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 test_update(self, config_mock):
"""
Test that mconfig updates are handled correctly
"""
# Set up fixture data
# Update will simulate gateway moving from
# test_mconfig -> updated_mconfig
TestUpdate = namedtuple('TestUpdate', ['value', 'key'])
test_mconfig = GatewayConfigs()
updated_mconfig = GatewayConfigs()
some_any = Any()
magmad = MagmaD(log_level=1)
some_any.Pack(magmad)
test_mconfig.configs_by_key['magmad'].CopyFrom(some_any)
updated_mconfig.configs_by_key['magmad'].CopyFrom(some_any)
metricsd = MetricsD(log_level=2)
some_any.Pack(metricsd)
test_mconfig.configs_by_key['metricsd'].CopyFrom(some_any)
metricsd = MetricsD(log_level=3)
some_any.Pack(metricsd)
updated_mconfig.configs_by_key['metricsd'].CopyFrom(some_any)
# Set up mock dependencies
config_mock.return_value = {
'magma_services': ['magmad', 'metricsd'],
}
@asyncio.coroutine
def _mock_restart_services(): return "blah"
service_manager_mock = MagicMock()
magmad_service_mock = MagicMock()
mconfig_manager_mock = MconfigManagerImpl()
load_mock = patch.object(
mconfig_manager_mock,
'load_mconfig', MagicMock(return_value=test_mconfig),
)
update_mock = patch.object(
mconfig_manager_mock,
'update_stored_mconfig', Mock(),
)
restart_service_mock = patch.object(
service_manager_mock,
'restart_services', MagicMock(wraps=_mock_restart_services),
)
with load_mock as loader, update_mock as updater, \
restart_service_mock as restarter:
loop = asyncio.new_event_loop()
config_manager = ConfigManager(
['magmad', 'metricsd'], service_manager_mock,
magmad_service_mock, mconfig_manager_mock,
allow_unknown_fields=False,
loop=loop,
)
# Verify that config update restarts all services
update_str = MessageToJson(updated_mconfig)
updates = [
TestUpdate(value='', key='some key'),
TestUpdate(
value=update_str.encode('utf-8'),
key='last key',
),
]
config_manager.process_update(CONFIG_STREAM_NAME, updates, False)
# Only metricsd config was updated, hence should be restarted
loader.assert_called_once_with()
restarter.assert_called_once_with(['metricsd'])
updater.assert_called_once_with(update_str)
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 test_update(self, config_mock):
"""
Test that mconfig updates are handled correctly
"""
# Set up fixture data
# Update will simulate gateway moving from
# test_mconfig -> updated_mconfig
test_mconfig = GatewayConfigs()
updated_mconfig = GatewayConfigs()
some_any = Any()
magmad = MagmaD(log_level=1)
some_any.Pack(magmad)
test_mconfig.configs_by_key['magmad'].CopyFrom(some_any)
updated_mconfig.configs_by_key['magmad'].CopyFrom(some_any)
metricsd = MetricsD(log_level=2)
some_any.Pack(metricsd)
test_mconfig.configs_by_key['metricsd'].CopyFrom(some_any)
metricsd = MetricsD(log_level=3)
some_any.Pack(metricsd)
updated_mconfig.configs_by_key['metricsd'].CopyFrom(some_any)
# Set up mock dependencies
config_mock.return_value = {
'magma_services': ['magmad', 'metricsd'],
}
@asyncio.coroutine
def _mock_restart_services():
return "blah"
@asyncio.coroutine
def _mock_update_dynamic_services():
return "mockResponse"
service_manager_mock = MagicMock()
magmad_service_mock = MagicMock()
mconfig_manager_mock = MconfigManagerImpl()
load_mock = patch.object(
mconfig_manager_mock,
'load_mconfig', MagicMock(return_value=test_mconfig),
)
update_mock = patch.object(
mconfig_manager_mock,
'update_stored_mconfig', Mock(),
)
restart_service_mock = patch.object(
service_manager_mock,
'restart_services', MagicMock(wraps=_mock_restart_services),
)
update_dynamic_services_mock = patch.object(
service_manager_mock,
'update_dynamic_services', MagicMock(wraps=_mock_update_dynamic_services),
)
processed_updates_mock = patch('magma.magmad.events.processed_updates')
class ServiceMconfigMock:
def __init__(self, service, mconfig_struct):
pass
dynamic_services = []
mock_mcfg = patch('magma.magmad.config_manager.load_service_mconfig', MagicMock(wraps=ServiceMconfigMock))
with load_mock as loader,\
update_mock as updater, \
restart_service_mock as restarter,\
update_dynamic_services_mock as dynamic_services,\
mock_mcfg as mock_c,\
processed_updates_mock as processed_updates:
loop = asyncio.new_event_loop()
config_manager = ConfigManager(
['magmad', 'metricsd'], service_manager_mock,
magmad_service_mock, mconfig_manager_mock,
allow_unknown_fields=False,
loop=loop,
)
# Process an empty set of updates
updates = []
config_manager.process_update(CONFIG_STREAM_NAME, updates, False)
# No services should be updated or restarted due to empty updates
restarter.assert_not_called()
updater.assert_not_called()
# Verify that config update restarts all services
update_str = MessageToJson(updated_mconfig)
updates = [
DataUpdate(value=''.encode('utf-8'), key='some key'),
DataUpdate(
value=update_str.encode('utf-8'),
key='last key',
),
]
config_manager.process_update(CONFIG_STREAM_NAME, updates, False)
# Only metricsd config was updated, hence should be restarted
self.assertEqual(loader.call_count, 1)
restarter.assert_called_once_with(['metricsd'])
updater.assert_called_once_with(update_str)
configs_by_service = {
'magmad': updated_mconfig.configs_by_key['magmad'],
'metricsd': updated_mconfig.configs_by_key['metricsd'],
}
processed_updates.assert_called_once_with(configs_by_service)
restarter.reset_mock()
updater.reset_mock()
processed_updates.reset_mock()
updated_mconfig.configs_by_key['shared_mconfig'].CopyFrom(some_any)
update_str = MessageToJson(updated_mconfig)
updates = [
DataUpdate(
value=update_str.encode('utf-8'),
key='last key',
),
]
config_manager.process_update(CONFIG_STREAM_NAME, updates, False)
# shared config update should restart all services
restarter.assert_called_once_with(['magmad', 'metricsd'])
updater.assert_called_once_with(update_str)
dynamic_services.assert_called_once_with([])
processed_updates.assert_called_once_with(configs_by_service)
self.assertEqual(mock_c.call_count, 1)
restarter.reset_mock()
updater.reset_mock()
processed_updates.reset_mock()
dynamic_services.reset_mock()
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()