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 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 _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 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 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 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))