def enableComponents(self, session: Session):
"""
Raises:
ConfigurationError
"""
self._logger.info('Enabling \'installer\' component')
base_kit = KitApi().getKit(session, 'base')
enabledComponents = ['installer']
# get list of components from 'base' kit
components = [c for c in base_kit.getComponentList()
if c.getName() in enabledComponents]
installerNode = NodeApi().getInstallerNode(session)
for component in components:
SoftwareProfileApi().enableComponent(
session,
installerNode.getSoftwareProfile().getName(),
base_kit.getName(),
base_kit.getVersion(),
base_kit.getIteration(),
component.getName(), compVersion=component.getVersion(),
)
def postAddHost(self, session: Session, hardwareProfileName: str,
softwareProfileName: Optional[str],
addHostSession: str) -> None:
"""
Perform post add host operations
"""
self._logger.debug(
'postAddHost(): hardwareProfileName=[%s]'
' softwareProfileName=[%s] addHostSession=[%s]' % (
hardwareProfileName, softwareProfileName, addHostSession))
# this query is redundant; in the calling method, we already have
# a list of Node (db) objects
from tortuga.node.nodeApi import NodeApi
nodes = NodeApi().getNodesByAddHostSession(session, addHostSession)
mgr = KitActionsManager()
mgr.session = session
mgr.post_add_host(
hardwareProfileName,
softwareProfileName,
nodes
)
# Always go over the web service for this call.
SyncWsApi().scheduleClusterUpdate(updateReason='Node(s) added')
def test_getNodeList(self, get_software_profile_metadata_mock,
get_os_boot_host_manager_mock): \
# pylint: disable=unused-argument
"""
Get all nodes
"""
fake_metadata = {
'mike': {
'tag': 'value',
}
}
get_software_profile_metadata_mock.return_value = fake_metadata
from tortuga.node.nodeApi import NodeApi
with self.dbm.session() as session:
result = NodeApi().getNodeList(session)
assert isinstance(result, TortugaObjectList)
node = result[0]
metadata = node.getSoftwareProfile().getMetadata()
assert metadata == fake_metadata
get_software_profile_metadata_mock.assert_called()
def __getNodeApi(self):
"""Get and cache the Node API"""
if self.__nodeApi is None:
from tortuga.node.nodeApi import NodeApi
self.__nodeApi = NodeApi()
return self.__nodeApi
def test_getNodeById_nonexistent(self, get_software_profile_metadata_mock,
get_os_boot_host_manager_mock): \
# pylint: disable=unused-argument
from tortuga.node.nodeApi import NodeApi
with pytest.raises(NodeNotFound):
with self.dbm.session() as session:
NodeApi().getNodeById(session, 99999)
def test_getNodeById(self, get_software_profile_metadata_mock,
get_os_boot_host_manager_mock): \
# pylint: disable=unused-argument
from tortuga.node.nodeApi import NodeApi
node_id = 1
with self.dbm.session() as session:
node = NodeApi().getNodeById(session, node_id)
assert node.getId() == node_id
# this is a bit of a cheat since it compares host names only
# but that should be sufficient here...
assert node.getName().split('.', 1)[0] == \
socket.getfqdn().split('.', 1)[0]
get_software_profile_metadata_mock.assert_called_with(
session, node.getSoftwareProfile().getName())
def __init__(self, kit):
"""
Initialise parent class.
"""
super().__init__(kit)
self._provider = DhcpdDhcpProvider(self)
self._manager = self._get_os_dhcpd_manager('dhcpd')
self._config = ConfigManager()
self._installer_node = NodeApi().getInstallerNode()
def runCommand(self):
self.parseArgs(
_("""
Display list of nodes able to use the specified software profile,
ordered by cost.
"""))
softwareProfileName = self.getArgs().softwareProfile
nodeApi = NodeApi()
softwareUsesHardwareDbApi = SoftwareUsesHardwareDbApi()
hardwareProfileDbApi = HardwareProfileDbApi()
load_kits()
with DbManager().session() as session:
hwPList = hardwareProfileDbApi.getHardwareProfileList(session)
hardwareProfileIdList = softwareUsesHardwareDbApi.\
getAllowedHardwareProfilesBySoftwareProfileName(
session, softwareProfileName)
nodeList = nodeApi.getNodeList(session)
usableNodes = []
for node in nodeList:
if (node.getHardwareProfile().getId() in hardwareProfileIdList) \
and node.getIsIdle():
usableNodes.append(node)
costNameList = []
for node in usableNodes:
nodeHwP = node.getHardwareProfile().getId()
for hwP in hwPList:
if hwP.getId() == nodeHwP:
costNameList.append(
[int(hwP.getCost()),
node.getName()])
break
costNameList.sort()
for node in costNameList:
print('%s' % (node[1]))
def process_delete_host_request(session: Session,
transaction_id: str,
nodespec: str,
force: bool = False):
try:
req = NodeRequestsDbHandler().get_by_addHostSession(
session, transaction_id)
if req is None:
# Session was deleted prior to being processed. Nothing to do...
logger.warning(
'Delete host request [%s] not found; nothing to do...',
transaction_id)
return
#
# Save this data so that we have it for firing the event below
#
evt_req_id = req.id
evt_req_request = {'name': nodespec}
ahm.update_session(transaction_id, running=True)
logger.debug('process_delete_host_request(): transaction_id=[{0}],'
' nodespec=[{1}]'.format(transaction_id, nodespec))
try:
NodeApi().deleteNode(session, nodespec, force=force)
ahm.delete_session(transaction_id)
session.delete(req)
DeleteNodeRequestComplete.fire(request_id=evt_req_id,
request=evt_req_request)
except (OperationFailed, NodeNotFound):
ahm.delete_session(transaction_id)
session.delete(req)
raise
except TortugaException as exc:
logger.exception('Exception while deleting nodes')
req.message = str(exc)
req.state = 'error'
req.last_update = datetime.datetime.utcnow()
finally:
ahm.update_session(transaction_id, running=False)
finally:
session.commit()
def test_invalid_transfer(self, node_state_change_object,
get_kit_installer_function, MockClass1,
get_os_object_factory_function, dbm):
"""
Attempt to transfer node to current software profile
"""
with dbm.session() as session:
with pytest.raises(NodeTransferNotValid):
NodeApi().transferNodes(session,
'compute',
nodespec='compute-02.private')
def configure(self):
#
# Write config file
#
fp = open(CONFIG_FILE, 'w')
try:
print("# File generated by genconfig", file=fp)
installer = self.kit_installer.config_manager.getInstaller()
all_node_list = NodeApi().getNodeList()
node_list = [
node for node in all_node_list if node.getName() != installer
and node.getState() != 'Deleted' and not node.getIsIdle()
]
for node in node_list:
print('{}'.format(node), file=fp)
finally:
fp.close()
#
# Write /etc/netgroup
#
fp = open('/etc/netgroup', 'w')
dbm = DbManager()
session = dbm.openSession()
try:
software_profiles = \
SoftwareProfilesDbHandler().getSoftwareProfileList(session)
for software_profile in software_profiles:
if not software_profile.nodes:
continue
software_profile_node_list = [
node.name for node in software_profile.nodes
if node.state != 'Deleted'
]
if not software_profile_node_list:
continue
fp.write('{} {}\n\n'.format(
software_profile.name, ' '.join([
'({},,)'.format(node)
for node in software_profile_node_list
])))
finally:
fp.close()
dbm.closeSession()
fp.close()
def test_invalid_software_profile_transferNode(
self, get_software_profile_metadata_mock,
get_os_boot_host_manager_mock): \
# pylint: disable=unused-argument
from tortuga.node.nodeApi import NodeApi
with pytest.raises(SoftwareProfileNotFound):
with self.dbm.session() as session:
NodeApi().transferNodes(session,
'Compute2',
nodespec='compute-01.private')
def enableComponents(self):
"""
Raises:
ConfigurationError
"""
self._logger.info('Enabling \'installer\' component')
k = self.__getBaseKit()
enabledComponents = ['installer']
components = [
c for c in k.getComponentList() if c.getName() in enabledComponents
]
installerNode = NodeApi().getInstallerNode()
for component in components:
SoftwareProfileApi().enableComponent(
installerNode.getSoftwareProfile().getName(), k.getName(),
k.getVersion(), k.getIteration(), component.getName(),
component.getVersion())
def _get_installer_ip(self, network_id):
"""
Return IP address of provisioning interface on installer
:raises NicNotFound:
"""
installer_node = NodeApi().getInstallerNode(self.session)
prov_nics = self._get_provisioning_nics(installer_node)
for prov_nic in prov_nics:
if prov_nic.getNetwork().getId() == network_id:
return ipaddress.IPv4Address(prov_nic.getIp())
raise NicNotFound(
'Network has no corresponding provisioning NIC on installer')
def _findUnmanagedNics(self, session: Session):
# Get list of all NICs on the installer
systemNics = set(self._getSystemNics())
# Get provisioning NICs on installer
nics = NodeApi().getInstallerNode(session).getNics()
# Filter out the NIC names
primaryInstallerNics = set(
[nic.getNetworkDevice().getName() for nic in nics])
# Determine difference between all system NICs and managed NICs
unmanagedNics = systemNics.difference(primaryInstallerNics)
print('The following NICs are not currently managed: %s' %
(' '.join(unmanagedNics)))
def _local_delete(self, resource_adapter_name: str,
resource_adapter_profile: str, cloudserver_id: str):
from tortuga.db.models.instanceMapping import InstanceMapping
from tortuga.db.models.instanceMetadata import InstanceMetadata
from tortuga.db.resourceAdapterConfigDbHandler import ResourceAdapterConfigDbHandler
from tortuga.node.nodeApi import NodeApi
node_api = NodeApi()
#
# Assume the node instance ID name is the last item in the
# delimited cloud server ID
#
instance = cloudserver_id.split(":")[-1]
#
# Lookup the resource adapter configuration profile...
#
rac_api = ResourceAdapterConfigDbHandler()
rac = rac_api.get(self._sess, resource_adapter_name,
resource_adapter_profile)
#
# Check the instance mapping to see if there is a matching
# instance id...
#
im_list = self._sess.query(InstanceMapping).filter(
InstanceMapping.instance == instance,
InstanceMapping.resource_adapter_configuration_id == rac.id)
#
# Found something? Delete it and then return...
#
for im in im_list:
node = im.node
node_api.deleteNode(self._sess, node.name)
#
# Check the instance metadata so see if there is a
# matching vm_name...
#
im_list = self._sess.query(InstanceMetadata).filter(
InstanceMetadata.key == "vm_name",
InstanceMetadata.value == instance)
#
# Found something? Delete it and then return...
#
for im in im_list:
if im.instance.resource_adapter_configuration_id != rac.id:
continue
node = im.instance.node
node_api.deleteNode(self._sess, node.name)
def action_configure(self, _, *args, **kwargs):
"""
Configure.
:param _: Unused
:param *args: Unused
:param **kwargs: Unused
:returns: None
"""
try:
result = GlobalParameterDbApi().getParameter(
self.session,
'DHCPLeaseTime'
)
dhcp_lease_time = int(result.getValue())
except ParameterNotFound:
dhcp_lease_time = 2400
try:
result = GlobalParameterDbApi().getParameter(
self.session,
'DNSZone')
dns_zone = result.getValue()
except ParameterNotFound:
dns_zone = ''
installer_node = NodeApi().getInstallerNode(self.session)
self._manager.configure(
dhcp_lease_time,
dns_zone,
self._get_provisioning_nics_ip(installer_node),
self._dhcp_subnets(),
installerNode=installer_node,
bUpdateSysconfig=kwargs.get('bUpdateSysconfig', True),
kit_settings=self._get_kit_settings_dictionary
)
def process_delete_host_request(request):
session = DbManager().openSession()
try:
req = NodeRequestsDbHandler().get_by_addHostSession(
session, request['transaction_id'])
if req is None:
# Session was deleted prior to being process. Nothing to do...
return
ahm.update_session(request['transaction_id'], running=True)
logger.debug('process_delete_host_request(): transaction_id=[{0}],'
' nodespec=[{1}]'.format(request['transaction_id'],
request['nodespec']))
try:
NodeApi().deleteNode(request['nodespec'])
ahm.delete_session(request['transaction_id'])
session.delete(req)
except NodeNotFound:
ahm.delete_session(request['transaction_id'])
session.delete(req)
except TortugaException as exc:
logger.exception('Exception while deleting nodes')
req.message = str(exc)
req.state = 'error'
req.last_update = datetime.datetime.utcnow()
finally:
ahm.update_session(request['transaction_id'], running=False)
finally:
session.commit()
DbManager().closeSession()
def delete_node(self, sir_id):
with spot_cache:
cfg = refresh_spot_instance_request_cache()
if not cfg.has_section(sir_id) or \
not cfg.has_option(sir_id, 'node'):
self._logger.warning('Spot instance [{0}] does not have an'
' associated node'.format(sir_id))
return
spot_instance_node_mapping = cfg.get(sir_id, 'node')
if spot_instance_node_mapping:
self._logger.info('Removing node [{0}] for spot instance'
' request [{1}]'.format(
spot_instance_node_mapping, sir_id))
try:
NodeApi().deleteNode(spot_instance_node_mapping)
except NodeNotFound:
pass
def test_transferNodes_single_node(
self,
get_resourceadapter_class_mock,
default_resource_adapter_mock,
# boot_host_manager_mock,
get_software_profile_metadata_mock,
get_os_boot_host_manager_mock): # pylint: disable=unused-argument
from tortuga.node.nodeApi import NodeApi
with patch('tortuga.node.nodeManager.KitActionsManager.refresh') as \
kit_actions_manager_refresh_mock:
nodeApi = NodeApi()
with self.dbm.session() as session:
nodeApi.transferNodes(session,
'compute2',
nodespec='compute-01.private')
kit_actions_manager_refresh_mock.assert_called()
nodeApi.transferNodes(session,
'compute',
nodespec='compute-01.private')
def __init__(self):
self.cm = ConfigManager()
self.node_api = NodeApi()
self.admin_api = AdminApi()
self.network_api = NetworkApi()
self.parameter_api = ParameterApi()
def node_api(self):
if not self._node_api:
self._node_api = NodeApi()
return self._node_api
class CancelSpotInstanceRequestsCLI(TortugaCli):
def __init__(self):
super(CancelSpotInstanceRequestsCLI, self).__init__(validArgCount=1)
self.nodeApi = NodeApi()
def parseArgs(self, usage=None):
self.addOption('--all', action='store_true', default=False,
help='Cancel all spot instance requests managed by'
' Tortuga')
self.addOption('--terminate', action='store_true', default=False,
help='Terminate any running (fulfilled) instance(s).')
super(CancelSpotInstanceRequestsCLI, self).parseArgs(usage=usage)
if not self.getOptions().all and not self.getArgs():
self.getParser().error(
'<spot instance request id> or --all argument must be'
' specified')
def runCommand(self):
self.parseArgs(
usage='%prog [--terminate] <spot instance request id|--all>')
sir_instance_cache_filename = \
os.path.join(ConfigManager().getRoot(), 'var',
'spot-instances.conf')
cfg = configparser.ConfigParser()
cfg.read(sir_instance_cache_filename)
if self.getOptions().all:
result = self.__get_spot_instance_request_ids(cfg)
else:
sir_id = self.getArgs()[0]
result = [self.__get_spot_instance_request_id(cfg, sir_id)]
if not result:
# Nothing to do...
sys.exit(0)
self.__cancel_spot_instances(result)
def __get_spot_instance_request_id(self, cfg, sir_id):
# Ensure spot instance request id
if not cfg.has_section(sir_id):
sys.stderr.write(
'Spot instance request [{0}] is not managed by'
' Tortuga\n'.format(sir_id))
sys.exit(0)
resource_adapter_configuration = \
self.__get_resource_adapter_configuration(cfg, sir_id)
return sir_id, resource_adapter_configuration
def __get_resource_adapter_configuration(self, cfg, sir_id): \
# pylint: disable=no-self-use
return cfg.get(sir_id, 'resource_adapter_configuration') \
if cfg.has_option(
sir_id, 'resource_adapter_configuration') else \
'default'
def __get_spot_instance_request_ids(self, cfg):
result = []
for sir_id in cfg.sections():
resource_adapter_configuration = \
self.__get_resource_adapter_configuration(cfg, sir_id)
result.append((sir_id, resource_adapter_configuration))
return result
def __get_spot_instance_request_map(self, result): \
# pylint: disable=no-self-use
sir_map = {}
adapter_cfg_map = {}
adapter = Aws()
# Create map of spot instance requests keyed on EC2 region
for sir_id, resource_adapter_configuration in result:
if resource_adapter_configuration not in adapter_cfg_map:
adapter_cfg = adapter.getResourceAdapterConfig(
resource_adapter_configuration)
adapter_cfg_map[resource_adapter_configuration] = \
adapter_cfg
else:
adapter_cfg = adapter_cfg_map[
resource_adapter_configuration]
if adapter_cfg['region'].name not in sir_map:
sir_map[adapter_cfg['region'].name] = []
sir_map[adapter_cfg['region'].name].append(sir_id)
return sir_map
def __cancel_spot_instances(self, result):
sir_map = self.__get_spot_instance_request_map(result)
aws_instance_cache = configparser.ConfigParser()
aws_instance_cache.read('/opt/tortuga/var/aws-instance.conf')
# Iterate on map cancelling requests in each region
for region_name, sir_ids in sir_map.iteritems():
session = boto3.session.Session(region_name=region_name)
ec2_conn = session.client('ec2')
if len(sir_ids) == 1:
print('Cancelling spot instance request [{0}]'
' in region [{1}]'.format(sir_ids[0], region_name))
else:
print('Cancelling {0} spot instance requests in'
' region [{1}]'.format(len(sir_ids), region_name))
response = ec2_conn.describe_spot_instance_requests(
SpotInstanceRequestIds=sir_ids)
# Create list of tuples (sir_id, bool) which indicate if the
# spot instance request should be terminated
cancelled_spot_instance_requests = []
for sir in response['SpotInstanceRequests']:
# All spot instance requests that are 'open' should be
# terminated to avoid leaving orphaned Tortuga node records
cancelled_spot_instance_requests.append(
(sir['SpotInstanceRequestId'],
self.getOptions().terminate or sir['State'] == 'open'))
result = ec2_conn.cancel_spot_instance_requests(
SpotInstanceRequestIds=sir_ids)
# Delete corresponding node entries
for sir_id, terminate in cancelled_spot_instance_requests:
if terminate:
node_name = self.__get_associated_node(
aws_instance_cache, sir_id)
if node_name:
print(' - Deleting node [{0}]'.format(node_name))
self.nodeApi.deleteNode(node_name)
def __get_associated_node(self, aws_instance_cache, sir_id): \
# pylint: disable=no-self-use
node_name = None
for node_name in aws_instance_cache.sections():
if aws_instance_cache.has_option(
node_name, 'spot_instance_request') and \
aws_instance_cache.get(
node_name, 'spot_instance_request') == sir_id:
break
else:
return None
return node_name
def __init__(self):
super(CancelSpotInstanceRequestsCLI, self).__init__(validArgCount=1)
self.nodeApi = NodeApi()
def __fulfilled_request_handler(self, ec2_conn, sir_id, instance_id,
spot_instance_request, hwp):
# Ensure node entries created
resvs = ec2_conn.get_all_instances(instance_ids=[instance_id])
instance = resvs[0].instances[0]
if instance.state not in ('pending', 'running'):
self._logger.info('Ignoring instance [{0}] in state [{1}]'.format(
instance.id, instance.state))
return
# Determine node from spot instance request id
adapter_cfg = configparser.ConfigParser()
adapter_cfg.read('/opt/tortuga/var/aws-instance.conf')
create_node = False
for node_name in adapter_cfg.sections():
if adapter_cfg.has_option(
node_name, 'spot_instance_request') and \
adapter_cfg.get(node_name,
'spot_instance_request') == sir_id:
break
else:
create_node = True
node_name = instance.private_dns_name \
if hwp.getNameFormat() == '*' else None
if create_node:
self._logger.info('Creating node for spot instance'
' [{0}]'.format(instance.id))
# Error: unable to find pre-allocated node record for spot
# instance request
addNodesRequest = {
'softwareProfile':
spot_instance_request['softwareprofile'],
'hardwareProfile':
spot_instance_request['hardwareprofile'],
'isIdle':
False,
'count':
1,
'nodeDetails': [{
'metadata': {
'ec2_instance_id': instance.id,
'ec2_ipaddress': instance.private_ip_address,
}
}],
}
if 'resource_adapter_configuration' in spot_instance_request:
addNodesRequest['resource_adapter_configuration'] = \
spot_instance_request['resource_adapter_configuration']
if node_name:
addNodesRequest['nodeDetails'][0]['name'] = node_name
try:
addHostSession = AddHostWsApi().addNodes(addNodesRequest)
with gevent.Timeout(300):
while True:
response = AddHostWsApi()\
.getStatus(session=addHostSession, getNodes=True)
if not response['running']:
self._logger.debug(
'response: {0}'.format(response))
node_name = response['nodes'][0]['name']
break
gevent.sleep(5)
except gevent.timeout.Timeout:
self._logger.error('Timeout waiting for add nodes operation'
' to complete')
except NodeAlreadyExists:
self._logger.error('Error adding node [{0}]:'
' already exists'.format(
instance.private_dns_name))
else:
self._logger.info('Updating existing node [{0}]'.format(node_name))
# Mark node as 'Provisioned' now that there's a backing instance
NodeApi().updateNode(node_name,
updateNodeRequest={
'state':
state.NODE_STATE_PROVISIONED,
'nics': [{
'ip':
instance.private_ip_address,
}],
'metadata': {
'ec2_instance_id': instance.id,
}
})
update_spot_instance_request_cache(sir_id,
metadata=dict(node=node_name,
status='fulfilled'))
def __init__(self):
super(CancelSpotInstanceRequestsCLI, self).__init__(validArgCount=1)
self.nodeApi = NodeApi()
self._logger = logging.getLogger('tortuga.console')
def test_basic(self, node_state_change_object, get_kit_installer_function,
MockClass1, get_os_object_factory_function, dbm):
"""
Transfer a single node
"""
name = 'compute-01'
with dbm.session() as session:
# xfer node 'compute-01' to 'compute2' software profile
result = NodeApi().transferNodes(session,
'compute2',
nodespec=name)
# get node after xfer
node = NodeApi().getNode(session, name)
# validate new software profile
assert node.getSoftwareProfile().getName() == 'compute2'
# validate state (which is fudged above)
assert node.getState() != 'Installed'
# update status from 'Expired' to 'Installed' to allow xfer
NodeApi().updateNodeStatus(session, name, state='Installed')
# ensure event fired to indicate node state change
node_state_change_object.assert_called()
# xfer node back to 'compute' software profile
NodeApi().transferNodes(session, 'compute', nodespec=name)
NodeApi().updateNodeStatus(session, name, state='Installed')
node = NodeApi().getNode(session, name)
assert node.getSoftwareProfile().getName() == 'compute'