def _fd_worker(queue, clients, result_handler, service_manager):
"""
Worker method to retrieve file descriptors
:param queue: Queue to use
:param clients: SSHClients to choose from
:param result_handler: Logging object
:param service_manager: Service manager instance
:return: None
:rtype: NoneType
"""
while not queue.empty():
cluster_name, _node_config, _results = queue.get(False)
errors = _results['errors']
output = _results['result']
identifier = 'Arakoon cluster {0} on node {1}'.format(cluster_name, _node_config.ip)
result_handler.info('Retrieving file descriptor information for {0}'.format(identifier), add_to_result=False)
try:
client = clients[_node_config.ip]
try:
# Handle config Arakoon
cluster_name = cluster_name if cluster_name != 'cacc' else 'config'
service_name = ArakoonInstaller.get_service_name_for_cluster(cluster_name)
pid = service_manager.get_service_pid(service_name, client)
file_descriptors = client.run(['lsof', '-i', '-a', '-p', pid]).splitlines()[1:]
except Exception as _ex:
errors.append(('lsof', _ex))
raise
output['fds'] = file_descriptors
except Exception as _ex:
result_handler.warning(
'Could not retrieve the file descriptor information for {0} ({1})'.format(identifier, str(_ex)), add_to_result=False)
finally:
queue.task_done()
def ensure_s3_transaction_safety(cls,
s3_cluster,
available_storagerouters,
s3_installer=None):
# type: (S3TransactionCluster, Dict[StorageRouter, DiskPartition], Optional[S3TransactionInstaller]) -> None
"""
Ensure that the S3 transaction cluster is safe and sound
:param s3_cluster: ABM Cluster object
:type s3_cluster: ABMCluster
:param available_storagerouters: All available storagerouters mapped with their DB partition
:type available_storagerouters: Dict[StorageRouter, DiskPartition]
:param s3_installer: The ABMInstaller to use. Defaults to creating a new one
:type s3_installer: ABMInstaller
:return: None
:rtype: NoneType
"""
s3_transaction_installer = s3_installer or S3TransactionInstaller()
metadata = ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=s3_cluster.name)
if 0 < len(s3_cluster.s3_transaction_services) < len(
available_storagerouters) and metadata['internal'] is True:
current_service_ips = [
j_service.service.storagerouter.ip
for j_service in s3_cluster.s3_transaction_services
]
for storagerouter, partition in available_storagerouters.iteritems(
):
if storagerouter.ip in current_service_ips:
continue
s3_transaction_installer.extend_s3_cluster(
storagerouter, s3_cluster)
current_service_ips.append(storagerouter.ip)
def _get_arakoon_clusters(cls, result_handler):
"""
Retrieves all Arakoon clusters registered in this OVSCluster
:param result_handler: Logging object
:type result_handler: ovs.extensions.healthcheck.result.HCResults
:return: Dict with the Arakoon cluster types as key and list with dicts which contain cluster names and pyrakoon clients
:rtype: dict(str, list[dict])
"""
result_handler.info('Fetching available arakoon clusters.', add_to_result=False)
arakoon_clusters = {}
for cluster_name in list(Configuration.list('/ovs/arakoon')) + ['cacc']:
# Determine Arakoon type
is_cacc = cluster_name == 'cacc'
arakoon_config = ArakoonClusterConfig(cluster_id=cluster_name, load_config=not is_cacc)
if is_cacc is True:
with open(Configuration.CACC_LOCATION) as config_file:
contents = config_file.read()
arakoon_config.read_config(contents=contents)
try:
arakoon_client = ArakoonInstaller.build_client(arakoon_config)
except (ArakoonNoMaster, ArakoonNoMasterResult) as ex:
result_handler.failure('Unable to find a master for Arakoon cluster {0}. (Message: {1})'.format(cluster_name, str(ex)),
code=ErrorCodes.master_none)
except Exception as ex:
msg = 'Unable to connect to Arakoon cluster {0}. (Message: {1})'.format(cluster_name, str(ex))
result_handler.exception(msg, code=ErrorCodes.unhandled_exception)
cls.logger.exception(msg)
continue
metadata = json.loads(arakoon_client.get(ArakoonInstaller.METADATA_KEY))
cluster_type = metadata['cluster_type']
if cluster_type not in arakoon_clusters:
arakoon_clusters[cluster_type] = []
arakoon_clusters[cluster_type].append({'cluster_name': cluster_name, 'client': arakoon_client, 'config': arakoon_config})
return arakoon_clusters
def _get_arakoon_clusters(cls, result_handler):
"""
Retrieves all Arakoon clusters registered in this OVSCluster
:param result_handler: Logging object
:type result_handler: ovs.extensions.healthcheck.result.HCResults
:return: Dict with the Arakoon cluster types as key and list with dicts which contain cluster names and pyrakoon clients
:rtype: dict(str, list[dict])
"""
result_handler.info('Fetching available arakoon clusters.',
add_to_result=False)
arakoon_clusters = {}
for cluster_name in list(
Configuration.list('/ovs/arakoon')) + ['cacc']:
# Determine Arakoon type
is_cacc = cluster_name == 'cacc'
arakoon_config = ArakoonClusterConfig(cluster_id=cluster_name,
load_config=not is_cacc)
if is_cacc is True:
with open(Configuration.CACC_LOCATION) as config_file:
contents = config_file.read()
arakoon_config.read_config(contents=contents)
try:
arakoon_client = ArakoonInstaller.build_client(arakoon_config)
except (ArakoonNoMaster, ArakoonNoMasterResult) as ex:
result_handler.failure(
'Unable to find a master for Arakoon cluster {0}. (Message: {1})'
.format(cluster_name, str(ex)),
code=ErrorCodes.master_none)
except Exception as ex:
msg = 'Unable to connect to Arakoon cluster {0}. (Message: {1})'.format(
cluster_name, str(ex))
result_handler.exception(msg,
code=ErrorCodes.unhandled_exception)
cls.logger.exception(msg)
continue
metadata = json.loads(
arakoon_client.get(ArakoonInstaller.METADATA_KEY))
cluster_type = metadata['cluster_type']
if cluster_type not in arakoon_clusters:
arakoon_clusters[cluster_type] = []
arakoon_clusters[cluster_type].append({
'cluster_name': cluster_name,
'client': arakoon_client,
'config': arakoon_config
})
return arakoon_clusters
def _on_demote(cluster_ip, master_ip, offline_node_ips=None):
"""
Handles the demote for the StorageDrivers
:param cluster_ip: IP of the node to demote
:type cluster_ip: str
:param master_ip: IP of the master node
:type master_ip: str
:param offline_node_ips: IPs of nodes which are offline
:type offline_node_ips: list
:return: None
"""
_ = master_ip
if offline_node_ips is None:
offline_node_ips = []
servicetype = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.ARAKOON)
current_service = None
remaining_ips = []
for service in servicetype.services:
if service.name == 'arakoon-voldrv' and service.is_internal is True: # Externally managed arakoon cluster services do not have StorageRouters
if service.storagerouter.ip == cluster_ip:
current_service = service
elif service.storagerouter.ip not in offline_node_ips:
remaining_ips.append(service.storagerouter.ip)
if current_service is not None:
if len(remaining_ips) == 0:
raise RuntimeError(
'Could not find any remaining arakoon nodes for the voldrv cluster'
)
StorageDriverController._logger.debug(
'* Shrink StorageDriver cluster')
cluster_name = str(
Configuration.get('/ovs/framework/arakoon_clusters|voldrv'))
arakoon_installer = ArakoonInstaller(cluster_name=cluster_name)
arakoon_installer.load()
arakoon_installer.shrink_cluster(removal_ip=cluster_ip,
offline_nodes=offline_node_ips)
arakoon_installer.restart_cluster_after_shrinking()
current_service.delete()
StorageDriverController._configure_arakoon_to_volumedriver(
cluster_name=cluster_name)
def _fd_worker(queue, clients, result_handler, service_manager):
"""
Worker method to retrieve file descriptors
:param queue: Queue to use
:param clients: SSHClients to choose from
:param result_handler: Logging object
:param service_manager: Service manager instance
:return: None
:rtype: NoneType
"""
while not queue.empty():
cluster_name, _node_config, _results = queue.get(False)
errors = _results['errors']
output = _results['result']
identifier = 'Arakoon cluster {0} on node {1}'.format(
cluster_name, _node_config.ip)
result_handler.info(
'Retrieving file descriptor information for {0}'.format(
identifier),
add_to_result=False)
try:
client = clients[_node_config.ip]
try:
# Handle config Arakoon
cluster_name = cluster_name if cluster_name != 'cacc' else 'config'
service_name = ArakoonInstaller.get_service_name_for_cluster(
cluster_name)
pid = service_manager.get_service_pid(service_name, client)
file_descriptors = client.run(
['lsof', '-i', '-a', '-p', pid]).splitlines()[1:]
except Exception as _ex:
errors.append(('lsof', _ex))
raise
output['fds'] = file_descriptors
except Exception as _ex:
result_handler.warning(
'Could not retrieve the file descriptor information for {0} ({1})'
.format(identifier, str(_ex)),
add_to_result=False)
finally:
queue.task_done()
def manual_alba_arakoon_checkup(alba_backend_guid,
nsm_clusters,
abm_cluster=None):
# type: (str, List[str], Optional[str]) -> Union[bool, None]
"""
Creates a new Arakoon cluster if required and extends cluster if possible on all available master nodes
:param alba_backend_guid: Guid of the ALBA Backend
:type alba_backend_guid: str
:param nsm_clusters: NSM clusters for this ALBA Backend
The code will claim the Arakoon clusters for this backend when provided
:type nsm_clusters: list[str]
:param abm_cluster: ABM cluster for this ALBA Backend
The code will claim the Arakoon cluster for this backend when provided
:type abm_cluster: str|None
:return: True if task completed, None if task was discarded (by decorator)
:rtype: bool|None
"""
if (abm_cluster is not None
and len(nsm_clusters) == 0) or (len(nsm_clusters) > 0
and abm_cluster is None):
raise ValueError(
'Both ABM cluster and NSM clusters must be provided')
if abm_cluster is not None:
# Check if the requested clusters are available
for cluster_name in [abm_cluster] + nsm_clusters:
try:
metadata = ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=cluster_name)
if metadata['in_use'] is True:
raise ValueError(
'Cluster {0} has already been claimed'.format(
cluster_name))
except NotFoundException:
raise ValueError(
'Could not find an Arakoon cluster with name: {0}'.
format(cluster_name))
AlbaArakoonController._alba_arakoon_checkup(
alba_backend_guid=alba_backend_guid,
abm_cluster=abm_cluster,
nsm_clusters=nsm_clusters)
return True
def remove_arakoon_cluster(cluster_name, master_storagerouter_ip):
"""
Delete a whole arakoon cluster
:param cluster_name: name of a existing arakoon cluster
:type cluster_name: str
:param master_storagerouter_ip: master ip address of a existing arakoon cluster
:type master_storagerouter_ip: str
"""
ArakoonRemover.LOGGER.info(
"Starting removing arakoon cluster with name `{0}`, master_ip `{1}`"
.format(cluster_name, master_storagerouter_ip))
arakoon_installer = ArakoonInstaller(cluster_name)
arakoon_installer.load()
arakoon_installer.delete_cluster()
ArakoonRemover.LOGGER.info(
"Finished removing arakoon cluster with name `{0}`, master_ip `{1}`"
.format(cluster_name, master_storagerouter_ip))
def test_node_config_checkup(self):
"""
Validates correct working of cluster registry checkup
"""
base_structure = {
'1': {
'vrouter_id': '1',
'message_host': '10.0.1.1',
'message_port': 1,
'xmlrpc_host': '10.0.0.1',
'xmlrpc_port': 2,
'failovercache_host': '10.0.1.1',
'failovercache_port': 3,
'network_server_uri': 'tcp://10.0.1.1:4',
'node_distance_map': None
},
'2': {
'vrouter_id': '2',
'message_host': '10.0.1.2',
'message_port': 1,
'xmlrpc_host': '10.0.0.2',
'xmlrpc_port': 2,
'failovercache_host': '10.0.1.2',
'failovercache_port': 3,
'network_server_uri': 'tcp://10.0.1.2:4',
'node_distance_map': None
}
}
def _validate_node_config(_config, _expected_map):
expected = copy.deepcopy(base_structure[_config.vrouter_id])
expected['node_distance_map'] = _expected_map[_config.vrouter_id]
self.assertDictEqual(
expected, {
'vrouter_id': _config.vrouter_id,
'message_host': _config.message_host,
'message_port': _config.message_port,
'xmlrpc_host': _config.xmlrpc_host,
'xmlrpc_port': _config.xmlrpc_port,
'failovercache_host': _config.failovercache_host,
'failovercache_port': _config.failovercache_port,
'network_server_uri': _config.network_server_uri,
'node_distance_map': _config.node_distance_map
})
structure = DalHelper.build_dal_structure({
'vpools': [1],
'domains': [1, 2],
'storagerouters': [1, 2],
'storagedrivers':
[(1, 1, 1), (2, 1, 2)], # (<id>, <vpool_id>, <storagerouter_id>)
'storagerouter_domains': [(1, 1, 1, False), (2, 2, 1, False)]
} # (id>, <storagerouter_id>, <domain_id>, <backup>)
)
storagerouters = structure['storagerouters']
vpool = structure['vpools'][1]
arakoon_installer = ArakoonInstaller(cluster_name='voldrv')
arakoon_installer.create_cluster(
cluster_type=ServiceType.ARAKOON_CLUSTER_TYPES.SD,
ip=storagerouters[1].ip,
base_dir='/tmp')
# Initial run, it will now be configured
StorageRouterClient.node_config_recordings = []
result = StorageDriverController.cluster_registry_checkup()
self.assertDictEqual(result,
{vpool.guid: {
'success': True,
'changes': True
}})
self.assertListEqual(
sorted(StorageRouterClient.node_config_recordings), ['1', '2'])
expected_map = {
'1': {
'2': StorageDriver.DISTANCES.NEAR
},
'2': {
'1': StorageDriver.DISTANCES.NEAR
}
}
configs = vpool.clusterregistry_client.get_node_configs()
for config in configs:
_validate_node_config(config, expected_map)
# Running it again should not change anything
StorageRouterClient.node_config_recordings = []
result = StorageDriverController.cluster_registry_checkup()
self.assertDictEqual(result,
{vpool.guid: {
'success': True,
'changes': False
}})
self.assertListEqual(
sorted(StorageRouterClient.node_config_recordings), [])
expected_map = {
'1': {
'2': StorageDriver.DISTANCES.NEAR
},
'2': {
'1': StorageDriver.DISTANCES.NEAR
}
}
configs = vpool.clusterregistry_client.get_node_configs()
for config in configs:
_validate_node_config(config, expected_map)
# Validate some error paths
domain = structure['domains'][2]
junction = structure['storagerouters'][1].domains[0]
junction.domain = domain
junction.save()
vpool_config_path = 'file://opt/OpenvStorage/config/framework.json?key=/ovs/vpools/{0}/hosts/1/config'.format(
vpool.guid)
StorageRouterClient.exceptions['server_revision'] = {
vpool_config_path: Exception('ClusterNotReachableException')
}
StorageRouterClient.node_config_recordings = []
result = StorageDriverController.cluster_registry_checkup()
self.assertDictEqual(result,
{vpool.guid: {
'success': True,
'changes': True
}})
self.assertListEqual(
sorted(StorageRouterClient.node_config_recordings), ['2'])
expected_map = {
'1': {
'2': StorageDriver.DISTANCES.INFINITE
},
'2': {
'1': StorageDriver.DISTANCES.INFINITE
}
}
configs = vpool.clusterregistry_client.get_node_configs()
for config in configs:
_validate_node_config(config, expected_map)
def _get_update_information_cluster_alba(cls, client, update_info,
package_info):
"""
In this function the services for each component / package combination are defined
This service information consists out of:
* Services to stop (before update) and start (after update of packages) -> 'services_stop_start'
* Services to restart after update (post-update logic) -> 'services_post_update'
* Down-times which will be caused due to service restarts -> 'downtime'
* Prerequisites that have not been met -> 'prerequisites'
Verify whether all relevant services have the correct binary active
Whether a service has the correct binary version in use, we use the ServiceFactory.get_service_update_versions functionality
When a service has an older binary version running, we add this information to the 'update_info'
This combined information is then stored in the 'package_information' of the StorageRouter DAL object
:param client: SSHClient on which to retrieve the service information required for an update
:type client: ovs.extensions.generic.sshclient.SSHClient
:param update_info: Dictionary passed in by the thread calling this function used to store all update information
:type update_info: dict
:param package_info: Dictionary containing the components and packages which have an update available for current SSHClient
:type package_info: dict
:return: None
:rtype: NoneType
"""
cls._logger.info(
'StorageRouter {0}: Refreshing ALBA update information'.format(
client.ip))
try:
binaries = cls._package_manager.get_binary_versions(client=client)
storagerouter = StorageRouterList.get_by_ip(ip=client.ip)
cls._logger.debug('StorageRouter {0}: Binary versions: {1}'.format(
client.ip, binaries))
# Retrieve Arakoon information
arakoon_info = {}
for service in storagerouter.services:
if service.type.name not in [
ServiceType.SERVICE_TYPES.ALBA_MGR,
ServiceType.SERVICE_TYPES.NS_MGR
]:
continue
if service.type.name == ServiceType.SERVICE_TYPES.ALBA_MGR:
cluster_name = service.abm_service.abm_cluster.name
alba_backend_name = service.abm_service.abm_cluster.alba_backend.name
else:
cluster_name = service.nsm_service.nsm_cluster.name
alba_backend_name = service.nsm_service.nsm_cluster.alba_backend.name
cls._logger.debug(
'StorageRouter {0}: Retrieving update information for Arakoon cluster {1}'
.format(client.ip, cluster_name))
arakoon_update_info = ArakoonInstaller.get_arakoon_update_info(
cluster_name=cluster_name)
cls._logger.debug(
'StorageRouter {0}: Arakoon update information for cluster {1}: {2}'
.format(client.ip, cluster_name, arakoon_update_info))
if arakoon_update_info['internal'] is True:
arakoon_info[arakoon_update_info['service_name']] = [
'backend', alba_backend_name
] if arakoon_update_info['downtime'] is True else None
for component, package_names in PackageFactory.get_package_info(
)['names'].iteritems():
package_names = sorted(package_names)
cls._logger.debug(
'StorageRouter {0}: Validating component {1} and related packages: {2}'
.format(client.ip, component, package_names))
if component not in update_info[client.ip]:
update_info[client.ip][component] = copy.deepcopy(
ServiceFactory.DEFAULT_UPDATE_ENTRY)
svc_component_info = update_info[client.ip][component]
pkg_component_info = package_info.get(component, {})
for package_name in package_names:
cls._logger.debug(
'StorageRouter {0}: Validating ALBA plugin related package {1}'
.format(client.ip, package_name))
if package_name == PackageFactory.PKG_OVS_BACKEND and package_name in pkg_component_info:
if ['gui', None] not in svc_component_info['downtime']:
svc_component_info['downtime'].append(
['gui', None])
if ['api', None] not in svc_component_info['downtime']:
svc_component_info['downtime'].append(
['api', None])
svc_component_info['services_stop_start'][10].append(
'ovs-watcher-framework')
svc_component_info['services_stop_start'][20].append(
'memcached')
cls._logger.debug(
'StorageRouter {0}: Added services "ovs-watcher-framework" and "memcached" to stop-start services'
.format(client.ip))
cls._logger.debug(
'StorageRouter {0}: Added GUI and API to downtime'.
format(client.ip))
elif package_name in [
PackageFactory.PKG_ALBA, PackageFactory.PKG_ALBA_EE
]:
# Retrieve proxy service information
for service in storagerouter.services:
if service.type.name != ServiceType.SERVICE_TYPES.ALBA_PROXY or service.alba_proxy is None:
continue
service_version = None
if package_name not in pkg_component_info:
service_version = ServiceFactory.get_service_update_versions(
client=client,
service_name=service.name,
binary_versions=binaries)
cls._logger.debug(
'StorageRouter {0}: Service {1} is running version {2}'
.format(client.ip, service.name,
service_version))
if package_name in pkg_component_info or service_version is not None:
if service_version is not None and package_name not in svc_component_info[
'packages']:
svc_component_info['packages'][
package_name] = service_version
svc_component_info['services_post_update'][
10].append('ovs-{0}'.format(service.name))
cls._logger.debug(
'StorageRouter {0}: Added service {1} to post-update services'
.format(client.ip,
'ovs-{0}'.format(service.name)))
downtime = [
'proxy',
service.alba_proxy.storagedriver.vpool.name
]
if downtime not in svc_component_info[
'downtime']:
svc_component_info['downtime'].append(
downtime)
cls._logger.debug(
'StorageRouter {0}: Added ALBA proxy downtime for vPool {1} to downtime'
.format(
client.ip, service.alba_proxy.
storagedriver.vpool.name))
if package_name in [
PackageFactory.PKG_ALBA,
PackageFactory.PKG_ALBA_EE,
PackageFactory.PKG_ARAKOON
]:
for service_name, downtime in arakoon_info.iteritems():
service_version = ServiceFactory.get_service_update_versions(
client=client,
service_name=service_name,
binary_versions=binaries,
package_name=package_name)
cls._logger.debug(
'StorageRouter {0}: Arakoon service {1} information: {2}'
.format(client.ip, service_name,
service_version))
if package_name in pkg_component_info or service_version is not None:
svc_component_info['services_post_update'][
10].append('ovs-{0}'.format(service_name))
cls._logger.debug(
'StorageRouter {0}: Added service {1} to post-update services'
.format(client.ip,
'ovs-{0}'.format(service_name)))
if service_version is not None and package_name not in svc_component_info[
'packages']:
svc_component_info['packages'][
package_name] = service_version
if downtime is not None and downtime not in svc_component_info[
'downtime']:
svc_component_info['downtime'].append(
downtime)
cls._logger.debug(
'StorageRouter {0}: Added Arakoon cluster for ALBA Backend {1} to downtime'
.format(client.ip, downtime[1]))
# Extend the service information with the package information related to this repository for current StorageRouter
if package_name in pkg_component_info and package_name not in svc_component_info[
'packages']:
cls._logger.debug(
'StorageRouter {0}: Adding package {1} because it has an update available'
.format(client.ip, package_name))
svc_component_info['packages'][
package_name] = pkg_component_info[package_name]
if component == PackageFactory.COMP_ALBA:
for alba_node in AlbaNodeList.get_albanodes():
try:
alba_node.client.get_metadata()
except:
svc_component_info['prerequisites'].append(
['alba_node_unresponsive', alba_node.ip])
cls._logger.debug(
'StorageRouter {0}: Added unresponsive ALBA Node {1} to prerequisites'
.format(client.ip, alba_node.ip))
# Verify whether migration (DAL and extension) code needs to be executed (only if no packages have an update available so far)
elif component == PackageFactory.COMP_FWK and PackageFactory.PKG_OVS_BACKEND not in svc_component_info[
'packages']:
cls._logger.debug(
'StorageRouter {0}: No updates detected, checking for required migrations'
.format(client.ip))
# Extension migration check
key = '/ovs/framework/hosts/{0}/versions'.format(
System.get_my_machine_id(client=client))
old_version = Configuration.get(key, default={}).get(
PackageFactory.COMP_MIGRATION_ALBA)
installed_version = str(
cls._package_manager.get_installed_versions(
client=client,
package_names=[PackageFactory.PKG_OVS_BACKEND
])[PackageFactory.PKG_OVS_BACKEND])
migrations_detected = False
if old_version is not None:
cls._logger.debug(
'StorageRouter {0}: Current running version for {1} extension migrations: {2}'
.format(client.ip, PackageFactory.COMP_ALBA,
old_version))
with remote(client.ip, [ExtensionMigrator]) as rem:
cls._logger.debug(
'StorageRouter {0}: Available version for {1} extension migrations: {2}'
.format(client.ip, PackageFactory.COMP_ALBA,
rem.ExtensionMigrator.THIS_VERSION))
if rem.ExtensionMigrator.THIS_VERSION > old_version:
migrations_detected = True
svc_component_info['packages'][
PackageFactory.PKG_OVS_BACKEND] = {
'installed': 'migrations',
'candidate': installed_version
}
# DAL migration check
if migrations_detected is False:
persistent_client = PersistentFactory.get_client()
old_version = persistent_client.get(
'ovs_model_version').get(
PackageFactory.COMP_MIGRATION_ALBA
) if persistent_client.exists(
'ovs_model_version') else None
if old_version is not None:
cls._logger.debug(
'StorageRouter {0}: Current running version for {1} DAL migrations: {2}'
.format(client.ip, PackageFactory.COMP_ALBA,
old_version))
with remote(client.ip, [DALMigrator]) as rem:
cls._logger.debug(
'StorageRouter {0}: Available version for {1} DAL migrations: {2}'
.format(client.ip,
PackageFactory.COMP_ALBA,
rem.DALMigrator.THIS_VERSION))
if rem.DALMigrator.THIS_VERSION > old_version:
svc_component_info['packages'][
PackageFactory.PKG_OVS_BACKEND] = {
'installed': 'migrations',
'candidate': installed_version
}
cls._logger.info(
'StorageRouter {0}: Refreshed ALBA update information'.format(
client.ip))
except Exception as ex:
cls._logger.exception(
'StorageRouter {0}: Refreshing ALBA update information failed'.
format(client.ip))
if 'errors' not in update_info[client.ip]:
update_info[client.ip]['errors'] = []
update_info[client.ip]['errors'].append(ex)
def migrate(previous_version):
"""
Migrates from a given version to the current version. It uses 'previous_version' to be smart
wherever possible, but the code should be able to migrate any version towards the expected version.
When this is not possible, the code can set a minimum version and raise when it is not met.
:param previous_version: The previous version from which to start the migration
:type previous_version: float
"""
working_version = previous_version
if working_version == 0:
from ovs.dal.hybrids.servicetype import ServiceType
# Initial version:
# * Add any basic configuration or model entries
# Add backends
for backend_type_info in [('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.NS_MGR,
ServiceType.SERVICE_TYPES.ALBA_MGR,
ServiceType.SERVICE_TYPES.ALBA_S3_TRANSACTION
]:
service_type = ServiceType()
service_type.name = service_type_info
service_type.save()
# From here on, all actual migration should happen to get to the expected state for THIS RELEASE
elif working_version < DALMigrator.THIS_VERSION:
import hashlib
from ovs.dal.exceptions import ObjectNotFoundException
from ovs.dal.helpers import HybridRunner, Descriptor
from ovs.dal.hybrids.albaabmcluster import ABMCluster
from ovs.dal.hybrids.albaosd import AlbaOSD
from ovs.dal.hybrids.albansmcluster import NSMCluster
from ovs.dal.hybrids.j_abmservice import ABMService
from ovs.dal.hybrids.j_nsmservice import NSMService
from ovs.dal.hybrids.service import Service
from ovs.dal.hybrids.servicetype import ServiceType
from ovs.dal.lists.albabackendlist import AlbaBackendList
from ovs.dal.lists.albanodelist import AlbaNodeList
from ovs.dal.lists.servicetypelist import ServiceTypeList
from ovs.dal.lists.storagerouterlist import StorageRouterList
from ovs.extensions.db.arakooninstaller import ArakoonClusterConfig, ArakoonInstaller
from ovs.extensions.generic.configuration import Configuration, NotFoundException
from ovs_extensions.generic.toolbox import ExtensionsToolbox
from ovs.extensions.plugins.albacli import AlbaCLI
from ovs.extensions.storage.persistentfactory import PersistentFactory
# Migrate unique constraints & indexes
client = PersistentFactory.get_client()
hybrid_structure = HybridRunner.get_hybrids()
for class_descriptor in hybrid_structure.values():
cls = Descriptor().load(class_descriptor).get_object()
classname = cls.__name__.lower()
unique_key = 'ovs_unique_{0}_{{0}}_'.format(classname)
index_prefix = 'ovs_index_{0}|{{0}}|'.format(classname)
index_key = 'ovs_index_{0}|{{0}}|{{1}}'.format(classname)
uniques = []
indexes = []
# noinspection PyProtectedMember
for prop in cls._properties:
if prop.unique is True and len([
k for k in client.prefix(
unique_key.format(prop.name))
]) == 0:
uniques.append(prop.name)
if prop.indexed is True and len([
k for k in client.prefix(
index_prefix.format(prop.name))
]) == 0:
indexes.append(prop.name)
if len(uniques) > 0 or len(indexes) > 0:
prefix = 'ovs_data_{0}_'.format(classname)
for key, data in client.prefix_entries(prefix):
for property_name in uniques:
ukey = '{0}{1}'.format(
unique_key.format(property_name),
hashlib.sha1(str(
data[property_name])).hexdigest())
client.set(ukey, key)
for property_name in indexes:
if property_name not in data:
continue # This is the case when there's a new indexed property added.
ikey = index_key.format(
property_name,
hashlib.sha1(str(
data[property_name])).hexdigest())
index = list(
client.get_multi([ikey], must_exist=False))[0]
transaction = client.begin_transaction()
if index is None:
client.assert_value(ikey,
None,
transaction=transaction)
client.set(ikey, [key],
transaction=transaction)
elif key not in index:
client.assert_value(ikey,
index[:],
transaction=transaction)
client.set(ikey,
index + [key],
transaction=transaction)
client.apply_transaction(transaction)
#############################################
# Introduction of ABMCluster and NSMCluster #
#############################################
# Verify presence of unchanged ALBA Backends
alba_backends = AlbaBackendList.get_albabackends()
changes_required = False
for alba_backend in alba_backends:
if alba_backend.abm_cluster is None or len(
alba_backend.nsm_clusters) == 0:
changes_required = True
break
if changes_required:
# Retrieve ABM and NSM clusters
abm_cluster_info = []
nsm_cluster_info = []
for cluster_name in Configuration.list('/ovs/arakoon'):
try:
metadata = ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=cluster_name)
if metadata[
'cluster_type'] == ServiceType.ARAKOON_CLUSTER_TYPES.ABM:
abm_cluster_info.append(metadata)
elif metadata[
'cluster_type'] == ServiceType.ARAKOON_CLUSTER_TYPES.NSM:
nsm_cluster_info.append(metadata)
except NotFoundException:
continue
# Retrieve NSM Arakoon cluster information
cluster_arakoon_map = {}
for cluster_info in abm_cluster_info + nsm_cluster_info:
cluster_name = cluster_info['cluster_name']
arakoon_config = ArakoonClusterConfig(
cluster_id=cluster_name)
cluster_arakoon_map[
cluster_name] = arakoon_config.export_dict()
storagerouter_map = dict(
(storagerouter.machine_id, storagerouter) for storagerouter
in StorageRouterList.get_storagerouters())
alba_backend_id_map = dict((alba_backend.alba_id, alba_backend)
for alba_backend in alba_backends)
for cluster_info in abm_cluster_info:
internal = cluster_info['internal']
cluster_name = cluster_info['cluster_name']
config_location = Configuration.get_configuration_path(
key=ArakoonClusterConfig.CONFIG_KEY.format(
cluster_name))
try:
alba_id = AlbaCLI.run(command='get-alba-id',
config=config_location,
named_params={'attempts':
3})['id']
nsm_hosts = AlbaCLI.run(command='list-nsm-hosts',
config=config_location,
named_params={'attempts': 3})
except RuntimeError:
continue
alba_backend = alba_backend_id_map.get(alba_id)
if alba_backend is None: # ALBA Backend with ID not found in model
continue
if alba_backend.abm_cluster is not None and len(
alba_backend.nsm_clusters
) > 0: # Clusters already exist
continue
# Create ABM Cluster
if alba_backend.abm_cluster is None:
abm_cluster = ABMCluster()
abm_cluster.name = cluster_name
abm_cluster.alba_backend = alba_backend
abm_cluster.config_location = ArakoonClusterConfig.CONFIG_KEY.format(
cluster_name)
abm_cluster.save()
else:
abm_cluster = alba_backend.abm_cluster
# Create ABM Services
abm_arakoon_config = cluster_arakoon_map[cluster_name]
abm_arakoon_config.pop('global')
arakoon_nodes = abm_arakoon_config.keys()
if internal is False:
services_to_create = 1
else:
if set(arakoon_nodes).difference(
set(storagerouter_map.keys())):
continue
services_to_create = len(arakoon_nodes)
for index in range(services_to_create):
service = Service()
service.name = 'arakoon-{0}-abm'.format(
alba_backend.name)
service.type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.ALBA_MGR)
if internal is True:
arakoon_node_config = abm_arakoon_config[
arakoon_nodes[index]]
service.ports = [
arakoon_node_config['client_port'],
arakoon_node_config['messaging_port']
]
service.storagerouter = storagerouter_map[
arakoon_nodes[index]]
else:
service.ports = []
service.storagerouter = None
service.save()
abm_service = ABMService()
abm_service.service = service
abm_service.abm_cluster = abm_cluster
abm_service.save()
# Create NSM Clusters
for cluster_index, nsm_host in enumerate(
sorted(nsm_hosts,
key=lambda host: ExtensionsToolbox.
advanced_sort(host['cluster_id'], '_'))):
nsm_cluster_name = nsm_host['cluster_id']
nsm_arakoon_config = cluster_arakoon_map.get(
nsm_cluster_name)
if nsm_arakoon_config is None:
continue
number = cluster_index if internal is False else int(
nsm_cluster_name.split('_')[-1])
nsm_cluster = NSMCluster()
nsm_cluster.name = nsm_cluster_name
nsm_cluster.number = number
nsm_cluster.alba_backend = alba_backend
nsm_cluster.config_location = ArakoonClusterConfig.CONFIG_KEY.format(
nsm_cluster_name)
nsm_cluster.save()
# Create NSM Services
nsm_arakoon_config.pop('global')
arakoon_nodes = nsm_arakoon_config.keys()
if internal is False:
services_to_create = 1
else:
if set(arakoon_nodes).difference(
set(storagerouter_map.keys())):
continue
services_to_create = len(arakoon_nodes)
for service_index in range(services_to_create):
service = Service()
service.name = 'arakoon-{0}-nsm_{1}'.format(
alba_backend.name, number)
service.type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.NS_MGR)
if internal is True:
arakoon_node_config = nsm_arakoon_config[
arakoon_nodes[service_index]]
service.ports = [
arakoon_node_config['client_port'],
arakoon_node_config['messaging_port']
]
service.storagerouter = storagerouter_map[
arakoon_nodes[service_index]]
else:
service.ports = []
service.storagerouter = None
service.save()
nsm_service = NSMService()
nsm_service.service = service
nsm_service.nsm_cluster = nsm_cluster
nsm_service.save()
# Clean up all junction services no longer linked to an ALBA Backend
all_nsm_services = [
service.nsm_service for service in ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.NS_MGR).services
if service.nsm_service.nsm_cluster is None
]
all_abm_services = [
service.abm_service for service in ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.ALBA_MGR).services
if service.abm_service.abm_cluster is None
]
for abm_service in all_abm_services:
abm_service.delete()
abm_service.service.delete()
for nsm_service in all_nsm_services:
nsm_service.delete()
nsm_service.service.delete()
################################
# Introduction of Active Drive #
################################
# Update slot_id and Alba Node relation for all OSDs
client = PersistentFactory.get_client()
disk_osd_map = {}
for key, data in client.prefix_entries('ovs_data_albaosd_'):
alba_disk_guid = data.get('alba_disk', {}).get('guid')
if alba_disk_guid is not None:
if alba_disk_guid not in disk_osd_map:
disk_osd_map[alba_disk_guid] = []
disk_osd_map[alba_disk_guid].append(
key.replace('ovs_data_albaosd_', ''))
try:
value = client.get(key)
value.pop('alba_disk', None)
client.set(key=key, value=value)
except Exception:
pass # We don't care if we would have any leftover AlbaDisk information in _data, but its cleaner not to
alba_guid_node_map = dict(
(an.guid, an) for an in AlbaNodeList.get_albanodes())
for key, data in client.prefix_entries('ovs_data_albadisk_'):
alba_disk_guid = key.replace('ovs_data_albadisk_', '')
alba_node_guid = data.get('alba_node', {}).get('guid')
if alba_disk_guid in disk_osd_map and alba_node_guid in alba_guid_node_map and len(
data.get('aliases', [])) > 0:
slot_id = data['aliases'][0].split('/')[-1]
for osd_guid in disk_osd_map[alba_disk_guid]:
try:
osd = AlbaOSD(osd_guid)
except ObjectNotFoundException:
continue
osd.slot_id = slot_id
osd.alba_node = alba_guid_node_map[alba_node_guid]
osd.save()
client.delete(key=key, must_exist=False)
# Remove unique constraints for AlbaNode IP
for key in client.prefix('ovs_unique_albanode_ip_'):
client.delete(key=key, must_exist=False)
# Remove relation for all Alba Disks
for key in client.prefix('ovs_reverseindex_albadisk_'):
client.delete(key=key, must_exist=False)
# Remove the relation between AlbaNode and AlbaDisk
for key in client.prefix('ovs_reverseindex_albanode_'):
if '|disks|' in key:
client.delete(key=key, must_exist=False)
return DALMigrator.THIS_VERSION
def add_arakoon(cluster_name,
storagerouter_ip,
cluster_basedir,
service_type=ServiceType.ARAKOON_CLUSTER_TYPES.FWK):
"""
Adds a external arakoon to a storagerouter
:param cluster_name: name of the new arakoon cluster
:type cluster_name: str
:param service_type: type of plugin for arakoon (DEFAULT=ServiceType.ARAKOON_CLUSTER_TYPES.FWK)
* FWK
* ABM
* NSM
:type service_type: ovs.dal.hybrids.ServiceType.ARAKOON_CLUSTER_TYPES
:param storagerouter_ip: ip of a storagerouter
:type storagerouter_ip: str
:param cluster_basedir: absolute path for the new arakoon cluster
:type cluster_basedir: str
:return:
"""
client = SSHClient(storagerouter_ip, username='******')
# create required directories
if not client.dir_exists(cluster_basedir):
client.dir_create(cluster_basedir)
# determine plugin
if service_type == ServiceType.ARAKOON_CLUSTER_TYPES.FWK:
plugins = None
elif service_type == ServiceType.ARAKOON_CLUSTER_TYPES.ABM:
plugins = {
AlbaController.ABM_PLUGIN: AlbaController.ALBA_VERSION_GET
}
elif service_type == ServiceType.ARAKOON_CLUSTER_TYPES.NSM:
plugins = {
AlbaController.NSM_PLUGIN: AlbaController.ALBA_VERSION_GET
}
else:
raise RuntimeError(
"Incompatible Arakoon cluster type selected: {0}".format(
service_type))
ArakoonSetup.LOGGER.info(
"Starting creation of new arakoon cluster with name `{0}`, servicetype `{1}`, ip `{2}`, base_dir `{3}`"
.format(cluster_name, service_type, storagerouter_ip,
cluster_basedir))
arakoon_installer = ArakoonInstaller(cluster_name)
arakoon_installer.create_cluster(
cluster_type=service_type,
ip=storagerouter_ip,
base_dir=cluster_basedir,
plugins=plugins,
locked=False,
internal=False,
log_sinks=Logger.get_sink_path('automation_lib_arakoon_server'),
crash_log_sinks=Logger.get_sink_path(
'automation_lib_arakoon_server_crash'))
if service_type == ServiceType.ARAKOON_CLUSTER_TYPES.ABM:
client.run([
'ln', '-s', '/usr/lib/alba/albamgr_plugin.cmxs',
'{0}/arakoon/{1}/db'.format(cluster_basedir, cluster_name)
])
elif service_type == ServiceType.ARAKOON_CLUSTER_TYPES.NSM:
client.run([
'ln', '-s', '/usr/lib/alba/nsm_host_plugin.cmxs',
'{0}/arakoon/{1}/db'.format(cluster_basedir, cluster_name)
])
arakoon_installer.start_cluster()
arakoon_installer.unclaim_cluster()
ArakoonSetup.LOGGER.info(
"Finished creation of new arakoon cluster with name `{0}`, servicetype `{1}`, ip `{2}`, base_dir `{3}`"
.format(cluster_name, service_type, storagerouter_ip,
cluster_basedir))
def test_alba_arakoon_checkup(self):
"""
Validates whether the ALBA Arakoon checkup works (Manual and Scheduled)
"""
ovs_structure = DalHelper.build_dal_structure(
structure={'storagerouters': [1]})
alba_structure = AlbaDalHelper.build_dal_structure(
structure={'alba_backends': [[1, 'LOCAL']]})
#############################
# SCHEDULED_ARAKOON_CHECKUP #
#############################
# Create an ABM and NSM cluster for ALBA Backend 1 and do some basic validations
sr_1 = ovs_structure['storagerouters'][1]
ab_1 = alba_structure['alba_backends'][1]
MockedSSHClient._run_returns[sr_1.ip] = {}
MockedSSHClient._run_returns[sr_1.ip][
'ln -s /usr/lib/alba/albamgr_plugin.cmxs /tmp/unittest/sr_1/disk_1/partition_1/arakoon/backend_1-abm/db'] = None
MockedSSHClient._run_returns[sr_1.ip][
'ln -s /usr/lib/alba/nsm_host_plugin.cmxs /tmp/unittest/sr_1/disk_1/partition_1/arakoon/backend_1-nsm_0/db'] = None
AlbaController.add_cluster(ab_1.guid)
abm_cluster_name = '{0}-abm'.format(ab_1.name)
nsm_cluster_name = '{0}-nsm_0'.format(ab_1.name)
arakoon_clusters = sorted(Configuration.list('/ovs/arakoon'))
self.assertListEqual(list1=[abm_cluster_name, nsm_cluster_name],
list2=arakoon_clusters)
abm_metadata = ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=abm_cluster_name)
nsm_metadata = ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=nsm_cluster_name)
self.assertTrue(expr=abm_metadata['in_use'])
self.assertTrue(expr=nsm_metadata['in_use'])
# Run scheduled Arakoon checkup and validate amount of Arakoon clusters did not change
AlbaArakoonController.scheduled_alba_arakoon_checkup()
self.assertListEqual(list1=[abm_cluster_name, nsm_cluster_name],
list2=arakoon_clusters)
self.assertEqual(first=len(ab_1.abm_cluster.abm_services), second=1)
self.assertEqual(first=len(ab_1.nsm_clusters), second=1)
self.assertEqual(first=len(ab_1.nsm_clusters[0].nsm_services),
second=1)
# Create 2 additional StorageRouters
srs = DalHelper.build_dal_structure(
structure={'storagerouters': [2, 3]},
previous_structure=ovs_structure)['storagerouters']
sr_2 = srs[2]
sr_3 = srs[3]
# Run scheduled checkup again and do some validations
MockedSSHClient._run_returns[sr_2.ip] = {}
MockedSSHClient._run_returns[sr_3.ip] = {}
MockedSSHClient._run_returns[sr_2.ip][
'ln -s /usr/lib/alba/albamgr_plugin.cmxs /tmp/unittest/sr_2/disk_1/partition_1/arakoon/backend_1-abm/db'] = None
MockedSSHClient._run_returns[sr_3.ip][
'ln -s /usr/lib/alba/albamgr_plugin.cmxs /tmp/unittest/sr_3/disk_1/partition_1/arakoon/backend_1-abm/db'] = None
MockedSSHClient._run_returns[sr_2.ip][
'arakoon --node {0} -config file://opt/OpenvStorage/config/framework.json?key=/ovs/arakoon/backend_1-abm/config -catchup-only'
.format(sr_2.machine_id)] = None
MockedSSHClient._run_returns[sr_3.ip][
'arakoon --node {0} -config file://opt/OpenvStorage/config/framework.json?key=/ovs/arakoon/backend_1-abm/config -catchup-only'
.format(sr_3.machine_id)] = None
AlbaArakoonController.scheduled_alba_arakoon_checkup()
self.assertListEqual(list1=[abm_cluster_name, nsm_cluster_name],
list2=arakoon_clusters)
self.assertEqual(first=len(ab_1.abm_cluster.abm_services),
second=3) # Gone up from 1 to 3
self.assertEqual(first=len(ab_1.nsm_clusters), second=1)
self.assertEqual(first=len(ab_1.nsm_clusters[0].nsm_services),
second=1) # Still 1 since NSM checkup hasn't ran yet
# Make sure 1 StorageRouter is unreachable
SSHClient._raise_exceptions[sr_3.ip] = {
'users': ['ovs'],
'exception': UnableToConnectException('No route to host')
}
AlbaArakoonController.scheduled_alba_arakoon_checkup()
alba_logs = Logger._logs.get('lib', [])
self.assertIn(
member='Storage Router with IP {0} is not reachable'.format(
sr_3.ip),
container=alba_logs)
##########################
# MANUAL_ARAKOON_CHECKUP #
##########################
AlbaDalHelper.setup() # Clear everything
ovs_structure = DalHelper.build_dal_structure(
structure={'storagerouters': [1]})
alba_structure = AlbaDalHelper.build_dal_structure(
structure={'alba_backends': [[1, 'LOCAL']]})
sr_1 = ovs_structure['storagerouters'][1]
ab_1 = alba_structure['alba_backends'][1]
MockedSSHClient._run_returns[sr_1.ip] = {}
MockedSSHClient._run_returns[sr_1.ip][
'ln -s /usr/lib/alba/albamgr_plugin.cmxs /tmp/unittest/sr_1/disk_1/partition_1/arakoon/backend_1-abm/db'] = None
MockedSSHClient._run_returns[sr_1.ip][
'ln -s /usr/lib/alba/nsm_host_plugin.cmxs /tmp/unittest/sr_1/disk_1/partition_1/arakoon/backend_1-nsm_0/db'] = None
AlbaController.add_cluster(ab_1.guid)
# Run manual Arakoon checkup and validate amount of Arakoon clusters did not change
AlbaArakoonController.manual_alba_arakoon_checkup(
alba_backend_guid=ab_1.guid, nsm_clusters=[], abm_cluster=None)
self.assertListEqual(list1=[abm_cluster_name, nsm_cluster_name],
list2=arakoon_clusters)
self.assertEqual(first=len(ab_1.abm_cluster.abm_services), second=1)
self.assertEqual(first=len(ab_1.nsm_clusters), second=1)
self.assertEqual(first=len(ab_1.nsm_clusters[0].nsm_services),
second=1)
# Test some error paths
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.manual_alba_arakoon_checkup(
alba_backend_guid=ab_1.guid,
nsm_clusters=['no_abm_cluster_passed'])
self.assertEqual(
first=raise_info.exception.message,
second='Both ABM cluster and NSM clusters must be provided')
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.manual_alba_arakoon_checkup(
alba_backend_guid=ab_1.guid,
nsm_clusters=[],
abm_cluster='no_nsm_clusters_passed')
self.assertEqual(
first=raise_info.exception.message,
second='Both ABM cluster and NSM clusters must be provided')
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.manual_alba_arakoon_checkup(
alba_backend_guid=ab_1.guid,
nsm_clusters=[nsm_cluster_name],
abm_cluster=abm_cluster_name)
self.assertEqual(first=raise_info.exception.message,
second='Cluster {0} has already been claimed'.format(
abm_cluster_name))
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.manual_alba_arakoon_checkup(
alba_backend_guid=ab_1.guid,
nsm_clusters=['non-existing-nsm-cluster'],
abm_cluster='non-existing-abm-cluster')
self.assertEqual(
first=raise_info.exception.message,
second=
'Could not find an Arakoon cluster with name: non-existing-abm-cluster'
)
# Recreate ALBA Backend with Arakoon clusters
AlbaDalHelper.setup() # Clear everything
ovs_structure = DalHelper.build_dal_structure(
structure={'storagerouters': [1]})
alba_structure = AlbaDalHelper.build_dal_structure(
structure={'alba_backends': [[1, 'LOCAL']]})
sr_1 = ovs_structure['storagerouters'][1]
ab_1 = alba_structure['alba_backends'][1]
# Create some Arakoon clusters to be claimed by the manual checkup
for cluster_name, cluster_type in {
'manual-abm-1': ServiceType.ARAKOON_CLUSTER_TYPES.ABM,
'manual-abm-2': ServiceType.ARAKOON_CLUSTER_TYPES.ABM,
'manual-nsm-1': ServiceType.ARAKOON_CLUSTER_TYPES.NSM,
'manual-nsm-2': ServiceType.ARAKOON_CLUSTER_TYPES.NSM,
'manual-nsm-3': ServiceType.ARAKOON_CLUSTER_TYPES.NSM
}.iteritems():
arakoon_installer = ArakoonInstaller(cluster_name=cluster_name)
arakoon_installer.create_cluster(
cluster_type=cluster_type,
ip=sr_1.ip,
base_dir=DalHelper.CLUSTER_DIR.format(cluster_name),
internal=False)
arakoon_installer.start_cluster()
arakoon_installer.unclaim_cluster()
AlbaArakoonController.manual_alba_arakoon_checkup(
alba_backend_guid=ab_1.guid,
nsm_clusters=['manual-nsm-1', 'manual-nsm-3'],
abm_cluster='manual-abm-2')
# Validate the correct clusters have been claimed by the manual checkup
unused_abms = ArakoonInstaller.get_unused_arakoon_clusters(
cluster_type=ServiceType.ARAKOON_CLUSTER_TYPES.ABM)
unused_nsms = ArakoonInstaller.get_unused_arakoon_clusters(
cluster_type=ServiceType.ARAKOON_CLUSTER_TYPES.NSM)
self.assertEqual(first=len(unused_abms), second=1)
self.assertEqual(first=len(unused_nsms), second=1)
self.assertEqual(first=unused_abms[0]['cluster_name'],
second='manual-abm-1')
self.assertEqual(first=unused_nsms[0]['cluster_name'],
second='manual-nsm-2')
def nsm_checkup(alba_backend_guid=None,
min_internal_nsms=1,
external_nsm_cluster_names=None):
# type: (Optional[str], Optional[int], Optional[List[str]]) -> None
"""
Validates the current NSM setup/configuration and takes actions where required.
Assumptions:
* A 2 node NSM is considered safer than a 1 node NSM.
* When adding an NSM, the nodes with the least amount of NSM participation are preferred
:param alba_backend_guid: Run for a specific ALBA Backend
:type alba_backend_guid: str
:param min_internal_nsms: Minimum amount of NSM hosts that need to be provided
:type min_internal_nsms: int
:param external_nsm_cluster_names: Information about the additional clusters to claim (only for externally managed Arakoon clusters)
:type external_nsm_cluster_names: list
:return: None
:rtype: NoneType
"""
###############
# Validations #
###############
if external_nsm_cluster_names is None:
external_nsm_cluster_names = []
AlbaArakoonController._logger.info('NSM checkup started')
if min_internal_nsms < 1:
raise ValueError(
'Minimum amount of NSM clusters must be 1 or more')
if not isinstance(external_nsm_cluster_names, list):
raise ValueError(
"'external_nsm_cluster_names' must be of type 'list'")
if len(external_nsm_cluster_names) > 0:
if alba_backend_guid is None:
raise ValueError(
'Additional NSMs can only be configured for a specific ALBA Backend'
)
if min_internal_nsms > 1:
raise ValueError(
"'min_internal_nsms' and 'external_nsm_cluster_names' are mutually exclusive"
)
external_nsm_cluster_names = list(set(
external_nsm_cluster_names)) # Remove duplicate cluster names
for cluster_name in external_nsm_cluster_names:
try:
ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=cluster_name)
except NotFoundException:
raise ValueError(
'Arakoon cluster with name {0} does not exist'.format(
cluster_name))
if alba_backend_guid is None:
alba_backends = [
alba_backend
for alba_backend in AlbaBackendList.get_albabackends()
if alba_backend.backend.status == 'RUNNING'
]
else:
alba_backends = [AlbaBackend(alba_backend_guid)]
masters = StorageRouterList.get_masters()
storagerouters = set()
for alba_backend in alba_backends:
if alba_backend.abm_cluster is None:
raise ValueError(
'No ABM cluster found for ALBA Backend {0}'.format(
alba_backend.name))
if len(alba_backend.abm_cluster.abm_services) == 0:
raise ValueError(
'ALBA Backend {0} does not have any registered ABM services'
.format(alba_backend.name))
if len(alba_backend.nsm_clusters) + len(
external_nsm_cluster_names) > MAX_NSM_AMOUNT:
raise ValueError(
'The maximum of {0} NSM Arakoon clusters will be exceeded. Amount of clusters that can be deployed for this ALBA Backend: {1}'
.format(MAX_NSM_AMOUNT,
MAX_NSM_AMOUNT - len(alba_backend.nsm_clusters)))
# Validate enough externally managed Arakoon clusters are available
if alba_backend.abm_cluster.abm_services[
0].service.is_internal is False:
unused_cluster_names = set([
cluster_info['cluster_name'] for cluster_info in
ArakoonInstaller.get_unused_arakoon_clusters(
cluster_type=ServiceType.ARAKOON_CLUSTER_TYPES.NSM)
])
if set(external_nsm_cluster_names).difference(
unused_cluster_names):
raise ValueError(
'Some of the provided cluster_names have already been claimed before'
)
storagerouters.update(
set(masters)
) # For externally managed we need an available master node
else:
for abm_service in alba_backend.abm_cluster.abm_services: # For internally managed we need all StorageRouters online
storagerouters.add(abm_service.service.storagerouter)
for nsm_cluster in alba_backend.nsm_clusters: # For internally managed we need all StorageRouters online
for nsm_service in nsm_cluster.nsm_services:
storagerouters.add(nsm_service.service.storagerouter)
ssh_clients = {}
for storagerouter in storagerouters:
try:
ssh_clients[storagerouter] = SSHClient(endpoint=storagerouter)
except UnableToConnectException:
raise RuntimeError(
'StorageRouter {0} with IP {1} is not reachable'.format(
storagerouter.name, storagerouter.ip))
version_str = AlbaArakoonInstaller.get_alba_version_string()
nsm_installer = NSMInstaller(version_str=version_str,
ssh_clients=ssh_clients)
##################
# Check Clusters #
##################
safety = Configuration.get(
'/ovs/framework/plugins/alba/config|nsm.safety')
maxload = Configuration.get(
'/ovs/framework/plugins/alba/config|nsm.maxload')
AlbaArakoonController._logger.debug(
'NSM safety is configured at: {0}'.format(safety))
AlbaArakoonController._logger.debug(
'NSM max load is configured at: {0}'.format(maxload))
master_client = None
failed_backends = []
for alba_backend in alba_backends:
try:
# Gather information
AlbaArakoonController._logger.info(
'ALBA Backend {0} - Ensuring NSM safety'.format(
alba_backend.name))
internal = AlbaArakoonInstaller.is_internally_managed(
alba_backend)
nsm_loads = AlbaArakoonController.get_nsm_loads(alba_backend)
nsm_storagerouters = AlbaArakoonController.get_nsms_per_storagerouter(
alba_backend)
sorted_nsm_clusters = sorted(alba_backend.nsm_clusters,
key=lambda k: k.number)
if not internal and len(external_nsm_cluster_names) > 0:
for sr, cl in ssh_clients.iteritems():
if sr.node_type == 'MASTER':
master_client = cl
break
if master_client is None:
# Internal is False and we specified the NSM clusters to claim, but no MASTER nodes online
raise ValueError(
'Could not find an online master node')
AlbaArakoonController._logger.debug(
'ALBA Backend {0} - Arakoon clusters are {1} managed'.
format(alba_backend.name,
'internally' if internal is True else 'externally'))
for nsm_number, nsm_load in nsm_loads.iteritems():
AlbaArakoonController._logger.debug(
'ALBA Backend {0} - NSM Cluster {1} - Load {2}'.format(
alba_backend.name, nsm_number, nsm_load))
for sr, count in nsm_storagerouters.iteritems():
AlbaArakoonController._logger.debug(
'ALBA Backend {0} - StorageRouter {1} - NSM Services {2}'
.format(alba_backend.name, sr.name, count))
if internal:
# Extend existing NSM clusters if safety not met
for nsm_cluster in sorted_nsm_clusters:
AlbaArakoonController._logger.debug(
'ALBA Backend {0} - Processing NSM {1} - Expected safety {2} - Current safety {3}'
.format(alba_backend.name, nsm_cluster.number,
safety, len(nsm_cluster.nsm_services)))
AlbaArakoonController.ensure_nsm_cluster_safety(
nsm_cluster,
nsm_storagerouters,
nsm_installer=nsm_installer)
AlbaArakoonController.ensure_nsm_clusters_load(
alba_backend,
nsms_per_storagerouter=nsm_storagerouters,
ssh_clients=ssh_clients,
version_str=version_str,
min_internal_nsms=min_internal_nsms,
external_nsm_cluster_names=external_nsm_cluster_names)
except Exception:
AlbaArakoonController._logger.exception(
'NSM Checkup failed for Backend {0}'.format(
alba_backend.name))
failed_backends.append(alba_backend.name)
def ensure_nsm_clusters_load(cls,
alba_backend,
nsms_per_storagerouter=None,
min_internal_nsms=1,
external_nsm_cluster_names=None,
version_str=None,
ssh_clients=None):
# type: (AlbaBackend, Optional[Dict[StorageRouter, int]], Optional[int], Optional[List[str], Optional[str]], Optional[StorageRouter, SSHClient]) -> None
"""
Ensure that all NSM clusters are not overloaded
:param alba_backend: Alba Backend to ensure NSM Cluster load for
:type alba_backend: AlbaBackend
:param nsms_per_storagerouter: Amount of NSMs mapped by StorageRouter
:type nsms_per_storagerouter: Dict[StorageRouter, int]
:param min_internal_nsms: Minimum amount of NSM hosts that need to be provided
:type min_internal_nsms: int
:param external_nsm_cluster_names: Information about the additional clusters to claim (only for externally managed Arakoon clusters)
:type external_nsm_cluster_names: list
:param version_str: Alba version string
:type version_str: str
:param ssh_clients: SSHClients to use
:type ssh_clients: Dict[Storagerouter, SSHClient]
:return: None
:rtype: NoneType
"""
if ssh_clients is None:
ssh_clients = {}
if external_nsm_cluster_names is None:
external_nsm_cluster_names = []
nsms_per_storagerouter = nsms_per_storagerouter if nsms_per_storagerouter is not None else cls.get_nsms_per_storagerouter(
alba_backend)
version_str = version_str or AlbaArakoonInstaller.get_alba_version_string(
)
nsm_loads = cls.get_nsm_loads(alba_backend)
internal = AlbaArakoonInstaller.is_internally_managed(alba_backend)
abm_cluster_name = alba_backend.abm_cluster.name
safety = Configuration.get(
'/ovs/framework/plugins/alba/config|nsm.safety')
maxload = Configuration.get(
'/ovs/framework/plugins/alba/config|nsm.maxload')
overloaded = min(nsm_loads.values()) >= maxload
if not overloaded:
# At least 1 NSM is not overloaded yet
AlbaArakoonController._logger.debug(
'ALBA Backend {0} - NSM load OK'.format(alba_backend.name))
if internal:
# When load is not OK, deploy at least 1 additional NSM
nsms_to_add = max(0, min_internal_nsms - len(nsm_loads))
else:
nsms_to_add = len(external_nsm_cluster_names)
if nsms_to_add == 0:
return
else:
AlbaArakoonController._logger.warning(
'ALBA Backend {0} - NSM load is NOT OK'.format(
alba_backend.name))
if internal:
# When load is not OK, deploy at least 1 additional NSM
nsms_to_add = max(1, min_internal_nsms - len(nsm_loads))
else:
# For externally managed clusters we only claim the specified clusters, if none provided, we just log it
nsms_to_add = len(external_nsm_cluster_names)
if nsms_to_add == 0:
cls._logger.critical(
'ALBA Backend {0} - All NSM clusters are overloaded'.
format(alba_backend.name))
return
# Deploy new (internal) or claim existing (external) NSM clusters
cls._logger.debug(
'ALBA Backend {0} - Currently {1} NSM cluster{2}'.format(
alba_backend.name, len(nsm_loads),
'' if len(nsm_loads) == 1 else 's'))
AlbaArakoonController._logger.debug(
'ALBA Backend {0} - Trying to add {1} NSM cluster{2}'.format(
alba_backend.name, nsms_to_add,
'' if nsms_to_add == 1 else 's'))
base_number = max(nsm_loads.keys()) + 1
for index, number in enumerate(
xrange(base_number, base_number + nsms_to_add)):
if not internal:
# External clusters
master_client = None
if not ssh_clients:
for storagerouter in StorageRouterList.get_masters():
try:
master_client = SSHClient(storagerouter)
except UnableToConnectException:
cls._logger.warning(
'StorageRouter {0} with IP {1} is not reachable'
.format(storagerouter.name, storagerouter.ip))
else:
for storagerouter, ssh_client in ssh_clients.iteritems():
if storagerouter.node_type == 'MASTER':
master_client = ssh_client
if not master_client:
raise ValueError('Could not find an online master node')
# @todo this might raise an indexerror?
nsm_cluster_name = external_nsm_cluster_names[index]
cls._logger.debug(
'ALBA Backend {0} - Claiming NSM cluster {1}'.format(
alba_backend.name, nsm_cluster_name))
metadata = ArakoonInstaller.get_unused_arakoon_metadata_and_claim(
cluster_type=ServiceType.ARAKOON_CLUSTER_TYPES.NSM,
cluster_name=nsm_cluster_name)
if metadata is None:
cls._logger.critical(
'ALBA Backend {0} - NSM cluster with name {1} could not be found'
.format(alba_backend.name, nsm_cluster_name))
continue
cls._logger.debug(
'ALBA Backend {0} - Modeling services'.format(
alba_backend.name))
AlbaArakoonInstaller.model_arakoon_service(
alba_backend=alba_backend,
cluster_name=nsm_cluster_name,
number=number)
cls._logger.debug('ALBA Backend {0} - Registering NSM'.format(
alba_backend.name))
NSMInstaller.register_nsm(abm_name=abm_cluster_name,
nsm_name=nsm_cluster_name,
ip=master_client.ip)
AlbaArakoonController._logger.debug(
'ALBA Backend {0} - Extended cluster'.format(
alba_backend.name))
else:
# Internal clusters
nsm_cluster_name = '{0}-nsm_{1}'.format(
alba_backend.name, number)
cls._logger.debug(
'ALBA Backend {0} - Adding NSM cluster {1}'.format(
alba_backend.name, nsm_cluster_name))
# One of the NSM nodes is overloaded. This means the complete NSM is considered overloaded
# Figure out which StorageRouters are the least occupied
loads = sorted(nsms_per_storagerouter.values())[:safety]
storagerouters = []
for storagerouter, load in nsms_per_storagerouter.iteritems():
if load in loads:
storagerouters.append(storagerouter)
if len(storagerouters) == safety:
break
# Creating a new NSM cluster
for sub_index, storagerouter in enumerate(storagerouters):
nsms_per_storagerouter[storagerouter] += 1
partition = AlbaArakoonInstaller.get_db_partition(
storagerouter)
arakoon_installer = ArakoonInstaller(
cluster_name=nsm_cluster_name)
# @todo Use deploy and extend code. (Disable register nsm in those parts)
if sub_index == 0:
arakoon_installer.create_cluster(
cluster_type=ServiceType.ARAKOON_CLUSTER_TYPES.NSM,
ip=storagerouter.ip,
base_dir=partition.folder,
plugins={NSM_PLUGIN: version_str})
else:
cls._logger.debug(
'ALBA Backend {0} - Extending NSM cluster {1}'.
format(alba_backend.name, nsm_cluster_name))
arakoon_installer.load()
arakoon_installer.extend_cluster(
new_ip=storagerouter.ip,
base_dir=partition.folder,
plugins={NSM_PLUGIN: version_str})
cls._logger.debug(
'ALBA Backend {0} - Linking plugins'.format(
alba_backend.name))
ssh_client = ssh_clients.get(storagerouter) or SSHClient(
StorageRouter)
AlbaArakoonInstaller.link_plugins(
client=ssh_client,
data_dir=partition.folder,
plugins=[NSM_PLUGIN],
cluster_name=nsm_cluster_name)
cls._logger.debug(
'ALBA Backend {0} - Modeling services'.format(
alba_backend.name))
AlbaArakoonInstaller.model_arakoon_service(
alba_backend=alba_backend,
cluster_name=nsm_cluster_name,
ports=arakoon_installer.ports[storagerouter.ip],
storagerouter=storagerouter,
number=number)
if sub_index == 0:
cls._logger.debug(
'ALBA Backend {0} - Starting cluster'.format(
alba_backend.name))
arakoon_installer.start_cluster()
else:
AlbaArakoonController._logger.debug(
'ALBA Backend {0} - Restarting cluster'.format(
alba_backend.name))
arakoon_installer.restart_cluster_after_extending(
new_ip=storagerouter.ip)
cls._logger.debug('ALBA Backend {0} - Registering NSM'.format(
alba_backend.name))
NSMInstaller.register_nsm(abm_name=abm_cluster_name,
nsm_name=nsm_cluster_name,
ip=storagerouters[0].ip)
cls._logger.debug(
'ALBA Backend {0} - Added NSM cluster {1}'.format(
alba_backend.name, nsm_cluster_name))
def _voldrv_arakoon_checkup(create_cluster):
def _add_service(service_storagerouter, arakoon_ports, service_name):
""" Add a service to the storage router """
new_service = Service()
new_service.name = service_name
new_service.type = service_type
new_service.ports = arakoon_ports
new_service.storagerouter = service_storagerouter
new_service.save()
return new_service
current_ips = []
current_services = []
service_type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.ARAKOON)
cluster_name = Configuration.get(
'/ovs/framework/arakoon_clusters').get('voldrv')
if cluster_name is not None:
arakoon_service_name = ArakoonInstaller.get_service_name_for_cluster(
cluster_name=cluster_name)
for service in service_type.services:
if service.name == arakoon_service_name:
current_services.append(service)
if service.is_internal is True:
current_ips.append(service.storagerouter.ip)
all_sr_ips = [
storagerouter.ip
for storagerouter in StorageRouterList.get_slaves()
]
available_storagerouters = {}
for storagerouter in StorageRouterList.get_masters():
storagerouter.invalidate_dynamics(['partition_config'])
if len(storagerouter.partition_config[DiskPartition.ROLES.DB]) > 0:
available_storagerouters[storagerouter] = DiskPartition(
storagerouter.partition_config[DiskPartition.ROLES.DB][0])
all_sr_ips.append(storagerouter.ip)
if create_cluster is True and len(
current_services) == 0: # Create new cluster
metadata = ArakoonInstaller.get_unused_arakoon_metadata_and_claim(
cluster_type=ServiceType.ARAKOON_CLUSTER_TYPES.SD)
if metadata is None: # No externally managed cluster found, we create 1 ourselves
if not available_storagerouters:
raise RuntimeError(
'Could not find any Storage Router with a DB role')
storagerouter, partition = available_storagerouters.items()[0]
arakoon_voldrv_cluster = 'voldrv'
arakoon_installer = ArakoonInstaller(
cluster_name=arakoon_voldrv_cluster)
arakoon_installer.create_cluster(
cluster_type=ServiceType.ARAKOON_CLUSTER_TYPES.SD,
ip=storagerouter.ip,
base_dir=partition.folder,
log_sinks=LogHandler.get_sink_path(
'arakoon-server_{0}'.format(arakoon_voldrv_cluster)),
crash_log_sinks=LogHandler.get_sink_path(
'arakoon-server-crash_{0}'.format(
arakoon_voldrv_cluster)))
arakoon_installer.start_cluster()
ports = arakoon_installer.ports[storagerouter.ip]
metadata = arakoon_installer.metadata
current_ips.append(storagerouter.ip)
else:
ports = []
storagerouter = None
cluster_name = metadata['cluster_name']
Configuration.set('/ovs/framework/arakoon_clusters|voldrv',
cluster_name)
StorageDriverController._logger.info(
'Claiming {0} managed arakoon cluster: {1}'.format(
'externally' if storagerouter is None else 'internally',
cluster_name))
StorageDriverController._configure_arakoon_to_volumedriver(
cluster_name=cluster_name)
current_services.append(
_add_service(
service_storagerouter=storagerouter,
arakoon_ports=ports,
service_name=ArakoonInstaller.get_service_name_for_cluster(
cluster_name=cluster_name)))
cluster_name = Configuration.get(
'/ovs/framework/arakoon_clusters').get('voldrv')
if cluster_name is None:
return
metadata = ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=cluster_name)
if 0 < len(current_services) < len(
available_storagerouters) and metadata['internal'] is True:
for storagerouter, partition in available_storagerouters.iteritems(
):
if storagerouter.ip in current_ips:
continue
arakoon_installer = ArakoonInstaller(cluster_name=cluster_name)
arakoon_installer.load()
arakoon_installer.extend_cluster(
new_ip=storagerouter.ip,
base_dir=partition.folder,
log_sinks=LogHandler.get_sink_path(
'arakoon-server_{0}'.format(cluster_name)),
crash_log_sinks=LogHandler.get_sink_path(
'arakoon-server-crash_{0}'.format(cluster_name)))
_add_service(
service_storagerouter=storagerouter,
arakoon_ports=arakoon_installer.ports[storagerouter.ip],
service_name=ArakoonInstaller.get_service_name_for_cluster(
cluster_name=cluster_name))
current_ips.append(storagerouter.ip)
arakoon_installer.restart_cluster_after_extending(
new_ip=storagerouter.ip)
StorageDriverController._configure_arakoon_to_volumedriver(
cluster_name=cluster_name)
def services_running(self):
# type: () -> bool
"""
Check if all services are running
:return: Boolean
"""
try:
key = 'ovs-watcher-{0}'.format(str(uuid.uuid4()))
value = str(time.time())
if self.target in [WatcherTypes.CONFIG, WatcherTypes.FWK]:
self.log_message('Testing configuration store...')
try:
Configuration.list('/')
except Exception as ex:
self.log_message(' Error during configuration store test: {0}'.format(ex), 2)
return False
with open(CACC_LOCATION) as config_file:
contents = config_file.read()
config = ArakoonClusterConfig(cluster_id=ARAKOON_NAME, load_config=False)
config.read_config(contents=contents)
client = ArakoonInstaller.build_client(config)
contents = client.get(ArakoonInstaller.INTERNAL_CONFIG_KEY, consistency=NoGuarantee())
if Watcher.LOG_CONTENTS != contents:
try:
config.read_config(contents=contents) # Validate whether the contents are not corrupt
except Exception as ex:
self.log_message(' Configuration stored in configuration store seems to be corrupt: {0}'.format(ex), 2)
return False
temp_filename = '{0}~'.format(CACC_LOCATION)
with open(temp_filename, 'w') as config_file:
config_file.write(contents)
config_file.flush()
os.fsync(config_file)
os.rename(temp_filename, CACC_LOCATION)
Watcher.LOG_CONTENTS = contents
self.log_message(' Configuration store OK', 0)
if self.target == WatcherTypes.FWK:
self._test_store('volatile', key, value)
self._test_store('persistent')
if self.target == WatcherTypes.VOLDRV:
# Arakoon, voldrv cluster
self._test_store('arakoon_voldrv')
if self.target in [WatcherTypes.FWK, WatcherTypes.VOLDRV]:
# RabbitMQ
self.log_message('Test rabbitMQ...', 0)
messagequeue = Configuration.get('/ovs/framework/messagequeue')
rmq_servers = messagequeue['endpoints']
good_node = False
for server in rmq_servers:
try:
connection_string = '{0}://{1}:{2}@{3}/%2F'.format(messagequeue['protocol'],
messagequeue['user'],
messagequeue['password'],
server)
connection = pika.BlockingConnection(pika.URLParameters(connection_string))
channel = connection.channel()
channel.basic_publish('', 'ovs-watcher', str(time.time()),
pika.BasicProperties(content_type='text/plain', delivery_mode=1))
connection.close()
good_node = True
except Exception as message:
self.log_message(' Error during rabbitMQ test on node {0}: {1}'.format(server, message), 2)
if good_node is False:
self.log_message(' No working rabbitMQ node could be found', 2)
return False
self.log_message(' RabbitMQ test OK')
self.log_message('All tests OK')
return True
except Exception as ex:
self.log_message('Unexpected exception: {0}'.format(ex), 2)
return False
def promote_node(cluster_ip, master_ip, ip_client_map, unique_id,
configure_memcached, configure_rabbitmq):
"""
Promotes a given node
"""
from ovs.dal.lists.storagerouterlist import StorageRouterList
from ovs.dal.lists.servicetypelist import ServiceTypeList
from ovs.dal.lists.servicelist import ServiceList
from ovs.dal.hybrids.service import Service
Toolbox.log(logger=NodeTypeController._logger,
messages='Promoting node',
title=True)
service_manager = ServiceFactory.get_manager()
if configure_memcached is True:
if NodeTypeController._validate_local_memcache_servers(
ip_client_map) is False:
raise RuntimeError(
'Not all memcache nodes can be reached which is required for promoting a node.'
)
target_client = ip_client_map[cluster_ip]
machine_id = System.get_my_machine_id(target_client)
node_name, _ = target_client.get_hostname()
master_client = ip_client_map[master_ip]
storagerouter = StorageRouterList.get_by_machine_id(unique_id)
storagerouter.node_type = 'MASTER'
storagerouter.save()
external_config = Configuration.get('/ovs/framework/external_config')
if external_config is None:
Toolbox.log(logger=NodeTypeController._logger,
messages='Joining Arakoon configuration cluster')
arakoon_installer = ArakoonInstaller(cluster_name='config')
arakoon_installer.load(ip=master_ip)
arakoon_installer.extend_cluster(
new_ip=cluster_ip,
base_dir=Configuration.get('/ovs/framework/paths|ovsdb'))
arakoon_installer.restart_cluster_after_extending(
new_ip=cluster_ip)
service_manager.register_service(
node_name=machine_id,
service_metadata=arakoon_installer.service_metadata[cluster_ip]
)
# Find other (arakoon) master nodes
arakoon_cluster_name = str(
Configuration.get('/ovs/framework/arakoon_clusters|ovsdb'))
arakoon_metadata = ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=arakoon_cluster_name)
config = ArakoonClusterConfig(cluster_id=arakoon_cluster_name)
master_node_ips = [node.ip for node in config.nodes]
if cluster_ip in master_node_ips:
master_node_ips.remove(cluster_ip)
if len(master_node_ips) == 0:
raise RuntimeError(
'There should be at least one other master node')
arakoon_ports = []
if arakoon_metadata['internal'] is True:
Toolbox.log(logger=NodeTypeController._logger,
messages='Joining Arakoon OVS DB cluster')
arakoon_installer = ArakoonInstaller(
cluster_name=arakoon_cluster_name)
arakoon_installer.load()
arakoon_installer.extend_cluster(
new_ip=cluster_ip,
base_dir=Configuration.get('/ovs/framework/paths|ovsdb'))
arakoon_installer.restart_cluster_after_extending(
new_ip=cluster_ip)
arakoon_ports = arakoon_installer.ports[cluster_ip]
if configure_memcached is True:
NodeTypeController.configure_memcached(
client=target_client, logger=NodeTypeController._logger)
NodeTypeController.add_services(client=target_client,
node_type='master',
logger=NodeTypeController._logger)
Toolbox.log(logger=NodeTypeController._logger,
messages='Update configurations')
if configure_memcached is True:
endpoints = Configuration.get('/ovs/framework/memcache|endpoints')
endpoint = '{0}:11211'.format(cluster_ip)
if endpoint not in endpoints:
endpoints.append(endpoint)
Configuration.set('/ovs/framework/memcache|endpoints',
endpoints)
if configure_rabbitmq is True:
endpoints = Configuration.get(
'/ovs/framework/messagequeue|endpoints')
endpoint = '{0}:5672'.format(cluster_ip)
if endpoint not in endpoints:
endpoints.append(endpoint)
Configuration.set('/ovs/framework/messagequeue|endpoints',
endpoints)
if arakoon_metadata['internal'] is True:
Toolbox.log(logger=NodeTypeController._logger,
messages='Restarting master node services')
PersistentFactory.store = None
VolatileFactory.store = None
if 'arakoon-ovsdb' not in [
s.name for s in ServiceList.get_services() if
s.is_internal is False or s.storagerouter.ip == cluster_ip
]:
service = Service()
service.name = 'arakoon-ovsdb'
service.type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.ARAKOON)
service.ports = arakoon_ports
service.storagerouter = storagerouter
service.save()
if configure_rabbitmq is True:
NodeTypeController.configure_rabbitmq(
client=target_client, logger=NodeTypeController._logger)
# Copy rabbitmq cookie
rabbitmq_cookie_file = '/var/lib/rabbitmq/.erlang.cookie'
Toolbox.log(logger=NodeTypeController._logger,
messages='Copying RabbitMQ cookie')
contents = master_client.file_read(rabbitmq_cookie_file)
master_hostname, _ = master_client.get_hostname()
target_client.dir_create(os.path.dirname(rabbitmq_cookie_file))
target_client.file_write(rabbitmq_cookie_file, contents)
target_client.file_chmod(rabbitmq_cookie_file, mode=0400)
target_client.run(['rabbitmq-server', '-detached'])
time.sleep(5)
target_client.run(['rabbitmqctl', 'stop_app'])
time.sleep(5)
target_client.run([
'rabbitmqctl', 'join_cluster',
'rabbit@{0}'.format(master_hostname)
])
time.sleep(5)
target_client.run(['rabbitmqctl', 'stop'])
time.sleep(5)
# Enable HA for the rabbitMQ queues
ServiceFactory.change_service_state(target_client,
'rabbitmq-server', 'start',
NodeTypeController._logger)
NodeTypeController.check_rabbitmq_and_enable_ha_mode(
client=target_client, logger=NodeTypeController._logger)
NodeTypeController._configure_amqp_to_volumedriver()
Toolbox.log(logger=NodeTypeController._logger,
messages='Starting services')
services = ['memcached', 'arakoon-ovsdb', 'rabbitmq-server']
if arakoon_metadata['internal'] is True:
services.remove('arakoon-ovsdb')
for service in services:
if service_manager.has_service(service, client=target_client):
ServiceFactory.change_service_state(target_client, service,
'start',
NodeTypeController._logger)
Toolbox.log(logger=NodeTypeController._logger,
messages='Restarting services')
NodeTypeController.restart_framework_and_memcache_services(
clients=ip_client_map, logger=NodeTypeController._logger)
if Toolbox.run_hooks(component='nodetype',
sub_component='promote',
logger=NodeTypeController._logger,
cluster_ip=cluster_ip,
master_ip=master_ip):
Toolbox.log(logger=NodeTypeController._logger,
messages='Restarting services')
NodeTypeController.restart_framework_and_memcache_services(
clients=ip_client_map, logger=NodeTypeController._logger)
if NodeTypeController.avahi_installed(
client=target_client,
logger=NodeTypeController._logger) is True:
NodeTypeController.configure_avahi(
client=target_client,
node_name=node_name,
node_type='master',
logger=NodeTypeController._logger)
Configuration.set('/ovs/framework/hosts/{0}/type'.format(machine_id),
'MASTER')
target_client.run(
['chown', '-R', 'ovs:ovs', '/opt/OpenvStorage/config'])
Configuration.set(
'/ovs/framework/hosts/{0}/promotecompleted'.format(machine_id),
True)
if target_client.file_exists('/tmp/ovs_rollback'):
target_client.file_delete('/tmp/ovs_rollback')
Toolbox.log(logger=NodeTypeController._logger,
messages='Promote complete')
def extend_arakoon(cluster_name,
master_storagerouter_ip,
storagerouter_ip,
cluster_basedir,
service_type=ServiceType.ARAKOON_CLUSTER_TYPES.FWK,
clustered_nodes=None):
"""
Adds a external arakoon to a storagerouter
:param cluster_name: name of the already existing arakoon cluster
:type cluster_name: str
:param master_storagerouter_ip: master ip address of the existing arakoon cluster
e.g. 10.100.199.11
:type master_storagerouter_ip: str
:param storagerouter_ip: ip of a new storagerouter to extend to
e.g. 10.100.199.12
:type storagerouter_ip: str
:param cluster_basedir: absolute path for the new arakoon cluster
:type cluster_basedir: str
:param service_type: type of plugin for arakoon (DEFAULT=ServiceType.ARAKOON_CLUSTER_TYPES.FWK)
* FWK
* ABM
* NSM
:type service_type: ovs.dal.hybrids.ServiceType.ARAKOON_CLUSTER_TYPES
:param clustered_nodes: nodes who are available for the arakoon (including the to be extended_arakoon)
e.g. ['10.100.199.11', '10.100.199.12'] (DEFAULT=[])
:type clustered_nodes: list
:return: is created or not
:rtype: bool
"""
if clustered_nodes is None:
clustered_nodes = []
client = SSHClient(storagerouter_ip, username='******')
# create required directories
if not client.dir_exists(cluster_basedir):
client.dir_create(cluster_basedir)
ArakoonSetup.LOGGER.info(
"Starting extending arakoon cluster with name `{0}`, master_ip `{1}`, slave_ip `{2}`, base_dir `{3}`"
.format(cluster_name, master_storagerouter_ip, storagerouter_ip,
cluster_basedir))
arakoon_installer = ArakoonInstaller(cluster_name)
arakoon_installer.load()
arakoon_installer.extend_cluster(
new_ip=storagerouter_ip,
base_dir=cluster_basedir,
locked=False,
log_sinks=Logger.get_sink_path('automation_lib_arakoon_server'),
crash_log_sinks=Logger.get_sink_path(
'automation_lib_arakoon_server_crash'))
if service_type == ServiceType.ARAKOON_CLUSTER_TYPES.ABM:
client.run([
'ln', '-s', '/usr/lib/alba/albamgr_plugin.cmxs',
'{0}/arakoon/{1}/db'.format(cluster_basedir, cluster_name)
])
elif service_type == ServiceType.ARAKOON_CLUSTER_TYPES.NSM:
client.run([
'ln', '-s', '/usr/lib/alba/nsm_host_plugin.cmxs',
'{0}/arakoon/{1}/db'.format(cluster_basedir, cluster_name)
])
# checking if we need to restart the given nodes
if len(clustered_nodes) != 0:
ArakoonSetup.LOGGER.info(
"Trying to restart all given nodes of arakoon: {0}".format(
clustered_nodes, cluster_name))
arakoon_installer.restart_cluster_after_extending(
new_ip=storagerouter_ip)
ArakoonSetup.LOGGER.info(
"Finished restarting all given nodes of arakoon: {0}".format(
clustered_nodes, cluster_name))
ArakoonSetup.LOGGER.info(
"Finished extending arakoon cluster with name `{0}`, master_ip `{1}`, slave_ip `{2}`, base_dir `{3}`"
.format(cluster_name, master_storagerouter_ip, storagerouter_ip,
cluster_basedir))
def demote_node(cluster_ip,
master_ip,
ip_client_map,
unique_id,
unconfigure_memcached,
unconfigure_rabbitmq,
offline_nodes=None):
"""
Demotes a given node
"""
from ovs.dal.lists.storagerouterlist import StorageRouterList
Toolbox.log(logger=NodeTypeController._logger,
messages='Demoting node',
title=True)
service_manager = ServiceFactory.get_manager()
if offline_nodes is None:
offline_nodes = []
if unconfigure_memcached is True and len(offline_nodes) == 0:
if NodeTypeController._validate_local_memcache_servers(
ip_client_map) is False:
raise RuntimeError(
'Not all memcache nodes can be reached which is required for demoting a node.'
)
# Find other (arakoon) master nodes
arakoon_cluster_name = str(
Configuration.get('/ovs/framework/arakoon_clusters|ovsdb'))
arakoon_metadata = ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=arakoon_cluster_name)
config = ArakoonClusterConfig(cluster_id=arakoon_cluster_name)
master_node_ips = [node.ip for node in config.nodes]
shrink = False
if cluster_ip in master_node_ips:
shrink = True
master_node_ips.remove(cluster_ip)
if len(master_node_ips) == 0:
raise RuntimeError(
'There should be at least one other master node')
storagerouter = StorageRouterList.get_by_machine_id(unique_id)
storagerouter.node_type = 'EXTRA'
storagerouter.save()
offline_node_ips = [node.ip for node in offline_nodes]
if arakoon_metadata['internal'] is True and shrink is True:
Toolbox.log(logger=NodeTypeController._logger,
messages='Leaving Arakoon {0} cluster'.format(
arakoon_cluster_name))
arakoon_installer = ArakoonInstaller(
cluster_name=arakoon_cluster_name)
arakoon_installer.load()
arakoon_installer.shrink_cluster(removal_ip=cluster_ip,
offline_nodes=offline_node_ips)
arakoon_installer.restart_cluster_after_shrinking()
try:
external_config = Configuration.get(
'/ovs/framework/external_config')
if external_config is None and shrink is True:
Toolbox.log(logger=NodeTypeController._logger,
messages='Leaving Arakoon config cluster')
arakoon_installer = ArakoonInstaller(cluster_name='config')
arakoon_installer.load(ip=master_node_ips[0])
arakoon_installer.shrink_cluster(
removal_ip=cluster_ip, offline_nodes=offline_node_ips)
arakoon_installer.restart_cluster_after_shrinking()
except Exception as ex:
Toolbox.log(
logger=NodeTypeController._logger,
messages=['\nFailed to leave configuration cluster', ex],
loglevel='exception')
Toolbox.log(logger=NodeTypeController._logger,
messages='Update configurations')
try:
if unconfigure_memcached is True:
endpoints = Configuration.get(
'/ovs/framework/memcache|endpoints')
endpoint = '{0}:{1}'.format(cluster_ip, 11211)
if endpoint in endpoints:
endpoints.remove(endpoint)
Configuration.set('/ovs/framework/memcache|endpoints',
endpoints)
if unconfigure_rabbitmq is True:
endpoints = Configuration.get(
'/ovs/framework/messagequeue|endpoints')
endpoint = '{0}:{1}'.format(cluster_ip, 5672)
if endpoint in endpoints:
endpoints.remove(endpoint)
Configuration.set('/ovs/framework/messagequeue|endpoints',
endpoints)
except Exception as ex:
Toolbox.log(logger=NodeTypeController._logger,
messages=['\nFailed to update configurations', ex],
loglevel='exception')
if arakoon_metadata['internal'] is True:
Toolbox.log(logger=NodeTypeController._logger,
messages='Restarting master node services')
remaining_nodes = ip_client_map.keys()[:]
if cluster_ip in remaining_nodes:
remaining_nodes.remove(cluster_ip)
PersistentFactory.store = None
VolatileFactory.store = None
for service in storagerouter.services:
if service.name == 'arakoon-ovsdb':
service.delete()
target_client = None
if storagerouter in offline_nodes:
if unconfigure_rabbitmq is True:
Toolbox.log(
logger=NodeTypeController._logger,
messages='Removing/unconfiguring offline RabbitMQ node')
client = ip_client_map[master_ip]
try:
client.run([
'rabbitmqctl', 'forget_cluster_node',
'rabbit@{0}'.format(storagerouter.name)
])
except Exception as ex:
Toolbox.log(logger=NodeTypeController._logger,
messages=[
'\nFailed to forget RabbitMQ cluster node',
ex
],
loglevel='exception')
else:
target_client = ip_client_map[cluster_ip]
if unconfigure_rabbitmq is True:
Toolbox.log(logger=NodeTypeController._logger,
messages='Removing/unconfiguring RabbitMQ')
try:
if service_manager.has_service('rabbitmq-server',
client=target_client):
ServiceFactory.change_service_state(
target_client, 'rabbitmq-server', 'stop',
NodeTypeController._logger)
target_client.run(['rabbitmq-server', '-detached'])
time.sleep(5)
target_client.run(['rabbitmqctl', 'stop_app'])
time.sleep(5)
target_client.run(['rabbitmqctl', 'reset'])
time.sleep(5)
target_client.run(['rabbitmqctl', 'stop'])
time.sleep(5)
target_client.file_unlink(
"/var/lib/rabbitmq/.erlang.cookie")
ServiceFactory.change_service_state(
target_client, 'rabbitmq-server', 'stop',
NodeTypeController._logger) # To be sure
except Exception as ex:
Toolbox.log(logger=NodeTypeController._logger,
messages=[
'\nFailed to remove/unconfigure RabbitMQ',
ex
],
loglevel='exception')
Toolbox.log(logger=NodeTypeController._logger,
messages='Stopping services')
services = ['memcached', 'rabbitmq-server']
if unconfigure_rabbitmq is False:
services.remove('rabbitmq-server')
if unconfigure_memcached is False:
services.remove('memcached')
for service in services:
if service_manager.has_service(service, client=target_client):
Toolbox.log(
logger=NodeTypeController._logger,
messages='Stopping service {0}'.format(service))
try:
ServiceFactory.change_service_state(
target_client, service, 'stop',
NodeTypeController._logger)
except Exception as ex:
Toolbox.log(
logger=NodeTypeController._logger,
messages=[
'\nFailed to stop service'.format(service), ex
],
loglevel='exception')
Toolbox.log(logger=NodeTypeController._logger,
messages='Removing services')
services = [
'scheduled-tasks', 'webapp-api', 'volumerouter-consumer'
]
for service in services:
if service_manager.has_service(service, client=target_client):
Toolbox.log(
logger=NodeTypeController._logger,
messages='Removing service {0}'.format(service))
try:
ServiceFactory.change_service_state(
target_client, service, 'stop',
NodeTypeController._logger)
service_manager.remove_service(service,
client=target_client)
except Exception as ex:
Toolbox.log(
logger=NodeTypeController._logger,
messages=[
'\nFailed to remove service'.format(service),
ex
],
loglevel='exception')
if service_manager.has_service('workers', client=target_client):
service_manager.add_service(
name='workers',
client=target_client,
params={'WORKER_QUEUE': '{0}'.format(unique_id)})
try:
NodeTypeController._configure_amqp_to_volumedriver()
except Exception as ex:
Toolbox.log(
logger=NodeTypeController._logger,
messages=['\nFailed to configure AMQP to Storage Driver', ex],
loglevel='exception')
Toolbox.log(logger=NodeTypeController._logger,
messages='Restarting services')
NodeTypeController.restart_framework_and_memcache_services(
clients=ip_client_map,
logger=NodeTypeController._logger,
offline_node_ips=offline_node_ips)
if Toolbox.run_hooks(component='nodetype',
sub_component='demote',
logger=NodeTypeController._logger,
cluster_ip=cluster_ip,
master_ip=master_ip,
offline_node_ips=offline_node_ips):
Toolbox.log(logger=NodeTypeController._logger,
messages='Restarting services')
NodeTypeController.restart_framework_and_memcache_services(
clients=ip_client_map,
logger=NodeTypeController._logger,
offline_node_ips=offline_node_ips)
if storagerouter not in offline_nodes:
target_client = ip_client_map[cluster_ip]
node_name, _ = target_client.get_hostname()
if NodeTypeController.avahi_installed(
client=target_client,
logger=NodeTypeController._logger) is True:
NodeTypeController.configure_avahi(
client=target_client,
node_name=node_name,
node_type='extra',
logger=NodeTypeController._logger)
Configuration.set(
'/ovs/framework/hosts/{0}/type'.format(storagerouter.machine_id),
'EXTRA')
if target_client is not None and target_client.file_exists(
'/tmp/ovs_rollback'):
target_client.file_write('/tmp/ovs_rollback', 'rollback')
Toolbox.log(logger=NodeTypeController._logger,
messages='Demote complete',
title=True)
def test_arakoon_collapse(self):
"""
Test the Arakoon collapse functionality
"""
# Set up the test
structure = DalHelper.build_dal_structure(
structure={'storagerouters': [1, 2]})
storagerouter_1 = structure['storagerouters'][1]
storagerouter_2 = structure['storagerouters'][2]
MockedSSHClient._run_returns[storagerouter_1.ip] = {}
MockedSSHClient._run_returns[storagerouter_2.ip] = {}
# Make sure we cover all Arakoon cluster types
clusters_to_create = {
ServiceType.ARAKOON_CLUSTER_TYPES.SD: [{
'name': 'unittest-voldrv',
'internal': True,
'success': True
}],
ServiceType.ARAKOON_CLUSTER_TYPES.CFG: [{
'name': 'unittest-cacc',
'internal': True,
'success': True
}],
ServiceType.ARAKOON_CLUSTER_TYPES.FWK: [{
'name': 'unittest-ovsdb',
'internal': True,
'success': False
}],
ServiceType.ARAKOON_CLUSTER_TYPES.ABM: [{
'name': 'unittest-cluster-1-abm',
'internal': True,
'success': False
}, {
'name': 'unittest-random-abm-name',
'internal': False,
'success': True
}],
ServiceType.ARAKOON_CLUSTER_TYPES.NSM: [{
'name': 'unittest-cluster-1-nsm_0',
'internal': True,
'success': True
}]
}
self.assertEqual(
first=sorted(clusters_to_create.keys()),
second=sorted(ServiceType.ARAKOON_CLUSTER_TYPES.keys()),
msg=
'An Arakoon cluster type has been removed or added, please update this test accordingly'
)
# Create all Arakoon clusters and related services
failed_clusters = []
external_clusters = []
successful_clusters = []
for cluster_type, cluster_infos in clusters_to_create.iteritems():
filesystem = cluster_type == ServiceType.ARAKOON_CLUSTER_TYPES.CFG
for cluster_info in cluster_infos:
internal = cluster_info['internal']
cluster_name = cluster_info['name']
base_dir = DalHelper.CLUSTER_DIR.format(cluster_name)
arakoon_installer = ArakoonInstaller(cluster_name=cluster_name)
arakoon_installer.create_cluster(cluster_type=cluster_type,
ip=storagerouter_1.ip,
base_dir=base_dir,
internal=internal)
arakoon_installer.start_cluster()
arakoon_installer.extend_cluster(new_ip=storagerouter_2.ip,
base_dir=base_dir)
service_name = ArakoonInstaller.get_service_name_for_cluster(
cluster_name=cluster_name)
if cluster_type == ServiceType.ARAKOON_CLUSTER_TYPES.ABM:
service_type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.ALBA_MGR)
elif cluster_type == ServiceType.ARAKOON_CLUSTER_TYPES.NSM:
service_type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.NS_MGR)
else:
service_type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.ARAKOON)
if internal is True:
DalHelper.create_service(
service_name=service_name,
service_type=service_type,
storagerouter=storagerouter_1,
ports=arakoon_installer.ports[storagerouter_1.ip])
DalHelper.create_service(
service_name=service_name,
service_type=service_type,
storagerouter=storagerouter_2,
ports=arakoon_installer.ports[storagerouter_2.ip])
else:
DalHelper.create_service(service_name=service_name,
service_type=service_type)
external_clusters.append(cluster_name)
continue
if cluster_info['success'] is True:
if filesystem is True:
config_path = ArakoonClusterConfig.CONFIG_FILE.format(
cluster_name)
else:
config_path = Configuration.get_configuration_path(
ArakoonClusterConfig.CONFIG_KEY.format(
cluster_name))
MockedSSHClient._run_returns[storagerouter_1.ip][
'arakoon --collapse-local 1 2 -config {0}'.format(
config_path)] = None
MockedSSHClient._run_returns[storagerouter_2.ip][
'arakoon --collapse-local 2 2 -config {0}'.format(
config_path)] = None
successful_clusters.append(cluster_name)
else: # For successful False clusters we don't emulate the collapse, thus making it fail
failed_clusters.append(cluster_name)
# Start collapse and make it fail for all clusters on StorageRouter 2
SSHClient._raise_exceptions[storagerouter_2.ip] = {
'users': ['ovs'],
'exception': UnableToConnectException('No route to host')
}
GenericController.collapse_arakoon()
# Verify all log messages for each type of cluster
generic_logs = Logger._logs.get('lib', {})
for cluster_name in successful_clusters + failed_clusters + external_clusters:
collect_msg = (
'DEBUG',
'Collecting info for cluster {0}'.format(cluster_name))
unreachable_msg = (
'ERROR',
'Could not collapse any cluster on {0} (not reachable)'.format(
storagerouter_2.name))
end_collapse_msg = (
'DEBUG', 'Collapsing cluster {0} on {1} completed'.format(
cluster_name, storagerouter_1.ip))
start_collapse_msg = ('DEBUG',
'Collapsing cluster {0} on {1}'.format(
cluster_name, storagerouter_1.ip))
failed_collapse_msg = (
'ERROR', 'Collapsing cluster {0} on {1} failed'.format(
cluster_name, storagerouter_1.ip))
messages_to_validate = []
if cluster_name in successful_clusters:
assert_function = self.assertIn
messages_to_validate.append(collect_msg)
messages_to_validate.append(unreachable_msg)
messages_to_validate.append(start_collapse_msg)
messages_to_validate.append(end_collapse_msg)
elif cluster_name in failed_clusters:
assert_function = self.assertIn
messages_to_validate.append(collect_msg)
messages_to_validate.append(unreachable_msg)
messages_to_validate.append(start_collapse_msg)
messages_to_validate.append(failed_collapse_msg)
else:
assert_function = self.assertNotIn
messages_to_validate.append(collect_msg)
messages_to_validate.append(start_collapse_msg)
messages_to_validate.append(end_collapse_msg)
for severity, message in messages_to_validate:
if assert_function == self.assertIn:
assert_message = 'Expected to find log message: {0}'.format(
message)
else:
assert_message = 'Did not expect to find log message: {0}'.format(
message)
assert_function(member=message,
container=generic_logs,
msg=assert_message)
if assert_function == self.assertIn:
self.assertEqual(
first=severity,
second=generic_logs[message],
msg='Log message {0} is of severity {1} expected {2}'.
format(message, generic_logs[message], severity))
# Collapse should always have a 'finished' message since each cluster should be attempted to be collapsed
for general_message in [
'Arakoon collapse started', 'Arakoon collapse finished'
]:
self.assertIn(member=general_message,
container=generic_logs,
msg='Expected to find log message: {0}'.format(
general_message))
def promote_or_demote_node(node_action,
cluster_ip=None,
execute_rollback=False):
"""
Promotes or demotes the local node
:param node_action: Demote or promote
:type node_action: str
:param cluster_ip: IP of node to promote or demote
:type cluster_ip: str
:param execute_rollback: In case of failure revert the changes made
:type execute_rollback: bool
:return: None
"""
if node_action not in ('promote', 'demote'):
raise ValueError('Nodes can only be promoted or demoted')
Toolbox.log(logger=NodeTypeController._logger,
messages='Open vStorage Setup - {0}'.format(
node_action.capitalize()),
boxed=True)
try:
Toolbox.log(logger=NodeTypeController._logger,
messages='Collecting information',
title=True)
machine_id = System.get_my_machine_id()
if Configuration.get('/ovs/framework/hosts/{0}/setupcompleted'.
format(machine_id)) is False:
raise RuntimeError('No local OVS setup found.')
if cluster_ip and not re.match(Toolbox.regex_ip, cluster_ip):
raise RuntimeError(
'Incorrect IP provided ({0})'.format(cluster_ip))
if cluster_ip:
client = SSHClient(endpoint=cluster_ip)
machine_id = System.get_my_machine_id(client)
node_type = Configuration.get(
'/ovs/framework/hosts/{0}/type'.format(machine_id))
if node_action == 'promote' and node_type == 'MASTER':
raise RuntimeError('This node is already master.')
elif node_action == 'demote' and node_type == 'EXTRA':
raise RuntimeError('This node should be a master.')
elif node_type not in ['MASTER', 'EXTRA']:
raise RuntimeError('This node is not correctly configured.')
master_ip = None
offline_nodes = []
online = True
target_client = None
if node_action == 'demote' and cluster_ip: # Demote an offline node
from ovs.dal.lists.storagerouterlist import StorageRouterList
from ovs.lib.storagedriver import StorageDriverController
ip = cluster_ip
unique_id = None
ip_client_map = {}
for storage_router in StorageRouterList.get_storagerouters():
try:
client = SSHClient(storage_router.ip, username='******')
if storage_router.node_type == 'MASTER':
master_ip = storage_router.ip
ip_client_map[storage_router.ip] = client
except UnableToConnectException:
if storage_router.ip == cluster_ip:
online = False
unique_id = storage_router.machine_id
StorageDriverController.mark_offline(
storagerouter_guid=storage_router.guid)
offline_nodes.append(storage_router)
if online is True:
raise RuntimeError(
"If the node is online, please use 'ovs setup demote' executed on the node you wish to demote"
)
if master_ip is None:
raise RuntimeError(
'Failed to retrieve another responsive MASTER node')
else:
target_password = Toolbox.ask_validate_password(
ip='127.0.0.1', logger=NodeTypeController._logger)
target_client = SSHClient('127.0.0.1',
username='******',
password=target_password)
unique_id = System.get_my_machine_id(target_client)
ip = Configuration.get(
'/ovs/framework/hosts/{0}/ip'.format(unique_id))
storagerouter_info = NodeTypeController.retrieve_storagerouter_info_via_host(
ip=target_client.ip, password=target_password)
node_ips = [
sr_info['ip']
for sr_info in storagerouter_info.itervalues()
]
master_node_ips = [
sr_info['ip']
for sr_info in storagerouter_info.itervalues()
if sr_info['type'] == 'master' and sr_info['ip'] != ip
]
if len(master_node_ips) == 0:
if node_action == 'promote':
raise RuntimeError('No master node could be found')
else:
raise RuntimeError(
'It is not possible to remove the only master')
master_ip = master_node_ips[0]
ip_client_map = dict(
(node_ip, SSHClient(node_ip, username='******'))
for node_ip in node_ips)
if node_action == 'demote':
for cluster_name in Configuration.list('/ovs/arakoon'):
config = ArakoonClusterConfig(cluster_id=cluster_name)
arakoon_client = ArakoonInstaller.build_client(config)
metadata = json.loads(
arakoon_client.get(ArakoonInstaller.METADATA_KEY))
if len(config.nodes) == 1 and config.nodes[
0].ip == ip and metadata.get('internal') is True:
raise RuntimeError(
'Demote is not supported when single node Arakoon cluster(s) are present on the node to be demoted.'
)
configure_rabbitmq = Toolbox.is_service_internally_managed(
service='rabbitmq')
configure_memcached = Toolbox.is_service_internally_managed(
service='memcached')
if node_action == 'promote':
try:
NodeTypeController.promote_node(
cluster_ip=ip,
master_ip=master_ip,
ip_client_map=ip_client_map,
unique_id=unique_id,
configure_memcached=configure_memcached,
configure_rabbitmq=configure_rabbitmq)
except Exception:
if execute_rollback is True:
NodeTypeController.demote_node(
cluster_ip=ip,
master_ip=master_ip,
ip_client_map=ip_client_map,
unique_id=unique_id,
unconfigure_memcached=configure_memcached,
unconfigure_rabbitmq=configure_rabbitmq,
offline_nodes=offline_nodes)
elif target_client is not None:
target_client.file_write('/tmp/ovs_rollback', 'demote')
raise
else:
try:
NodeTypeController.demote_node(
cluster_ip=ip,
master_ip=master_ip,
ip_client_map=ip_client_map,
unique_id=unique_id,
unconfigure_memcached=configure_memcached,
unconfigure_rabbitmq=configure_rabbitmq,
offline_nodes=offline_nodes)
except Exception:
if execute_rollback is True:
NodeTypeController.promote_node(
cluster_ip=ip,
master_ip=master_ip,
ip_client_map=ip_client_map,
unique_id=unique_id,
configure_memcached=configure_memcached,
configure_rabbitmq=configure_rabbitmq)
elif target_client is not None:
target_client.file_write('/tmp/ovs_rollback',
'promote')
raise
Toolbox.log(logger=NodeTypeController._logger, messages='\n')
Toolbox.log(logger=NodeTypeController._logger,
messages='{0} complete.'.format(
node_action.capitalize()),
boxed=True)
except Exception as exception:
Toolbox.log(logger=NodeTypeController._logger, messages='\n')
Toolbox.log(
logger=NodeTypeController._logger,
messages=['An unexpected error occurred:',
str(exception)],
boxed=True,
loglevel='exception')
sys.exit(1)
except KeyboardInterrupt:
Toolbox.log(logger=NodeTypeController._logger, messages='\n')
Toolbox.log(
logger=NodeTypeController._logger,
messages=
'This setup was aborted. Open vStorage may be in an inconsistent state, make sure to validate the installation.',
boxed=True,
loglevel='error')
sys.exit(1)
def shrink_vpool(cls,
storagedriver_guid,
offline_storage_router_guids=list()):
"""
Removes a StorageDriver (if its the last StorageDriver for a vPool, the vPool is removed as well)
:param storagedriver_guid: Guid of the StorageDriver to remove
:type storagedriver_guid: str
:param offline_storage_router_guids: Guids of StorageRouters which are offline and will be removed from cluster.
WHETHER VPOOL WILL BE DELETED DEPENDS ON THIS
:type offline_storage_router_guids: list
:return: None
:rtype: NoneType
"""
# TODO: Add logging
# TODO: Unit test individual pieces of code
# Validations
storagedriver = StorageDriver(storagedriver_guid)
storagerouter = storagedriver.storagerouter
cls._logger.info(
'StorageDriver {0} - Deleting StorageDriver {1}'.format(
storagedriver.guid, storagedriver.name))
vp_installer = VPoolInstaller(name=storagedriver.vpool.name)
vp_installer.validate(storagedriver=storagedriver)
sd_installer = StorageDriverInstaller(vp_installer=vp_installer,
storagedriver=storagedriver)
cls._logger.info(
'StorageDriver {0} - Checking availability of related StorageRouters'
.format(storagedriver.guid, storagedriver.name))
sr_client_map = SSHClient.get_clients(endpoints=[
sd.storagerouter for sd in vp_installer.vpool.storagedrivers
],
user_names=['root'])
sr_installer = StorageRouterInstaller(root_client=sr_client_map.get(
storagerouter, {}).get('root'),
storagerouter=storagerouter,
vp_installer=vp_installer,
sd_installer=sd_installer)
offline_srs = sr_client_map.pop('offline')
if sorted([sr.guid for sr in offline_srs
]) != sorted(offline_storage_router_guids):
raise RuntimeError('Not all StorageRouters are reachable')
if storagerouter not in offline_srs:
mtpt_pids = sr_installer.root_client.run(
"lsof -t +D '/mnt/{0}' || true".format(
vp_installer.name.replace(r"'", r"'\''")),
allow_insecure=True).splitlines()
if len(mtpt_pids) > 0:
raise RuntimeError(
'vPool cannot be deleted. Following processes keep the vPool mount point occupied: {0}'
.format(', '.join(mtpt_pids)))
# Retrieve reachable StorageDrivers
reachable_storagedrivers = []
for sd in vp_installer.vpool.storagedrivers:
if sd.storagerouter not in sr_client_map:
# StorageRouter is offline
continue
sd_key = '/ovs/vpools/{0}/hosts/{1}/config'.format(
vp_installer.vpool.guid, sd.storagedriver_id)
if Configuration.exists(sd_key) is True:
path = Configuration.get_configuration_path(sd_key)
with remote(sd.storagerouter.ip,
[LocalStorageRouterClient]) as rem:
try:
lsrc = rem.LocalStorageRouterClient(path)
lsrc.server_revision(
) # 'Cheap' call to verify whether volumedriver is responsive
cls._logger.info(
'StorageDriver {0} - Responsive StorageDriver {1} on node with IP {2}'
.format(storagedriver.guid, sd.name,
sd.storagerouter.ip))
reachable_storagedrivers.append(sd)
except Exception as exception:
if not is_connection_failure(exception):
raise
if len(reachable_storagedrivers) == 0:
raise RuntimeError(
'Could not find any responsive node in the cluster')
# Start removal
if vp_installer.storagedriver_amount > 1:
vp_installer.update_status(status=VPool.STATUSES.SHRINKING)
else:
vp_installer.update_status(status=VPool.STATUSES.DELETING)
# Clean up stale vDisks
cls._logger.info('StorageDriver {0} - Removing stale vDisks'.format(
storagedriver.guid))
VDiskController.remove_stale_vdisks(vpool=vp_installer.vpool)
# Reconfigure the MDSes
cls._logger.info('StorageDriver {0} - Reconfiguring MDSes'.format(
storagedriver.guid))
for vdisk_guid in storagerouter.vdisks_guids:
try:
MDSServiceController.ensure_safety(
vdisk_guid=vdisk_guid,
excluded_storagerouter_guids=[storagerouter.guid] +
offline_storage_router_guids)
except Exception:
cls._logger.exception(
'StorageDriver {0} - vDisk {1} - Ensuring MDS safety failed'
.format(storagedriver.guid, vdisk_guid))
# Validate that all MDSes on current StorageRouter have been moved away
# Ensure safety does not always throw an error, that's why we perform this check here instead of in the Exception clause of above code
vdisks = []
for mds in vp_installer.mds_services:
for junction in mds.vdisks:
vdisk = junction.vdisk
if vdisk in vdisks:
continue
vdisks.append(vdisk)
cls._logger.critical(
'StorageDriver {0} - vDisk {1} {2} - MDS Services have not been migrated away'
.format(storagedriver.guid, vdisk.guid, vdisk.name))
if len(vdisks) > 0:
# Put back in RUNNING, so it can be used again. Errors keep on displaying in GUI now anyway
vp_installer.update_status(status=VPool.STATUSES.RUNNING)
raise RuntimeError(
'Not all MDS Services have been successfully migrated away')
# Start with actual removal
errors_found = False
if storagerouter not in offline_srs:
errors_found &= sd_installer.stop_services()
errors_found &= vp_installer.configure_cluster_registry(
exclude=[storagedriver], apply_on=reachable_storagedrivers)
errors_found &= vp_installer.update_node_distance_map()
errors_found &= vp_installer.remove_mds_services()
errors_found &= sd_installer.clean_config_management()
errors_found &= sd_installer.clean_model()
if storagerouter not in offline_srs:
errors_found &= sd_installer.clean_directories(
mountpoints=StorageRouterController.get_mountpoints(
client=sr_installer.root_client))
try:
DiskController.sync_with_reality(
storagerouter_guid=storagerouter.guid)
except Exception:
cls._logger.exception(
'StorageDriver {0} - Synchronizing disks with reality failed'
.format(storagedriver.guid))
errors_found = True
if vp_installer.storagedriver_amount > 1:
# Update the vPool metadata and run DTL checkup
vp_installer.vpool.metadata['caching_info'].pop(
sr_installer.storagerouter.guid, None)
vp_installer.vpool.save()
try:
VDiskController.dtl_checkup(vpool_guid=vp_installer.vpool.guid,
ensure_single_timeout=600)
except Exception:
cls._logger.exception(
'StorageDriver {0} - DTL checkup failed for vPool {1} with guid {2}'
.format(storagedriver.guid, vp_installer.name,
vp_installer.vpool.guid))
else:
cls._logger.info(
'StorageDriver {0} - Removing vPool from model'.format(
storagedriver.guid))
# Clean up model
try:
vp_installer.vpool.delete()
except Exception:
errors_found = True
cls._logger.exception(
'StorageDriver {0} - Cleaning up vPool from the model failed'
.format(storagedriver.guid))
Configuration.delete('/ovs/vpools/{0}'.format(
vp_installer.vpool.guid))
cls._logger.info('StorageDriver {0} - Running MDS checkup'.format(
storagedriver.guid))
try:
MDSServiceController.mds_checkup()
except Exception:
cls._logger.exception(
'StorageDriver {0} - MDS checkup failed'.format(
storagedriver.guid))
# Update vPool status
if errors_found is True:
if vp_installer.storagedriver_amount > 1:
vp_installer.update_status(status=VPool.STATUSES.FAILURE)
raise RuntimeError(
'1 or more errors occurred while trying to remove the StorageDriver. Please check the logs for more information'
)
if vp_installer.storagedriver_amount > 1:
vp_installer.update_status(status=VPool.STATUSES.RUNNING)
cls._logger.info(
'StorageDriver {0} - Deleted StorageDriver {1}'.format(
storagedriver.guid, storagedriver.name))
if len(VPoolList.get_vpools()) == 0:
cluster_name = ArakoonInstaller.get_cluster_name('voldrv')
if ArakoonInstaller.get_arakoon_metadata_by_cluster_name(
cluster_name=cluster_name)['internal'] is True:
cls._logger.debug(
'StorageDriver {0} - Removing Arakoon cluster {1}'.format(
storagedriver.guid, cluster_name))
try:
installer = ArakoonInstaller(cluster_name=cluster_name)
installer.load()
installer.delete_cluster()
except Exception:
cls._logger.exception(
'StorageDriver {0} - Delete voldrv Arakoon cluster failed'
.format(storagedriver.guid))
service_type = ServiceTypeList.get_by_name(
ServiceType.SERVICE_TYPES.ARAKOON)
service_name = ArakoonInstaller.get_service_name_for_cluster(
cluster_name=cluster_name)
for service in list(service_type.services):
if service.name == service_name:
service.delete()
# Remove watcher volumedriver service if last StorageDriver on current StorageRouter
if len(
storagerouter.storagedrivers
) == 0 and storagerouter not in offline_srs: # ensure client is initialized for StorageRouter
try:
if cls._service_manager.has_service(
ServiceFactory.SERVICE_WATCHER_VOLDRV,
client=sr_installer.root_client):
cls._service_manager.stop_service(
ServiceFactory.SERVICE_WATCHER_VOLDRV,
client=sr_installer.root_client)
cls._service_manager.remove_service(
ServiceFactory.SERVICE_WATCHER_VOLDRV,
client=sr_installer.root_client)
except Exception:
cls._logger.exception(
'StorageDriver {0} - {1} service deletion failed'.format(
storagedriver.guid,
ServiceFactory.SERVICE_WATCHER_VOLDRV))
def services_running(self, target):
"""
Check all services are running
:param target: Target to check
:return: Boolean
"""
try:
key = 'ovs-watcher-{0}'.format(str(uuid.uuid4()))
value = str(time.time())
if target in ['config', 'framework']:
self.log_message(target, 'Testing configuration store...', 0)
from ovs.extensions.generic.configuration import Configuration
try:
Configuration.list('/')
except Exception as ex:
self.log_message(
target,
' Error during configuration store test: {0}'.format(
ex), 2)
return False
from ovs.extensions.db.arakooninstaller import ArakoonInstaller, ArakoonClusterConfig
from ovs_extensions.db.arakoon.pyrakoon.pyrakoon.compat import NoGuarantee
from ovs.extensions.generic.configuration import Configuration
with open(Configuration.CACC_LOCATION) as config_file:
contents = config_file.read()
config = ArakoonClusterConfig(
cluster_id=Configuration.ARAKOON_NAME, load_config=False)
config.read_config(contents=contents)
client = ArakoonInstaller.build_client(config)
contents = client.get(ArakoonInstaller.INTERNAL_CONFIG_KEY,
consistency=NoGuarantee())
if Watcher.LOG_CONTENTS != contents:
try:
config.read_config(
contents=contents
) # Validate whether the contents are not corrupt
except Exception as ex:
self.log_message(
target,
' Configuration stored in configuration store seems to be corrupt: {0}'
.format(ex), 2)
return False
temp_filename = '{0}~'.format(Configuration.CACC_LOCATION)
with open(temp_filename, 'w') as config_file:
config_file.write(contents)
config_file.flush()
os.fsync(config_file)
os.rename(temp_filename, Configuration.CACC_LOCATION)
Watcher.LOG_CONTENTS = contents
self.log_message(target, ' Configuration store OK', 0)
if target == 'framework':
# Volatile
self.log_message(target, 'Testing volatile store...', 0)
max_tries = 5
tries = 0
while tries < max_tries:
try:
try:
logging.disable(logging.WARNING)
from ovs.extensions.storage.volatilefactory import VolatileFactory
VolatileFactory.store = None
volatile = VolatileFactory.get_client()
volatile.set(key, value)
if volatile.get(key) == value:
volatile.delete(key)
break
volatile.delete(key)
finally:
logging.disable(logging.NOTSET)
except Exception as message:
self.log_message(
target,
' Error during volatile store test: {0}'.format(
message), 2)
key = 'ovs-watcher-{0}'.format(str(
uuid.uuid4())) # Get another key
time.sleep(1)
tries += 1
if tries == max_tries:
self.log_message(target,
' Volatile store not working correctly',
2)
return False
self.log_message(
target,
' Volatile store OK after {0} tries'.format(tries), 0)
# Persistent
self.log_message(target, 'Testing persistent store...', 0)
max_tries = 5
tries = 0
while tries < max_tries:
try:
try:
logging.disable(logging.WARNING)
persistent = PersistentFactory.get_client()
persistent.nop()
break
finally:
logging.disable(logging.NOTSET)
except Exception as message:
self.log_message(
target,
' Error during persistent store test: {0}'.format(
message), 2)
time.sleep(1)
tries += 1
if tries == max_tries:
self.log_message(
target, ' Persistent store not working correctly', 2)
return False
self.log_message(
target,
' Persistent store OK after {0} tries'.format(tries), 0)
if target == 'volumedriver':
# Arakoon, voldrv cluster
self.log_message(target, 'Testing arakoon (voldrv)...', 0)
max_tries = 5
tries = 0
while tries < max_tries:
try:
from ovs.extensions.generic.configuration import Configuration
from ovs_extensions.storage.persistent.pyrakoonstore import PyrakoonStore
cluster_name = str(
Configuration.get(
'/ovs/framework/arakoon_clusters|voldrv'))
configuration = Configuration.get(
'/ovs/arakoon/{0}/config'.format(cluster_name),
raw=True)
client = PyrakoonStore(cluster=cluster_name,
configuration=configuration)
client.nop()
break
except Exception as message:
self.log_message(
target,
' Error during arakoon (voldrv) test: {0}'.format(
message), 2)
time.sleep(1)
tries += 1
if tries == max_tries:
self.log_message(
target, ' Arakoon (voldrv) not working correctly', 2)
return False
self.log_message(target, ' Arakoon (voldrv) OK', 0)
if target in ['framework', 'volumedriver']:
# RabbitMQ
self.log_message(target, 'Test rabbitMQ...', 0)
import pika
from ovs.extensions.generic.configuration import Configuration
messagequeue = Configuration.get('/ovs/framework/messagequeue')
rmq_servers = messagequeue['endpoints']
good_node = False
for server in rmq_servers:
try:
connection_string = '{0}://{1}:{2}@{3}/%2F'.format(
messagequeue['protocol'], messagequeue['user'],
messagequeue['password'], server)
connection = pika.BlockingConnection(
pika.URLParameters(connection_string))
channel = connection.channel()
channel.basic_publish(
'', 'ovs-watcher', str(time.time()),
pika.BasicProperties(content_type='text/plain',
delivery_mode=1))
connection.close()
good_node = True
except Exception as message:
self.log_message(
target,
' Error during rabbitMQ test on node {0}: {1}'.
format(server, message), 2)
if good_node is False:
self.log_message(
target, ' No working rabbitMQ node could be found', 2)
return False
self.log_message(target, ' RabbitMQ test OK', 0)
self.log_message(target, 'All tests OK', 0)
return True
except Exception as ex:
self.log_message(target, 'Unexpected exception: {0}'.format(ex), 2)
return False
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 test_nsm_checkup_external(self):
"""
Validates whether the NSM checkup works for externally managed Arakoon clusters
"""
Configuration.set('/ovs/framework/plugins/alba/config|nsm.safety', 1)
Configuration.set('/ovs/framework/plugins/alba/config|nsm.maxload', 10)
structure = DalHelper.build_dal_structure(structure={'storagerouters': [1, 2, 3]})
alba_structure = AlbaDalHelper.build_dal_structure(structure={'alba_backends': [[1, 'LOCAL']]})
alba_backend = alba_structure['alba_backends'][1]
storagerouter_1 = structure['storagerouters'][1]
storagerouter_2 = structure['storagerouters'][2]
# Validate some logic for externally managed arakoons during NSM checkup
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.nsm_checkup(external_nsm_cluster_names=['test']) # No ALBA Backend specified
self.assertEqual(first=str(raise_info.exception), second='Additional NSMs can only be configured for a specific ALBA Backend')
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.nsm_checkup(alba_backend_guid=alba_backend.guid, min_internal_nsms=2, external_nsm_cluster_names=['test'])
self.assertEqual(first=str(raise_info.exception), second="'min_internal_nsms' and 'external_nsm_cluster_names' are mutually exclusive")
with self.assertRaises(ValueError) as raise_info:
# noinspection PyTypeChecker
AlbaArakoonController.nsm_checkup(alba_backend_guid=alba_backend.guid, external_nsm_cluster_names={}) # NSM cluster names must be a list
self.assertEqual(first=str(raise_info.exception), second="'external_nsm_cluster_names' must be of type 'list'")
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.nsm_checkup(alba_backend_guid=alba_backend.guid, external_nsm_cluster_names=['non-existing-cluster']) # non-existing cluster names should raise
self.assertEqual(first=str(raise_info.exception), second="Arakoon cluster with name non-existing-cluster does not exist")
# Create an external ABM and NSM Arakoon cluster
external_abm_1 = 'backend_1-abm'
external_nsm_1 = 'backend_1-nsm_0'
external_nsm_2 = 'backend_1-nsm_1'
for cluster_name, cluster_type in {external_abm_1: 'ABM', external_nsm_1: 'NSM', external_nsm_2: 'NSM'}.iteritems():
arakoon_installer = ArakoonInstaller(cluster_name=cluster_name)
arakoon_installer.create_cluster(cluster_type=cluster_type, ip=storagerouter_1.ip, base_dir='/tmp', internal=False)
arakoon_installer.extend_cluster(new_ip=storagerouter_2.ip, base_dir='/tmp')
arakoon_installer.start_cluster()
arakoon_installer.unclaim_cluster()
self.assertDictEqual(d1={'cluster_name': cluster_name,
'cluster_type': cluster_type,
'internal': False,
'in_use': False},
d2=arakoon_installer.get_arakoon_metadata_by_cluster_name(cluster_name=cluster_name))
# Let the 'add_cluster` claim the externally managed clusters and model the services
Logger._logs = {}
AlbaController.add_cluster(alba_backend_guid=alba_backend.guid,
abm_cluster=external_abm_1,
nsm_clusters=[external_nsm_1]) # Only claim external_nsm_1
for cluster_name, cluster_type in {external_abm_1: 'ABM', external_nsm_1: 'NSM', external_nsm_2: 'NSM'}.iteritems():
arakoon_installer = ArakoonInstaller(cluster_name=cluster_name)
self.assertDictEqual(d1={'cluster_name': cluster_name,
'cluster_type': cluster_type,
'internal': False,
'in_use': False if cluster_name == external_nsm_2 else True},
d2=arakoon_installer.get_arakoon_metadata_by_cluster_name(cluster_name=cluster_name))
log_found = False
for log_record in Logger._logs.get('lib', []):
if 'NSM load OK' in log_record:
log_found = True
break
self.assertTrue(expr=log_found)
self.assertEqual(first=1, second=len(alba_backend.abm_cluster.abm_services))
self.assertEqual(first=1, second=len(alba_backend.nsm_clusters))
self.assertEqual(first=1, second=len(alba_backend.nsm_clusters[0].nsm_services))
self.assertIsNone(obj=alba_backend.abm_cluster.abm_services[0].service.storagerouter)
self.assertIsNone(obj=alba_backend.nsm_clusters[0].nsm_services[0].service.storagerouter)
self.assertListEqual(VirtualAlbaBackend.run_log['backend_1-abm'], [['update_abm_client_config'],
['add_nsm_host', 'backend_1-nsm_0'],
['update_maintenance_config','--eviction-type-random'],
['update_maintenance_config','enable-auto-cleanup-deleted-namespaces-days']])
# Add cluster already invokes a NSM checkup, so nothing should have changed
VirtualAlbaBackend.run_log['backend_1-abm'] = []
AlbaArakoonController.nsm_checkup()
self.assertListEqual(list1=[], list2=VirtualAlbaBackend.run_log['backend_1-abm'])
# Overload the only NSM and run NSM checkup. This should log a critical message, but change nothing
VirtualAlbaBackend.data['backend_1-abm']['nsms'][0]['namespaces_count'] = 25
Logger._logs = {}
AlbaArakoonController.nsm_checkup()
log_found = False
for log_record in Logger._logs.get('lib', []):
if 'All NSM clusters are overloaded' in log_record:
log_found = True
break
self.assertTrue(expr=log_found)
self.assertEqual(first=1, second=len(alba_backend.abm_cluster.abm_services))
self.assertEqual(first=1, second=len(alba_backend.nsm_clusters))
self.assertEqual(first=1, second=len(alba_backend.nsm_clusters[0].nsm_services))
self.assertIsNone(obj=alba_backend.abm_cluster.abm_services[0].service.storagerouter)
self.assertIsNone(obj=alba_backend.nsm_clusters[0].nsm_services[0].service.storagerouter)
self.assertListEqual(list1=[], list2=VirtualAlbaBackend.run_log['backend_1-abm'])
# Validate a maximum of 50 NSMs can be deployed
current_nsms = [nsm_cluster.number for nsm_cluster in alba_backend.nsm_clusters]
alba_structure = AlbaDalHelper.build_dal_structure(
structure={'alba_nsm_clusters': [(1, 50)]}, # (<abackend_id>, <amount_of_nsm_clusters>)
previous_structure=alba_structure
)
# Try to add 1 additional NSM
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.nsm_checkup(alba_backend_guid=alba_backend.guid, external_nsm_cluster_names=[external_nsm_2])
self.assertEqual(first=str(raise_info.exception), second='The maximum of 50 NSM Arakoon clusters will be exceeded. Amount of clusters that can be deployed for this ALBA Backend: 0')
# Remove the unused NSM clusters again
for nsm_cluster in alba_structure['alba_nsm_clusters'][1][len(current_nsms):]:
for nsm_service in nsm_cluster.nsm_services:
nsm_service.delete()
nsm_service.service.delete()
nsm_cluster.delete()
# Try to add a previously claimed NSM cluster
with self.assertRaises(ValueError) as raise_info:
AlbaArakoonController.nsm_checkup(alba_backend_guid=alba_backend.guid, external_nsm_cluster_names=[external_nsm_1]) # The provided cluster_name to claim has already been claimed
self.assertEqual(first=str(raise_info.exception), second='Some of the provided cluster_names have already been claimed before')
# Add a 2nd NSM cluster
AlbaArakoonController.nsm_checkup(alba_backend_guid=alba_backend.guid, external_nsm_cluster_names=[external_nsm_2])
self.assertEqual(first=1, second=len(alba_backend.abm_cluster.abm_services))
self.assertEqual(first=2, second=len(alba_backend.nsm_clusters))
self.assertEqual(first=1, second=len(alba_backend.nsm_clusters[0].nsm_services))
self.assertEqual(first=1, second=len(alba_backend.nsm_clusters[1].nsm_services))
self.assertIsNone(obj=alba_backend.abm_cluster.abm_services[0].service.storagerouter)
self.assertIsNone(obj=alba_backend.nsm_clusters[0].nsm_services[0].service.storagerouter)
self.assertIsNone(obj=alba_backend.nsm_clusters[1].nsm_services[0].service.storagerouter)
self.assertListEqual(list1=[['add_nsm_host', 'backend_1-nsm_1']], list2=VirtualAlbaBackend.run_log['backend_1-abm'])
for cluster_name, cluster_type in {external_abm_1: 'ABM', external_nsm_1: 'NSM', external_nsm_2: 'NSM'}.iteritems():
arakoon_installer = ArakoonInstaller(cluster_name=cluster_name)
self.assertDictEqual(d1={'cluster_name': cluster_name,
'cluster_type': cluster_type,
'internal': False,
'in_use': True},
d2=arakoon_installer.get_arakoon_metadata_by_cluster_name(cluster_name=cluster_name))