vpool_guid == vpool.guid:
service_number = max(service.mds_service.number, service_number)
client = SSHClient(endpoint=storagerouter)
mds_port_range = Configuration.get('/ovs/framework/hosts/{0}/ports|mds'.format(System.get_my_machine_id(client)))
free_ports = System.get_free_ports(selected_range=mds_port_range,
exclude=occupied_ports,
amount=1,
client=client)
if len(free_ports) != 1:
raise RuntimeError('Failed to find an available port on StorageRouter {0} within range {1}'.format(storagerouter.name, mds_port_range))
# Partition check
db_partition = None
for disk in storagerouter.disks:
for partition in disk.partitions:
if DiskPartition.ROLES.DB in partition.roles:
db_partition = partition
break
if db_partition is None:
raise RuntimeError('Could not find DB partition on StorageRouter {0}'.format(storagerouter.name))
# Verify StorageDriver configured
storagedrivers = [sd for sd in vpool.storagedrivers if sd.storagerouter_guid == storagerouter.guid]
if len(storagedrivers) != 1:
raise RuntimeError('Expected to find a configured StorageDriver for vPool {0} on StorageRouter {1}'.format(vpool.name, storagerouter.name))
# MODEL UPDATES
# Service and MDS service
service_number += 1
cls._logger.info('StorageRouter {0} - vPool {1}: Adding junction service with number {2}'.format(storagerouter.name, vpool.name, service_number))
service = Service()
service.name = 'metadataserver_{0}_{1}'.format(vpool.name, service_number)
service.type = mds_service_type
service.ports = free_ports
service.storagerouter = storagerouter
service.save()
mds_service.vpool = vpool
mds_service.number = service_number
mds_service.service = service
mds_service.save()
# StorageDriver partitions
cls._logger.info('StorageRouter {0} - vPool {1}: Adding StorageDriverPartition on partition with mount point {2}'.format(storagerouter.name, vpool.name, db_partition.mountpoint))
storagedriver = storagedrivers[0]
sdp = StorageDriverController.add_storagedriverpartition(storagedriver, {'size': None,
'role': DiskPartition.ROLES.DB,
'sub_role': StorageDriverPartition.SUBROLE.MDS,
'partition': db_partition,
'mds_service': mds_service})
# CONFIGURATIONS
# Volumedriver
mds_nodes = []
for sd_partition in storagedriver.partitions:
if sd_partition.role == DiskPartition.ROLES.DB and sd_partition.sub_role == StorageDriverPartition.SUBROLE.MDS and sd_partition.mds_service is not None:
service = sd_partition.mds_service.service
mds_nodes.append({'host': service.storagerouter.ip,
'port': service.ports[0],
'db_directory': '{0}/db'.format(sd_partition.path),
'scratch_directory': '{0}/scratch'.format(sd_partition.path)})
cls._logger.info('StorageRouter {0} - vPool {1}: Configuring StorageDriver with MDS nodes: {2}'.format(storagerouter.name, vpool.name, mds_nodes))
# Generate the correct section in the StorageDriver's configuration
try:
storagedriver_config = StorageDriverConfiguration(vpool.guid, storagedriver.storagedriver_id)
storagedriver_config.configure_metadata_server(mds_nodes=mds_nodes)
storagedriver_config.save(client)
except Exception:
cls._logger.exception('StorageRouter {0} - vPool {1}: Configuring StorageDriver failed. Reverting model changes'.format(storagerouter.name, vpool.name))
# Clean up model changes if error occurs
sdp.delete()
mds_service.delete() # Must be removed before the service
service.delete()
return mds_service
def test_first_free_port_after_system_range_succeeds(self):
result = System.get_free_ports([1025])
expected = [1025]
self.assertTrue(result == expected,
'Expected {0} got: {1}'.format(expected, result))
def test_get_2_free_ports(self):
result = System.get_free_ports(self.single_port_range,
range(8870, 8874), 2)
expected = [8874, 8875]
self.assertTrue(result == expected,
'Expected {0} got: {1}'.format(expected, result))
def test_first_free_port_after_system_range_succeeds(self):
result = System.get_free_ports([1025])
expected = [1025]
self.assertTrue(result == expected, 'Expected {0} got: {1}'.format(expected, result))
def _deploy_stack_and_scrub(queue, vpool, scrub_info, error_messages):
"""
Executes scrub work for a given vDisk queue and vPool, based on scrub_info
:param queue: a Queue with vDisk guids that need to be scrubbed (they should only be member of a single vPool)
:type queue: Queue
:param vpool: the vPool object of the vDisks
:type vpool: VPool
:param scrub_info: A dict containing scrub information:
`scrub_path` with the path where to scrub
`storage_router` with the StorageRouter that needs to do the work
:type scrub_info: dict
:param error_messages: A list of error messages to be filled (by reference)
:type error_messages: list
:return: None
:rtype: NoneType
"""
if len(vpool.storagedrivers
) == 0 or not vpool.storagedrivers[0].storagedriver_id:
error_messages.append(
'vPool {0} does not have any valid StorageDrivers configured'.
format(vpool.name))
return
service_manager = ServiceFactory.get_manager()
client = None
lock_time = 5 * 60
storagerouter = scrub_info['storage_router']
partition_guid = scrub_info['partition_guid']
alba_proxy_service = 'ovs-albaproxy_{0}_{1}_{2}_scrub'.format(
vpool.name, storagerouter.name, partition_guid)
scrub_directory = '{0}/scrub_work_{1}_{2}'.format(
scrub_info['scrub_path'], vpool.name, partition_guid)
scrub_config_key = 'ovs/vpools/{0}/proxies/scrub/scrub_config_{1}'.format(
vpool.guid, partition_guid)
backend_config_key = 'ovs/vpools/{0}/proxies/scrub/backend_config_{1}'.format(
vpool.guid, partition_guid)
# Deploy a proxy
try:
with file_mutex(name='ovs_albaproxy_scrub', wait=lock_time):
GenericController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Deploying ALBA proxy {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
client = SSHClient(storagerouter, 'root')
client.dir_create(scrub_directory)
client.dir_chmod(
scrub_directory, 0777
) # Celery task executed by 'ovs' user and should be able to write in it
if service_manager.has_service(
name=alba_proxy_service, client=client
) is True and service_manager.get_service_status(
name=alba_proxy_service, client=client) == 'active':
GenericController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Re-using existing proxy service {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
scrub_config = Configuration.get(scrub_config_key)
else:
machine_id = System.get_my_machine_id(client)
port_range = Configuration.get(
'/ovs/framework/hosts/{0}/ports|storagedriver'.format(
machine_id))
with volatile_mutex('deploy_proxy_for_scrub_{0}'.format(
storagerouter.guid),
wait=30):
port = System.get_free_ports(selected_range=port_range,
nr=1,
client=client)[0]
scrub_config = Configuration.get(
'ovs/vpools/{0}/proxies/scrub/generic_scrub'.format(
vpool.guid))
scrub_config['port'] = port
scrub_config['transport'] = 'tcp'
Configuration.set(scrub_config_key,
json.dumps(scrub_config, indent=4),
raw=True)
params = {
'VPOOL_NAME':
vpool.name,
'LOG_SINK':
LogHandler.get_sink_path(alba_proxy_service),
'CONFIG_PATH':
Configuration.get_configuration_path(scrub_config_key)
}
service_manager.add_service(name='ovs-albaproxy',
params=params,
client=client,
target_name=alba_proxy_service)
service_manager.start_service(name=alba_proxy_service,
client=client)
GenericController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Deployed ALBA proxy {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
backend_config = Configuration.get(
'ovs/vpools/{0}/hosts/{1}/config'.format(
vpool.guid, vpool.storagedrivers[0].storagedriver_id
))['backend_connection_manager']
if backend_config.get('backend_type') != 'MULTI':
backend_config['alba_connection_host'] = '127.0.0.1'
backend_config['alba_connection_port'] = scrub_config[
'port']
else:
for value in backend_config.itervalues():
if isinstance(value, dict):
value['alba_connection_host'] = '127.0.0.1'
value['alba_connection_port'] = scrub_config[
'port']
# Copy backend connection manager information in separate key
Configuration.set(
backend_config_key,
json.dumps({"backend_connection_manager": backend_config},
indent=4),
raw=True)
except Exception:
message = 'Scrubber - vPool {0} - StorageRouter {1} - An error occurred deploying ALBA proxy {2}'.format(
vpool.name, storagerouter.name, alba_proxy_service)
error_messages.append(message)
GenericController._logger.exception(message)
if client is not None and service_manager.has_service(
name=alba_proxy_service, client=client) is True:
if service_manager.get_service_status(
name=alba_proxy_service, client=client) == 'active':
service_manager.stop_service(name=alba_proxy_service,
client=client)
service_manager.remove_service(name=alba_proxy_service,
client=client)
if Configuration.exists(scrub_config_key):
Configuration.delete(scrub_config_key)
# Execute the actual scrubbing
threads = []
threads_key = '/ovs/framework/hosts/{0}/config|scrub_stack_threads'.format(
storagerouter.machine_id)
amount_threads = Configuration.get(
key=threads_key) if Configuration.exists(key=threads_key) else 2
if not isinstance(amount_threads, int):
error_messages.append(
'Amount of threads to spawn must be an integer for StorageRouter with ID {0}'
.format(storagerouter.machine_id))
return
amount_threads = max(amount_threads,
1) # Make sure amount_threads is at least 1
amount_threads = min(min(queue.qsize(), amount_threads),
20) # Make sure amount threads is max 20
GenericController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Spawning {2} threads for proxy service {3}'
.format(vpool.name, storagerouter.name, amount_threads,
alba_proxy_service))
for index in range(amount_threads):
thread = Thread(name='execute_scrub_{0}_{1}_{2}'.format(
vpool.guid, partition_guid, index),
target=GenericController._execute_scrub,
args=(queue, vpool, scrub_info, scrub_directory,
error_messages))
thread.start()
threads.append(thread)
for thread in threads:
thread.join()
# Delete the proxy again
try:
with file_mutex(name='ovs_albaproxy_scrub', wait=lock_time):
GenericController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Removing service {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
client = SSHClient(storagerouter, 'root')
client.dir_delete(scrub_directory)
if service_manager.has_service(alba_proxy_service,
client=client):
service_manager.stop_service(alba_proxy_service,
client=client)
service_manager.remove_service(alba_proxy_service,
client=client)
if Configuration.exists(scrub_config_key):
Configuration.delete(scrub_config_key)
GenericController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Removed service {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
except Exception:
message = 'Scrubber - vPool {0} - StorageRouter {1} - Removing service {2} failed'.format(
vpool.name, storagerouter.name, alba_proxy_service)
error_messages.append(message)
GenericController._logger.exception(message)
def test_support_for_multiple_port_ranges(self):
result = System.get_free_ports(self.multiple_port_range, range(3000, 8874), 2)
expected = [8874, 8875]
self.assertTrue(result == expected, 'Expected: {0} got: {1}'.format(expected, result))
def test_first_free_port_after_system_range_succeeds(self):
result = System.get_free_ports(1025)
self.assertTrue(result == [1025], 'Expected 1025 got: {0}'.format(result))
def test_local_remote_check(self):
local_result = System.get_free_ports(self.single_port_range)
client = SSHClient(self.ip)
remote_result = System.get_free_ports(self.single_port_range, client=client)
self.assertEqual(local_result, remote_result)
def execute_scrub_work(queue, vpool, scrub_info, error_messages):
"""
Executes scrub work for a given vDisk queue and vPool, based on scrub_info
:param queue: a Queue with vDisk guids that need to be scrubbed (they should only be member of a single vPool)
:type queue: Queue
:param vpool: the vPool object of the vDisks
:type vpool: VPool
:param scrub_info: A dict containing scrub information: `scrub_path` with the path where to scrub and `storage_router` with the StorageRouter
that needs to do the work
:type scrub_info: dict
:param error_messages: A list of error messages to be filled
:type error_messages: list
:return: a list of error messages
:rtype: list
"""
def _verify_mds_config(current_vdisk):
current_vdisk.invalidate_dynamics('info')
vdisk_configs = current_vdisk.info['metadata_backend_config']
if len(vdisk_configs) == 0:
raise RuntimeError('Could not load MDS configuration')
return vdisk_configs
client = None
lock_time = 5 * 60
storagerouter = scrub_info['storage_router']
scrub_directory = '{0}/scrub_work_{1}_{2}'.format(
scrub_info['scrub_path'], vpool.name, storagerouter.name)
scrub_config_key = 'ovs/vpools/{0}/proxies/scrub/scrub_config_{1}'.format(
vpool.guid, storagerouter.guid)
backend_config_key = 'ovs/vpools/{0}/proxies/scrub/backend_config_{1}'.format(
vpool.guid, storagerouter.guid)
alba_proxy_service = 'ovs-albaproxy_{0}_{1}_scrub'.format(
vpool.name, storagerouter.name)
# Deploy a proxy
try:
with file_mutex(name='ovs_albaproxy_scrub', wait=lock_time):
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Deploying ALBA proxy {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
client = SSHClient(storagerouter, 'root')
client.dir_create(scrub_directory)
client.dir_chmod(
scrub_directory, 0777
) # Celery task executed by 'ovs' user and should be able to write in it
if ServiceManager.has_service(
name=alba_proxy_service, client=client
) is True and ServiceManager.get_service_status(
name=alba_proxy_service, client=client) is True:
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Re-using existing proxy service {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
scrub_config = Configuration.get(scrub_config_key)
else:
machine_id = System.get_my_machine_id(client)
port_range = Configuration.get(
'/ovs/framework/hosts/{0}/ports|storagedriver'.format(
machine_id))
port = System.get_free_ports(selected_range=port_range,
nr=1,
client=client)[0]
# Scrub config
# {u'albamgr_cfg_url': u'arakoon://config/ovs/vpools/71e2f717-f270-4a41-bbb0-d4c8c084d43e/proxies/64759516-3471-4321-b912-fb424568fc5b/config/abm?ini=%2Fopt%2FOpenvStorage%2Fconfig%2Farakoon_cacc.ini',
# u'fragment_cache': [u'none'],
# u'ips': [u'127.0.0.1'],
# u'log_level': u'info',
# u'manifest_cache_size': 17179869184,
# u'port': 0,
# u'transport': u'tcp'}
# Backend config
# {u'alba_connection_host': u'10.100.193.155',
# u'alba_connection_port': 26204,
# u'alba_connection_preset': u'preset',
# u'alba_connection_timeout': 15,
# u'alba_connection_transport': u'TCP',
# u'backend_interface_retries_on_error': 5,
# u'backend_interface_retry_backoff_multiplier': 2.0,
# u'backend_interface_retry_interval_secs': 1,
# u'backend_type': u'ALBA'}
scrub_config = Configuration.get(
'ovs/vpools/{0}/proxies/scrub/generic_scrub'.format(
vpool.guid))
scrub_config['port'] = port
scrub_config['transport'] = 'tcp'
Configuration.set(scrub_config_key,
json.dumps(scrub_config, indent=4),
raw=True)
params = {
'VPOOL_NAME':
vpool.name,
'LOG_SINK':
LogHandler.get_sink_path('alba_proxy'),
'CONFIG_PATH':
Configuration.get_configuration_path(scrub_config_key)
}
ServiceManager.add_service(name='ovs-albaproxy',
params=params,
client=client,
target_name=alba_proxy_service)
ServiceManager.start_service(name=alba_proxy_service,
client=client)
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Deployed ALBA proxy {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
backend_config = Configuration.get(
'ovs/vpools/{0}/hosts/{1}/config'.format(
vpool.guid, vpool.storagedrivers[0].storagedriver_id
))['backend_connection_manager']
backend_config['alba_connection_host'] = '127.0.0.1'
backend_config['alba_connection_port'] = scrub_config['port']
Configuration.set(
backend_config_key,
json.dumps({"backend_connection_manager": backend_config},
indent=4),
raw=True)
except Exception:
message = 'Scrubber - vPool {0} - StorageRouter {1} - An error occurred deploying ALBA proxy {2}'.format(
vpool.name, storagerouter.name, alba_proxy_service)
error_messages.append(message)
ScheduledTaskController._logger.exception(message)
if client is not None and ServiceManager.has_service(
name=alba_proxy_service, client=client) is True:
if ServiceManager.get_service_status(name=alba_proxy_service,
client=client) is True:
ServiceManager.stop_service(name=alba_proxy_service,
client=client)
ServiceManager.remove_service(name=alba_proxy_service,
client=client)
if Configuration.exists(scrub_config_key):
Configuration.delete(scrub_config_key)
try:
# Empty the queue with vDisks to scrub
with remote(storagerouter.ip, [VDisk]) as rem:
while True:
vdisk = None
vdisk_guid = queue.get(False)
try:
# Check MDS master is local. Trigger MDS handover if necessary
vdisk = rem.VDisk(vdisk_guid)
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - Started scrubbing at location {3}'
.format(vpool.name, storagerouter.name, vdisk.name,
scrub_directory))
configs = _verify_mds_config(current_vdisk=vdisk)
storagedriver = StorageDriverList.get_by_storagedriver_id(
vdisk.storagedriver_id)
if configs[0].get(
'ip') != storagedriver.storagerouter.ip:
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - MDS master is not local, trigger handover'
.format(vpool.name, storagerouter.name,
vdisk.name))
MDSServiceController.ensure_safety(
VDisk(vdisk_guid)
) # Do not use a remote VDisk instance here
configs = _verify_mds_config(current_vdisk=vdisk)
if configs[0].get(
'ip') != storagedriver.storagerouter.ip:
ScheduledTaskController._logger.warning(
'Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - Skipping because master MDS still not local'
.format(vpool.name, storagerouter.name,
vdisk.name))
continue
# Do the actual scrubbing
with vdisk.storagedriver_client.make_locked_client(
str(vdisk.volume_id)) as locked_client:
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - Retrieve and apply scrub work'
.format(vpool.name, storagerouter.name,
vdisk.name))
work_units = locked_client.get_scrubbing_workunits(
)
for work_unit in work_units:
res = locked_client.scrub(
work_unit=work_unit,
scratch_dir=scrub_directory,
log_sinks=[
LogHandler.get_sink_path(
'scrubber', allow_override=True)
],
backend_config=Configuration.
get_configuration_path(backend_config_key))
locked_client.apply_scrubbing_result(
scrubbing_work_result=res)
if work_units:
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - {3} work units successfully applied'
.format(vpool.name, storagerouter.name,
vdisk.name, len(work_units)))
else:
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - No scrubbing required'
.format(vpool.name, storagerouter.name,
vdisk.name))
except Exception:
if vdisk is None:
message = 'Scrubber - vPool {0} - StorageRouter {1} - vDisk with guid {2} could not be found'.format(
vpool.name, storagerouter.name, vdisk_guid)
else:
message = 'Scrubber - vPool {0} - StorageRouter {1} - vDisk {2} - Scrubbing failed'.format(
vpool.name, storagerouter.name, vdisk.name)
error_messages.append(message)
ScheduledTaskController._logger.exception(message)
except Empty: # Raised when all items have been fetched from the queue
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Queue completely processed'
.format(vpool.name, storagerouter.name))
except Exception:
message = 'Scrubber - vPool {0} - StorageRouter {1} - Scrubbing failed'.format(
vpool.name, storagerouter.name)
error_messages.append(message)
ScheduledTaskController._logger.exception(message)
# Delete the proxy again
try:
with file_mutex(name='ovs_albaproxy_scrub', wait=lock_time):
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Removing service {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
client = SSHClient(storagerouter, 'root')
client.dir_delete(scrub_directory)
if ServiceManager.has_service(alba_proxy_service,
client=client):
ServiceManager.stop_service(alba_proxy_service,
client=client)
ServiceManager.remove_service(alba_proxy_service,
client=client)
if Configuration.exists(scrub_config_key):
Configuration.delete(scrub_config_key)
ScheduledTaskController._logger.info(
'Scrubber - vPool {0} - StorageRouter {1} - Removed service {2}'
.format(vpool.name, storagerouter.name,
alba_proxy_service))
except Exception:
message = 'Scrubber - vPool {0} - StorageRouter {1} - Removing service {2} failed'.format(
vpool.name, storagerouter.name, alba_proxy_service)
error_messages.append(message)
ScheduledTaskController._logger.exception(message)
