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,
)
# 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()
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
if 'td-agent-bit' in services:
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,
)
# 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 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, 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 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()