def ovs_reset_bridges():
mconfig = get_mconfig_manager().load_service_mconfig(
'pipelined', mconfigs_pb2.PipelineD(),
)
service_config = load_service_config('pipelined')
if service_config.get('enable_nat', mconfig.nat_enabled):
non_nat_sgi_interface = ""
sgi_management_iface_ip_addr = ""
sgi_management_iface_gw = ""
else:
non_nat_sgi_interface = service_config['nat_iface']
sgi_management_iface_ip_addr = service_config.get(
'sgi_management_iface_ip_addr', mconfig.sgi_management_iface_ip_addr,
)
sgi_management_iface_gw = service_config.get(
'sgi_management_iface_gw', mconfig.sgi_management_iface_gw,
)
sgi_bridge_name = service_config.get('uplink_bridge', "uplink_br0")
reset_br = "magma-bridge-reset.sh -n %s %s %s %s" % \
(
sgi_bridge_name,
non_nat_sgi_interface,
sgi_management_iface_ip_addr,
sgi_management_iface_gw,
)
print("ovs-restart: ", reset_br)
subprocess.call(reset_br.split())
def test_happy_path_feature_unspecified(self, new_manager_mock):
manager_instance = Mock()
manager_instance.load_service_mconfig.return_value = MagmaD()
new_manager_mock.return_value = manager_instance
actual = get_mconfig_manager()
self.assertIsInstance(actual, MconfigManagerImpl)
manager_instance.load_service_mconfig.assert_called_once_with('magmad')
def reload_mconfig(self):
"""
Reloads the managed config for the service
"""
try:
# reload mconfig manager in case feature flag for streaming changed
self._mconfig_manager = get_mconfig_manager()
self._mconfig = self._mconfig_manager.load_service_mconfig(
self._name, )
except LoadConfigError as e:
logging.warning(e)
def __init__(self, name, empty_mconfig, loop=None):
self._name = name
self._port = 0
self._get_status_callback = None
self._get_operational_states_cb = None
self._log_count_handler = MsgCounterHandler()
# Init logging before doing anything
logging.basicConfig(
level=logging.INFO,
format='[%(asctime)s %(levelname)s %(name)s] %(message)s')
# Add a handler to count errors
logging.root.addHandler(self._log_count_handler)
# Set gRPC polling strategy
self._set_grpc_poll_strategy()
# Load the managed config if present
self._mconfig = empty_mconfig
self._mconfig_metadata = None
self._mconfig_manager = get_mconfig_manager()
self.reload_mconfig()
self._state = ServiceInfo.STARTING
self._health = ServiceInfo.APP_UNHEALTHY
if loop is None:
loop = asyncio.get_event_loop()
self._loop = loop
self._start_time = int(time.time())
self._register_signal_handlers()
# Load the service config if present
self._config = None
self.reload_config()
# Count errors
self.log_counter = ServiceLogErrorReporter(
loop=self._loop,
service_config=self._config,
handler=self._log_count_handler,
)
self.log_counter.start()
# Operational States
self._operational_states = []
self._version = '0.0.0'
# Load the service version if available
try:
self._version = pkg_resources.get_distribution('orc8r').version
except pkg_resources.ResolutionError as e:
logging.info(e)
self._server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
add_Service303Servicer_to_server(self, self._server)
def test_happy_path_feature_off(self, new_manager_mock):
"""
Test happy path with feature flag turned off explicitly
"""
manager_instance = Mock()
manager_instance.load_service_mconfig.return_value = MagmaD(
feature_flags={'kafka_config_streamer': False}, )
new_manager_mock.return_value = manager_instance
actual = get_mconfig_manager()
self.assertIsInstance(actual, MconfigManagerImpl)
manager_instance.load_service_mconfig.assert_called_once_with('magmad')
def test_happy_path_feature_on(self, new_manager_mock):
"""
Test happy path with feature flag turned on
"""
manager_instance = Mock()
manager_instance.load_service_mconfig.return_value = MagmaD(
feature_flags={'kafka_config_streamer': True}, )
new_manager_mock.return_value = manager_instance
actual = get_mconfig_manager()
# We should have instantiated a StreamedMconfigManager, which in the
# context of this test method will return the mock instance.
self.assertIs(actual, manager_instance)
manager_instance.load_service_mconfig.assert_called_once_with('magmad')
def reload_mconfig(self):
"""Reload the managed config for the service"""
# reload mconfig manager in case feature flag for streaming changed
self._mconfig_manager = get_mconfig_manager()
try:
self._mconfig_metadata = self._mconfig_manager.load_mconfig_metadata(
)
self._mconfig = self._mconfig_manager.load_service_mconfig(
self._name,
self._mconfig,
)
self._shared_mconfig = self._mconfig_manager.load_shared_mconfig(
self._shared_mconfig)
except LoadConfigError as e:
logging.warning(e)
def __init__(self, name, loop=None):
self._name = name
self._port = 0
self._get_status_callback = None
self._reload_config_callback = None
# Init logging before doing anything
logging.basicConfig(
level=logging.INFO,
format='[%(asctime)s %(levelname)s %(name)s] %(message)s')
# Set gRPC polling strategy
self._set_grpc_poll_strategy()
# Load the managed config if present
self._mconfig = None
self._mconfig_manager = get_mconfig_manager()
self.reload_mconfig()
self._state = ServiceInfo.STARTING
self._health = ServiceInfo.APP_UNHEALTHY
if loop is None:
loop = asyncio.get_event_loop()
self._loop = loop
self._start_time = int(time.time())
self._register_signal_handlers()
# Load the service config if present
self._config = None
try:
self._config = load_service_config(name)
except LoadConfigError as e:
logging.warning(e)
self._setup_logging()
self._version = '0.0.0'
# Load the service version if available
try:
self._version = pkg_resources.get_distribution('orc8r').version
except pkg_resources.ResolutionError as e:
logging.info(e)
self._server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
add_Service303Servicer_to_server(self, self._server)
def test_new_mconfig_load_error(self, new_manager_mock, old_manager_mock):
"""
Test feature flag on, but new mconfig manager errors out on load
"""
new_manager_instance = Mock()
new_manager_instance.load_service_mconfig = Mock(
side_effect=LoadConfigError('mock'), )
new_manager_mock.return_value = new_manager_instance
old_manager_instance = Mock()
old_manager_instance.load_service_mconfig.return_value = MagmaD(
feature_flags={'kafka_config_streamer': True}, )
old_manager_mock.return_value = old_manager_instance
actual = get_mconfig_manager()
self.assertIs(actual, new_manager_instance)
new_manager_instance.load_service_mconfig\
.assert_called_once_with('magmad')
old_manager_instance.load_service_mconfig\
.assert_called_once_with('magmad')
def test_both_mconfigs_load_error(self, new_manager_mock,
old_manager_mock):
"""
Test both mconfig managers erroring out on load
"""
new_manager_instance = Mock()
new_manager_instance.load_service_mconfig = Mock(
side_effect=LoadConfigError('mock'), )
new_manager_mock.return_value = new_manager_instance
old_manager_instance = Mock()
old_manager_instance.load_service_mconfig = Mock(
side_effect=LoadConfigError('mock2'), )
old_manager_mock.return_value = old_manager_instance
actual = get_mconfig_manager()
self.assertIs(actual, old_manager_instance)
new_manager_instance.load_service_mconfig \
.assert_called_once_with('magmad')
old_manager_instance.load_service_mconfig \
.assert_called_once_with('magmad')
async def bootstrap_success_cb(certs_generated: bool):
nonlocal first_time_bootstrap
if first_time_bootstrap:
if stream_client:
stream_client.start()
if sync_rpc_client:
sync_rpc_client.start()
first_time_bootstrap = False
if certs_generated:
svcs_to_restart = []
if 'control_proxy' in services:
svcs_to_restart.append('control_proxy')
# fluent-bit caches TLS client certs in memory, so we need to
# restart it whenever the certs change
fresh_mconfig = get_mconfig_manager().load_service_mconfig(
'magmad',
mconfigs_pb2.MagmaD(),
)
dynamic_svcs = fresh_mconfig.dynamic_services or []
if 'td-agent-bit' in dynamic_svcs:
svcs_to_restart.append('td-agent-bit')
await service_manager.restart_services(services=svcs_to_restart)
def main():
"""
Main magmad function
"""
service = MagmaService('magmad', mconfigs_pb2.MagmaD())
# Optionally pipe errors to Sentry
sentry_init(service_name=service.name)
logging.info('Starting magmad for UUID: %s', snowflake.make_snowflake())
# Create service manager
services = service.config.get('magma_services')
init_system = service.config.get('init_system', 'systemd')
registered_dynamic_services = service.config.get(
'registered_dynamic_services',
[],
)
enabled_dynamic_services = []
if service.mconfig is not None:
enabled_dynamic_services = service.mconfig.dynamic_services
# Poll the services' Service303 interface
service_poller = ServicePoller(
service.loop,
service.config,
enabled_dynamic_services,
)
service_poller.start()
service_manager = ServiceManager(
services,
init_system,
service_poller,
registered_dynamic_services,
enabled_dynamic_services,
)
# Get metrics service config
metrics_config = service.config.get('metricsd')
metrics_services = metrics_config['services']
collect_interval = metrics_config['collect_interval']
sync_interval = metrics_config['sync_interval']
grpc_timeout = metrics_config['grpc_timeout']
grpc_msg_size = metrics_config.get('max_grpc_msg_size_mb', 4)
metrics_post_processor_fn = metrics_config.get('post_processing_fn')
metric_scrape_targets = [
ScrapeTarget(t['url'], t['name'], t['interval'])
for t in metrics_config.get('metric_scrape_targets', [])
]
# Create local metrics collector
metrics_collector = MetricsCollector(
services=metrics_services,
collect_interval=collect_interval,
sync_interval=sync_interval,
grpc_timeout=grpc_timeout,
grpc_max_msg_size_mb=grpc_msg_size,
loop=service.loop,
post_processing_fn=get_metrics_postprocessor_fn(
metrics_post_processor_fn, ),
scrape_targets=metric_scrape_targets,
)
# Poll and sync the metrics collector loops
metrics_collector.run()
# Start a background thread to stream updates from the cloud
stream_client = None
if service.config.get('enable_config_streamer', False):
stream_client = StreamerClient(
{
CONFIG_STREAM_NAME:
ConfigManager(
services,
service_manager,
service,
MconfigManagerImpl(),
),
},
service.loop,
)
# Create sync rpc client with a heartbeat of 30 seconds (timeout = 60s)
sync_rpc_client = None
if service.config.get('enable_sync_rpc', False):
sync_rpc_client = SyncRPCClient(
service.loop,
30,
service.config.get('print_grpc_payload', False),
)
first_time_bootstrap = True
# This is called when bootstrap succeeds and when _bootstrap_check is
# invoked but bootstrap is not needed. If it's invoked right after certs
# are generated, certs_generated is true, control_proxy will restart.
async def bootstrap_success_cb(certs_generated: bool):
nonlocal first_time_bootstrap
if first_time_bootstrap:
if stream_client:
stream_client.start()
if sync_rpc_client:
sync_rpc_client.start()
first_time_bootstrap = False
if certs_generated:
svcs_to_restart = []
if 'control_proxy' in services:
svcs_to_restart.append('control_proxy')
# fluent-bit caches TLS client certs in memory, so we need to
# restart it whenever the certs change
fresh_mconfig = get_mconfig_manager().load_service_mconfig(
'magmad',
mconfigs_pb2.MagmaD(),
)
dynamic_svcs = fresh_mconfig.dynamic_services or []
if 'td-agent-bit' in dynamic_svcs:
svcs_to_restart.append('td-agent-bit')
await service_manager.restart_services(services=svcs_to_restart)
# Create bootstrap manager
bootstrap_manager = BootstrapManager(service, bootstrap_success_cb)
# Initialize kernel version poller if it is enabled
kernel_version_poller = None
if service.config.get('enable_kernel_version_checking', False):
kernel_version_poller = KernelVersionsPoller(service)
kernel_version_poller.start()
# gateway status generator to bundle various information about this
# gateway into an object.
gateway_status_factory = GatewayStatusFactory(
service=service,
service_poller=service_poller,
kernel_version_poller=kernel_version_poller,
)
# _grpc_client_manager to manage grpc client recycling
grpc_client_manager = GRPCClientManager(
service_name="state",
service_stub=StateServiceStub,
max_client_reuse=60,
)
# Initialize StateReporter
state_reporter = StateReporter(
config=service.config,
mconfig=service.mconfig,
loop=service.loop,
bootstrap_manager=bootstrap_manager,
gw_status_factory=gateway_status_factory,
grpc_client_manager=grpc_client_manager,
)
# Initialize ServiceHealthWatchdog
service_health_watchdog = ServiceHealthWatchdog(
config=service.config,
loop=service.loop,
service_poller=service_poller,
service_manager=service_manager,
)
# Start _bootstrap_manager
bootstrap_manager.start_bootstrap_manager()
# Start all services when magmad comes up
service.loop.create_task(service_manager.start_services())
# Start state reporting loop
state_reporter.start()
# Start service timeout health check loop
service_health_watchdog.start()
# Start upgrade manager loop
if service.config.get('enable_upgrade_manager', False):
upgrader = _get_upgrader_impl(service)
service.loop.create_task(start_upgrade_loop(service, upgrader))
# Start network health metric collection
if service.config.get('enable_network_monitor', False):
service.loop.create_task(metrics_collection_loop(service.config))
# Create generic command executor
command_executor = None
if service.config.get('generic_command_config', None):
command_executor = get_command_executor_impl(service)
# Start loop to monitor unattended upgrade status
service.loop.create_task(monitor_unattended_upgrade_status())
# Add all servicers to the server
magmad_servicer = MagmadRpcServicer(
service,
services,
service_manager,
get_mconfig_manager(),
command_executor,
service.loop,
service.config.get('print_grpc_payload', False),
)
magmad_servicer.add_to_server(service.rpc_server)
if SDWatchdog.has_notify():
# Create systemd watchdog
sdwatchdog = SDWatchdog(
tasks=[bootstrap_manager, state_reporter],
update_status=True,
)
# Start watchdog loop
service.loop.create_task(sdwatchdog.run())
# Run the service loop
service.run()
# Cleanup the service
service.close()
def __init__(self, name, empty_mconfig, loop=None):
self._name = name
self._port = 0
self._get_status_callback = None
self._get_operational_states_cb = None
self._log_count_handler = MsgCounterHandler()
# Init logging before doing anything
logging.basicConfig(
level=logging.INFO,
format='[%(asctime)s %(levelname)s %(name)s] %(message)s')
# Add a handler to count errors
logging.root.addHandler(self._log_count_handler)
# Set gRPC polling strategy
self._set_grpc_poll_strategy()
# Load the managed config if present
self._mconfig = empty_mconfig
self._mconfig_metadata = None
self._mconfig_manager = get_mconfig_manager()
self.reload_mconfig()
self._state = ServiceInfo.STARTING
self._health = ServiceInfo.APP_UNHEALTHY
if loop is None:
loop = asyncio.get_event_loop()
self._loop = loop
self._start_time = int(time.time())
self._register_signal_handlers()
# Load the service config if present
self._config = None
self.reload_config()
# Count errors
self.log_counter = ServiceLogErrorReporter(
loop=self._loop,
service_config=self._config,
handler=self._log_count_handler,
)
self.log_counter.start()
# Operational States
self._operational_states = []
self._version = '0.0.0'
# Load the service version if available
try:
# Check if service on docker
if self._config and 'init_system' in self._config \
and self._config['init_system'] == 'docker':
# image comes in form of "feg_gateway_python:<IMAGE_TAG>\n"
# Skip the "feg_gateway_python:" part
image = os.popen(
'docker ps --filter name=magmad --format "{{.Image}}" | '
'cut -d ":" -f 2')
image_tag = image.read().strip('\n')
self._version = image_tag
else:
self._version = pkg_resources.get_distribution('orc8r').version
except pkg_resources.ResolutionError as e:
logging.info(e)
self._server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
add_Service303Servicer_to_server(self, self._server)
def main():
"""
Main magmad function
"""
service = MagmaService('magmad')
logging.info('Starting magmad for UUID: %s', snowflake.make_snowflake())
# Create service manager
services = service.config['magma_services']
init_system = service.config.get('init_system', 'systemd')
registered_dynamic_services = service.config.get(
'registered_dynamic_services', [])
enabled_dynamic_services = []
if service.mconfig is not None:
enabled_dynamic_services = service.mconfig.dynamic_services
# Poll the services' Service303 interface
service_poller = ServicePoller(service.loop, service.config)
service_poller.start()
service_manager = ServiceManager(services, init_system, service_poller,
registered_dynamic_services,
enabled_dynamic_services)
# Start a background thread to stream updates from the cloud
stream_client = None
if service.config.get('enable_config_streamer', False):
stream_client = StreamerClient(
{
CONFIG_STREAM_NAME:
ConfigManager(
services,
service_manager,
service,
MconfigManagerImpl(),
),
MCONFIG_VIEW_STREAM_NAME:
StreamingMconfigCallback(
services,
service_manager,
service,
StreamedMconfigManager(),
)
},
service.loop,
)
# Schedule periodic checkins
checkin_manager = CheckinManager(service, service_poller)
# Create sync rpc client with a timeout of 60 seconds
sync_rpc_client = None
if service.config.get('enable_sync_rpc', False):
sync_rpc_client = SyncRPCClient(service.loop, 60)
first_time_bootstrap = True
# This is called when bootstrap succeeds and when _bootstrap_check is
# invoked but bootstrap is not needed. If it's invoked right after certs
# are generated, certs_generated is true, control_proxy will restart.
def bootstrap_success_cb(certs_generated):
nonlocal first_time_bootstrap
if first_time_bootstrap:
if stream_client:
stream_client.start()
checkin_manager.try_checkin()
if sync_rpc_client:
sync_rpc_client.start()
first_time_bootstrap = False
if certs_generated and 'control_proxy' in services:
service.loop.create_task(
service_manager.restart_services(services=['control_proxy']))
# Create bootstrap manager
bootstrap_manager = BootstrapManager(service, bootstrap_success_cb)
def checkin_failure_cb(err_code):
bootstrap_manager.on_checkin_fail(err_code)
checkin_manager.set_failure_cb(checkin_failure_cb)
# Start bootstrap_manager after checkin_manager's callback is set
bootstrap_manager.start_bootstrap_manager()
# Start all services when magmad comes up
service.loop.create_task(service_manager.start_services())
# Start upgrade manager loop
if service.config.get('enable_upgrade_manager', False):
upgrader = _get_upgrader_impl(service)
service.loop.create_task(start_upgrade_loop(service, upgrader))
# Start network health metric collection
if service.config.get('enable_network_monitor', False):
service.loop.create_task(metrics_collection_loop(service.config))
if service.config.get('enable_systemd_tailer', False):
service.loop.create_task(start_systemd_tailer(service.config))
# Start loop to monitor unattended upgrade status
service.loop.create_task(monitor_unattended_upgrade_status(service.loop))
# Add all servicers to the server
magmad_servicer = MagmadRpcServicer(
service,
services,
service_manager,
get_mconfig_manager(),
service.loop,
)
magmad_servicer.add_to_server(service.rpc_server)
if SDWatchdog.has_notify():
# Create systemd watchdog
sdwatchdog = SDWatchdog(tasks=[bootstrap_manager, checkin_manager],
update_status=True)
# Start watchdog loop
service.loop.create_task(sdwatchdog.run())
# Run the service loop
service.run()
# Cleanup the service
service.close()
def main():
"""
Main magmad function
"""
service = MagmaService('magmad', mconfigs_pb2.MagmaD())
logging.info('Starting magmad for UUID: %s', snowflake.make_snowflake())
# Create service manager
services = service.config['magma_services']
init_system = service.config.get('init_system', 'systemd')
registered_dynamic_services = service.config.get(
'registered_dynamic_services', [])
enabled_dynamic_services = []
if service.mconfig is not None:
enabled_dynamic_services = service.mconfig.dynamic_services
# Poll the services' Service303 interface
service_poller = ServicePoller(service.loop, service.config)
service_poller.start()
service_manager = ServiceManager(services, init_system, service_poller,
registered_dynamic_services,
enabled_dynamic_services)
# Get metrics service config
metrics_config = service.config['metricsd']
metrics_services = metrics_config['services']
collect_interval = metrics_config['collect_interval']
sync_interval = metrics_config['sync_interval']
grpc_timeout = metrics_config['grpc_timeout']
queue_length = metrics_config['queue_length']
metrics_post_processor_fn = metrics_config.get('post_processing_fn')
# Create local metrics collector
metrics_collector = MetricsCollector(
metrics_services,
collect_interval,
sync_interval,
grpc_timeout,
queue_length,
service.loop,
get_metrics_postprocessor_fn(metrics_post_processor_fn),
)
# Poll and sync the metrics collector loops
metrics_collector.run()
# Start a background thread to stream updates from the cloud
stream_client = None
if service.config.get('enable_config_streamer', False):
stream_client = StreamerClient(
{
CONFIG_STREAM_NAME:
ConfigManager(
services,
service_manager,
service,
MconfigManagerImpl(),
),
},
service.loop,
)
# Schedule periodic checkins
checkin_manager = CheckinManager(service, service_poller)
# Create sync rpc client with a heartbeat of 30 seconds (timeout = 60s)
sync_rpc_client = None
if service.config.get('enable_sync_rpc', False):
sync_rpc_client = SyncRPCClient(service.loop, 30)
first_time_bootstrap = True
# This is called when bootstrap succeeds and when _bootstrap_check is
# invoked but bootstrap is not needed. If it's invoked right after certs
# are generated, certs_generated is true, control_proxy will restart.
async def bootstrap_success_cb(certs_generated):
nonlocal first_time_bootstrap
if first_time_bootstrap:
if stream_client:
stream_client.start()
await checkin_manager.try_checkin()
if sync_rpc_client:
sync_rpc_client.start()
first_time_bootstrap = False
if certs_generated and 'control_proxy' in services:
service.loop.create_task(
service_manager.restart_services(services=['control_proxy']))
# Create bootstrap manager
bootstrap_manager = BootstrapManager(service, bootstrap_success_cb)
async def checkin_failure_cb(err_code):
await bootstrap_manager.on_checkin_fail(err_code)
checkin_manager.set_failure_cb(checkin_failure_cb)
# Start bootstrap_manager after checkin_manager's callback is set
bootstrap_manager.start_bootstrap_manager()
# Schedule periodic state reporting
state_manager = StateReporter(service, checkin_manager)
state_manager.start()
# Start all services when magmad comes up
service.loop.create_task(service_manager.start_services())
# Start upgrade manager loop
if service.config.get('enable_upgrade_manager', False):
upgrader = _get_upgrader_impl(service)
service.loop.create_task(start_upgrade_loop(service, upgrader))
# Start network health metric collection
if service.config.get('enable_network_monitor', False):
service.loop.create_task(metrics_collection_loop(service.config))
if service.config.get('enable_systemd_tailer', False):
service.loop.create_task(start_systemd_tailer(service.config))
# Create generic command executor
command_executor = None
if service.config.get('generic_command_config', None):
command_executor = get_command_executor_impl(service)
# Start loop to monitor unattended upgrade status
service.loop.create_task(monitor_unattended_upgrade_status(service.loop))
# Add all servicers to the server
magmad_servicer = MagmadRpcServicer(
service,
services,
service_manager,
get_mconfig_manager(),
command_executor,
service.loop,
)
magmad_servicer.add_to_server(service.rpc_server)
if SDWatchdog.has_notify():
# Create systemd watchdog
sdwatchdog = SDWatchdog(tasks=[bootstrap_manager, checkin_manager],
update_status=True)
# Start watchdog loop
service.loop.create_task(sdwatchdog.run())
# Run the service loop
service.run()
# Cleanup the service
service.close()