def get_failover_flow(self,
role=constants.ROLE_STANDALONE,
load_balancer=None):
"""Creates a flow to failover a stale amphora
:returns: The flow for amphora failover
"""
failover_amphora_flow = linear_flow.Flow(
constants.FAILOVER_AMPHORA_FLOW)
failover_amphora_flow.add(
lifecycle_tasks.AmphoraToErrorOnRevertTask(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
failover_amphora_flow.add(
network_tasks.FailoverPreparationForAmphora(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
# Note: It seems intuitive to boot an amphora prior to deleting
# the old amphora, however this is a complicated issue.
# If the target host (due to anit-affinity) is resource
# constrained, this will fail where a post-delete will
# succeed. Since this is async with the API it would result
# in the LB ending in ERROR though the amps are still alive.
# Consider in the future making this a complicated
# try-on-failure-retry flow, or move upgrade failovers to be
# synchronous with the API. For now spares pool and act/stdby
# will mitigate most of this delay.
# Delete the old amphora
failover_amphora_flow.add(
database_tasks.MarkAmphoraPendingDeleteInDB(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
failover_amphora_flow.add(
database_tasks.MarkAmphoraHealthBusy(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
failover_amphora_flow.add(
compute_tasks.ComputeDelete(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
failover_amphora_flow.add(
network_tasks.WaitForPortDetach(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
failover_amphora_flow.add(
database_tasks.MarkAmphoraDeletedInDB(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
# If this is an unallocated amp (spares pool), we're done
if not load_balancer:
failover_amphora_flow.add(
database_tasks.DisableAmphoraHealthMonitoring(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
return failover_amphora_flow
# Save failed amphora details for later
failover_amphora_flow.add(
database_tasks.GetAmphoraDetails(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA,
provides=constants.AMP_DATA))
# Get a new amphora
# Note: Role doesn't matter here. We will update it later.
get_amp_subflow = self.get_amphora_for_lb_subflow(
prefix=constants.FAILOVER_AMPHORA_FLOW)
failover_amphora_flow.add(get_amp_subflow)
# Update the new amphora with the failed amphora details
failover_amphora_flow.add(
database_tasks.UpdateAmpFailoverDetails(
requires=(constants.AMPHORA, constants.AMP_DATA)))
# Update the data stored in the flow from the database
failover_amphora_flow.add(
database_tasks.ReloadLoadBalancer(
requires=constants.LOADBALANCER_ID,
provides=constants.LOADBALANCER))
failover_amphora_flow.add(
database_tasks.ReloadAmphora(requires=constants.AMPHORA_ID,
provides=constants.AMPHORA))
# Prepare to reconnect the network interface(s)
failover_amphora_flow.add(
network_tasks.GetAmphoraeNetworkConfigs(
requires=constants.LOADBALANCER,
provides=constants.AMPHORAE_NETWORK_CONFIG))
failover_amphora_flow.add(
database_tasks.GetListenersFromLoadbalancer(
requires=constants.LOADBALANCER, provides=constants.LISTENERS))
failover_amphora_flow.add(
database_tasks.GetAmphoraeFromLoadbalancer(
requires=constants.LOADBALANCER, provides=constants.AMPHORAE))
# Plug the VIP ports into the new amphora
# The reason for moving these steps here is the udp listeners want to
# do some kernel configuration before Listener update for forbidding
# failure during rebuild amphora.
failover_amphora_flow.add(
network_tasks.PlugVIPPort(
requires=(constants.AMPHORA,
constants.AMPHORAE_NETWORK_CONFIG)))
failover_amphora_flow.add(
amphora_driver_tasks.AmphoraPostVIPPlug(
requires=(constants.AMPHORA, constants.LOADBALANCER,
constants.AMPHORAE_NETWORK_CONFIG)))
# Listeners update needs to be run on all amphora to update
# their peer configurations. So parallelize this with an
# unordered subflow.
update_amps_subflow = unordered_flow.Flow(
constants.UPDATE_AMPS_SUBFLOW)
timeout_dict = {
constants.CONN_MAX_RETRIES:
CONF.haproxy_amphora.active_connection_max_retries,
constants.CONN_RETRY_INTERVAL:
CONF.haproxy_amphora.active_connection_rety_interval
}
# Setup parallel flows for each amp. We don't know the new amp
# details at flow creation time, so setup a subflow for each
# amp on the LB, they let the task index into a list of amps
# to find the amphora it should work on.
amp_index = 0
for amp in load_balancer.amphorae:
if amp.status == constants.DELETED:
continue
update_amps_subflow.add(
amphora_driver_tasks.AmpListenersUpdate(
name=constants.AMP_LISTENER_UPDATE + '-' + str(amp_index),
requires=(constants.LOADBALANCER, constants.AMPHORAE),
inject={
constants.AMPHORA_INDEX: amp_index,
constants.TIMEOUT_DICT: timeout_dict
}))
amp_index += 1
failover_amphora_flow.add(update_amps_subflow)
# Plug the member networks into the new amphora
failover_amphora_flow.add(
network_tasks.CalculateAmphoraDelta(
requires=(constants.LOADBALANCER, constants.AMPHORA),
provides=constants.DELTA))
failover_amphora_flow.add(
network_tasks.HandleNetworkDelta(requires=(constants.AMPHORA,
constants.DELTA),
provides=constants.ADDED_PORTS))
failover_amphora_flow.add(
amphora_driver_tasks.AmphoraePostNetworkPlug(
requires=(constants.LOADBALANCER, constants.ADDED_PORTS)))
failover_amphora_flow.add(
database_tasks.ReloadLoadBalancer(
name='octavia-failover-LB-reload-2',
requires=constants.LOADBALANCER_ID,
provides=constants.LOADBALANCER))
# Handle the amphora role and VRRP if necessary
if role == constants.ROLE_MASTER:
failover_amphora_flow.add(
database_tasks.MarkAmphoraMasterInDB(
name=constants.MARK_AMP_MASTER_INDB,
requires=constants.AMPHORA))
vrrp_subflow = self.get_vrrp_subflow(role)
failover_amphora_flow.add(vrrp_subflow)
elif role == constants.ROLE_BACKUP:
failover_amphora_flow.add(
database_tasks.MarkAmphoraBackupInDB(
name=constants.MARK_AMP_BACKUP_INDB,
requires=constants.AMPHORA))
vrrp_subflow = self.get_vrrp_subflow(role)
failover_amphora_flow.add(vrrp_subflow)
elif role == constants.ROLE_STANDALONE:
failover_amphora_flow.add(
database_tasks.MarkAmphoraStandAloneInDB(
name=constants.MARK_AMP_STANDALONE_INDB,
requires=constants.AMPHORA))
failover_amphora_flow.add(
amphora_driver_tasks.ListenersStart(
requires=(constants.LOADBALANCER, constants.AMPHORA)))
failover_amphora_flow.add(
database_tasks.DisableAmphoraHealthMonitoring(
rebind={constants.AMPHORA: constants.FAILED_AMPHORA},
requires=constants.AMPHORA))
return failover_amphora_flow
def get_amphora_for_lb_failover_subflow(self,
prefix,
role=constants.ROLE_STANDALONE,
failed_amp_vrrp_port_id=None,
is_vrrp_ipv6=False,
is_spare=False):
"""Creates a new amphora that will be used in a failover flow.
:requires: loadbalancer_id, flavor, vip, vip_sg_id, loadbalancer
:provides: amphora_id, amphora
:param prefix: The flow name prefix to use on the flow and tasks.
:param role: The role this amphora will have in the topology.
:param failed_amp_vrrp_port_id: The base port ID of the failed amp.
:param is_vrrp_ipv6: True if the base port IP is IPv6.
:param is_spare: True if we are getting a spare amphroa.
:return: A Taskflow sub-flow that will create the amphora.
"""
sf_name = prefix + '-' + constants.CREATE_AMP_FOR_FAILOVER_SUBFLOW
amp_for_failover_flow = linear_flow.Flow(sf_name)
# Try to allocate or boot an amphora instance (unconfigured)
amp_for_failover_flow.add(
self.get_amphora_for_lb_subflow(
prefix=prefix + '-' + constants.FAILOVER_LOADBALANCER_FLOW,
role=role,
is_spare=is_spare))
# If we are getting a spare amphora, this is all we need to do.
if is_spare:
return amp_for_failover_flow
# Create the VIP base (aka VRRP) port for the amphora.
amp_for_failover_flow.add(
network_tasks.CreateVIPBasePort(
name=prefix + '-' + constants.CREATE_VIP_BASE_PORT,
requires=(constants.VIP, constants.VIP_SG_ID,
constants.AMPHORA_ID),
provides=constants.BASE_PORT))
# Attach the VIP base (aka VRRP) port to the amphora.
amp_for_failover_flow.add(
compute_tasks.AttachPort(
name=prefix + '-' + constants.ATTACH_PORT,
requires=(constants.AMPHORA, constants.PORT),
rebind={constants.PORT: constants.BASE_PORT}))
# Update the amphora database record with the VIP base port info.
amp_for_failover_flow.add(
database_tasks.UpdateAmpFailoverDetails(
name=prefix + '-' + constants.UPDATE_AMP_FAILOVER_DETAILS,
requires=(constants.AMPHORA, constants.VIP,
constants.BASE_PORT)))
# Make sure the amphora in the flow storage is up to date
# or the vrrp_ip will be empty
amp_for_failover_flow.add(
database_tasks.ReloadAmphora(name=prefix + '-' +
constants.RELOAD_AMPHORA,
requires=constants.AMPHORA_ID,
provides=constants.AMPHORA))
# Update the amphora networking for the plugged VIP port
amp_for_failover_flow.add(
network_tasks.GetAmphoraNetworkConfigsByID(
name=prefix + '-' +
constants.GET_AMPHORA_NETWORK_CONFIGS_BY_ID,
requires=(constants.LOADBALANCER_ID, constants.AMPHORA_ID),
provides=constants.AMPHORAE_NETWORK_CONFIG))
# Disable the base (vrrp) port on the failed amphora
# This prevents a DAD failure when bringing up the new amphora.
# Keepalived will handle this for act/stdby.
if (role == constants.ROLE_STANDALONE and failed_amp_vrrp_port_id
and is_vrrp_ipv6):
amp_for_failover_flow.add(
network_tasks.AdminDownPort(
name=prefix + '-' + constants.ADMIN_DOWN_PORT,
inject={constants.PORT_ID: failed_amp_vrrp_port_id}))
amp_for_failover_flow.add(
amphora_driver_tasks.AmphoraPostVIPPlug(
name=prefix + '-' + constants.AMPHORA_POST_VIP_PLUG,
requires=(constants.AMPHORA, constants.LOADBALANCER,
constants.AMPHORAE_NETWORK_CONFIG)))
# Plug member ports
amp_for_failover_flow.add(
network_tasks.CalculateAmphoraDelta(
name=prefix + '-' + constants.CALCULATE_AMPHORA_DELTA,
requires=(constants.LOADBALANCER, constants.AMPHORA,
constants.AVAILABILITY_ZONE, constants.VRRP_PORT),
rebind={constants.VRRP_PORT: constants.BASE_PORT},
provides=constants.DELTA))
amp_for_failover_flow.add(
network_tasks.HandleNetworkDelta(
name=prefix + '-' + constants.HANDLE_NETWORK_DELTA,
requires=(constants.AMPHORA, constants.DELTA),
provides=constants.ADDED_PORTS))
amp_for_failover_flow.add(
amphora_driver_tasks.AmphoraePostNetworkPlug(
name=prefix + '-' + constants.AMPHORAE_POST_NETWORK_PLUG,
requires=(constants.LOADBALANCER, constants.ADDED_PORTS)))
return amp_for_failover_flow
def test_handle_network_delta(self, mock_get_net_driver):
mock_net_driver = mock.MagicMock()
mock_get_net_driver.return_value = mock_net_driver
nic1 = mock.MagicMock()
nic1.network_id = uuidutils.generate_uuid()
nic2 = mock.MagicMock()
nic2.network_id = uuidutils.generate_uuid()
interface1 = mock.MagicMock()
interface1.port_id = uuidutils.generate_uuid()
port1 = mock.MagicMock()
port1.network_id = uuidutils.generate_uuid()
fixed_ip = mock.MagicMock()
fixed_ip.subnet_id = uuidutils.generate_uuid()
port1.fixed_ips = [fixed_ip]
subnet = mock.MagicMock()
network = mock.MagicMock()
delta = data_models.Delta(amphora_id=self.amphora_mock.id,
compute_id=self.amphora_mock.compute_id,
add_nics=[nic1],
delete_nics=[nic2, nic2, nic2])
mock_net_driver.plug_network.return_value = interface1
mock_net_driver.get_port.return_value = port1
mock_net_driver.get_network.return_value = network
mock_net_driver.get_subnet.return_value = subnet
mock_net_driver.unplug_network.side_effect = [
None, net_base.NetworkNotFound, Exception
]
handle_net_delta_obj = network_tasks.HandleNetworkDelta()
result = handle_net_delta_obj.execute(self.amphora_mock, delta)
mock_net_driver.plug_network.assert_called_once_with(
self.amphora_mock.compute_id, nic1.network_id)
mock_net_driver.get_port.assert_called_once_with(interface1.port_id)
mock_net_driver.get_network.assert_called_once_with(port1.network_id)
mock_net_driver.get_subnet.assert_called_once_with(fixed_ip.subnet_id)
self.assertEqual({self.amphora_mock.id: [port1]}, result)
mock_net_driver.unplug_network.assert_called_with(
self.amphora_mock.compute_id, nic2.network_id)
# Revert
delta2 = data_models.Delta(amphora_id=self.amphora_mock.id,
compute_id=self.amphora_mock.compute_id,
add_nics=[nic1, nic1],
delete_nics=[nic2, nic2, nic2])
mock_net_driver.unplug_network.reset_mock()
handle_net_delta_obj.revert(
failure.Failure.from_exception(Exception('boom')), None, None)
mock_net_driver.unplug_network.assert_not_called()
mock_net_driver.unplug_network.reset_mock()
handle_net_delta_obj.revert(None, None, None)
mock_net_driver.unplug_network.assert_not_called()
mock_net_driver.unplug_network.reset_mock()
handle_net_delta_obj.revert(None, None, delta2)
