Startup of openstack service

Basically everything openstack Services all depend on evenlet Complete various concurrent tasks , Its processes can be divided into two categories ：
1、 WSGIService: Receive and process http request , rely on eventlet.wsgi  Of  wsgi server  Handle http request , such as nova-api
2、 Service: Receive and process rpc request , Such as  nova-operation etc.
Whether it's  WSGIService  still  Service  Type of process , Every time a request is received (http or rpc), Will allocate a process in the process pool to process the request

One 、WSGIService Start of

Let's say nova Service as an example .
nova-api from  nova/cmd/api.py  start-up , It initializes a WSGIService( from  service.py  Definition ) object .

def main():
    objects.register_all()
    CONF(sys.argv[1:], project='nova',
         version=version.version_string())
    logging.setup(CONF, "nova")

    rpc.init(CONF)
    launcher = service.get_launcher()
    server = service.WSGIService('osapi_nova')
    launcher.launch_service(server, workers=server.workers)
    launcher.wait()

api In the from service Layer to get an initiator object , The final will be server Object passed into the initiator object launch_service In the method ,launch_service（server, workers=server.workers） The method is defined as follows ：

class Launcher(object):
    def __init__(self):
        super(Launcher, self).__init__()
        self.launch_service = serve
        self.wait = wait

This method is referenced to serve Method ,serve The method is defined as follows ：

def serve(server, workers=None):
    global _launcher
    if _launcher:
        raise RuntimeError(_('serve() can only be called once'))

    _launcher = service.launch(CONF, server, workers=workers)

Finally called oslo_service/service.py Under the launch Method ,launch The method is defined as follows ：

def launch(conf, service, workers=1, restart_method='reload'):
    …

    if workers is not None and workers <= 0:
        raise ValueError(_("Number of workers should be positive!"))

    if workers is None or workers == 1:
        launcher = ServiceLauncher(conf, restart_method=restart_method)
    else:
        launcher = ProcessLauncher(conf, restart_method=restart_method)
    launcher.launch_service(service, workers=workers)

You can see that there are two kinds of starters used here , In the process of further explaining the startup, first introduce openstack The starter in

Two 、Openstack Medium Launcher

Openstack One of them is called Launcher The concept of , Specifically used to start the service , This class is placed in oslo_service In this bag ,Launcher Divided into two :
One is ServiceLauncher;
The other is ProcessLauncher.
ServiceLauncher A service used to start a single process ;
and ProcessLauncher Used to start multiple worker Child process services , Such as all kinds api service (nova-api、cinder-api) etc.

oslo_service/service.py

1、ServiceLauncher

ServiceLauncher Inherited from Launcher, An important member of the startup service is launcher_service,ServiceLauncher This member of is inherited from Launcher

def launch_service(self, service, workers=1):
 	…
    if workers is not None and workers != 1:
        raise ValueError(_("Launcher asked to start multiple workers"))
    _check_service_base(service)
    service.backdoor_port = self.backdoor_port
    self.services.add(service)

aucher_service Is to add services to self.services Inside the members ,services The type of members is class Services, Look at it. add Method

class Services(object):

    def __init__(self):
        self.services = []
        self.tg = threadgroup.ThreadGroup()
        self.done = event.Event()

    def add(self, service):
        """Add a service to a list and create a thread to run it.

        :param service: service to run
        """
        self.services.append(service)
        self.tg.add_thread(self.run_service, service, self.done)

Services The initialization of this class is very simple , Create a ThreadGroup,ThreadGroup It's actually eventlet Of GreenPool,Openstack utilize eventlet Implement concurrency ,add Method , take self.run_service This method is put into pool in , and service It's its parameters .run_service It's easy , It's called service Of start Method , This completes the start of the service

2、ProcessLauncher

ProcessLauncher Directly inherited from Object, There are also launch_service Method

def launch_service(self, service, workers=1):
 	…
    _check_service_base(service)
    wrap = ServiceWrapper(service, workers)

    LOG.info('Starting %d workers', wrap.workers)
    while self.running and len(wrap.children) < wrap.workers:
        self._start_child(wrap)

lauch_service Except for acceptance service outside , You also need to accept one workers Parameters , That is, the number of child processes , And then call _start_child Start multiple subprocesses

def _start_child(self, wrap):
    if len(wrap.forktimes) > wrap.workers:
        # Limit ourselves to one process a second (over the period of
        # number of workers * 1 second). This will allow workers to
        # start up quickly but ensure we don't fork off children that
        # die instantly too quickly.
        if time.time() - wrap.forktimes[0] < wrap.workers:
            LOG.info('Forking too fast, sleeping')
            time.sleep(1)

        wrap.forktimes.pop(0)

    wrap.forktimes.append(time.time())

    pid = os.fork()
    if pid == 0:
        self.launcher = self._child_process(wrap.service)
        while True:
            self._child_process_handle_signal()
            status, signo = self._child_wait_for_exit_or_signal(
                self.launcher)
            if not _is_sighup_and_daemon(signo):
                self.launcher.wait()
                break
            self.launcher.restart()

        os._exit(status)

    LOG.debug('Started child %d', pid)

    wrap.children.add(pid)
    self.children[pid] = wrap

See the familiar fork No, , Just a simple call os.fork(), Then the child process starts running , Subprocess call _child_process

def _child_process(self, service):
    self._child_process_handle_signal()

    # Reopen the eventlet hub to make sure we don't share an epoll
    # fd with parent and/or siblings, which would be bad
    eventlet.hubs.use_hub()

    # Close write to ensure only parent has it open
    os.close(self.writepipe)
    # Create greenthread to watch for parent to close pipe
    eventlet.spawn_n(self._pipe_watcher)

    # Reseed random number generator
    random.seed()

    launcher = Launcher(self.conf, restart_method=self.restart_method)
    launcher.launch_service(service)
    return launcher

_child_process It's very simple , Create a Launcher, call Laucher.launch_service Method , As mentioned earlier , Actually ServiceLauncher Inherited from Launcher, Also called launcher_service Method , Start the service , Therefore, the next steps can refer to the previous , In the end service.start Method start service

3、 ... and 、WSGIService Start of — To continue

Back to the previous startup section , from launcher Section , We know that the startup of the service finally calls service Of start Method , And here it is service That's where we started api.py Created in the service, Then it is transmitted layer by layer into the starter behind , Let's go back to WSGIService Class start(self) Method

def start(self):
    …
    if self.manager:
        self.manager.init_host()
    self.server.start()
    self.port = self.server.port

Here we call oslo_service/wsgi.py Medium start(self) Method

def start(self):
    …
    self.dup_socket = self.socket.dup()

    if self._use_ssl:
        self.dup_socket = sslutils.wrap(self.conf, self.dup_socket)

    wsgi_kwargs = {
        'func': eventlet.wsgi.server,
        'sock': self.dup_socket,
        'site': self.app,
        'protocol': self._protocol,
        'custom_pool': self._pool,
        'log': self._logger,
        'log_format': self.conf.wsgi_log_format,
        'debug': False,
        'keepalive': self.conf.wsgi_keep_alive,
        'socket_timeout': self.client_socket_timeout
        }

    if self._max_url_len:
        wsgi_kwargs['url_length_limit'] = self._max_url_len

    self._server = eventlet.spawn(**wsgi_kwargs)

Be careful wsgi_kwargs Parameters in func, Its value is eventlet.wsgi.server, stay  eventlet/wsgi.py  Is defined as follows ：

def server(sock, site,
    …
     try:
        serv.log.info("(%s) wsgi starting up on %s" % (
            serv.pid, socket_repr(sock)))
        while is_accepting:
            try:
                client_socket = sock.accept()
                client_socket[0].settimeout(serv.socket_timeout)
                serv.log.debug("(%s) accepted %r" % (
                    serv.pid, client_socket[1]))
                try:
                    pool.spawn_n(serv.process_request, client_socket)
                except AttributeError:
                    warnings.warn("wsgi's pool should be an instance of "
                                  "eventlet.greenpool.GreenPool, is %s. Please convert your"
                                  " call site to use GreenPool instead" % type(pool),
                                  DeprecationWarning, stacklevel=2)
                    pool.execute_async(serv.process_request, client_socket)
            except ACCEPT_EXCEPTIONS as e:
                if support.get_errno(e) not in ACCEPT_ERRNO:
                    raise
            except (KeyboardInterrupt, SystemExit):
                serv.log.info("wsgi exiting")
                break
    finally:
        pool.waitall()
        …

see , Did you see a familiar scene ！sock.accept() Monitor requests , Whenever a new request is received , call pool.spawn_n() Start a process to process the request

Four 、Service Start of

Service Types of processes are also controlled by nova/cmd/* Create some files in the directory ：

• nova-schedule: nova/cmd/schedule.py
• ……
  As a consumer of message oriented middleware , They monitor their own queue, Whenever there is rpc When the request comes , They create a new co process rpc request . With nova-schedule For example , Initialize a at startup Server( from  service.py  Definition ) object .
  Whole Launcher Process follow WSGIServer equally , It's just service Of start() There are just some differences

def start(self):
    …
    target = messaging.Target(topic=self.topic, server=self.host)
    endpoints = [self.manager]
    endpoints.extend(self.manager.additional_endpoints)
    serializer = objects_base.KarborObjectSerializer()
    self.rpcserver = rpc.get_server(target, endpoints, serializer)
    self.rpcserver.start()

Call through , Finally, such a RPCServer object

class RPCServer(msg_server.MessageHandlingServer):
    def __init__(self, transport, target, dispatcher, executor='blocking'):
        super(RPCServer, self).__init__(transport, dispatcher, executor)
        self._target = target

This class inherits from MessageHandlingServer;
notes ：nova All components of depend on oslo.messaging Access the message server , adopt  oslo/messaging/server.py  Initialize a MessageHandlingServer The object of , Listen message queue .
Finally, the service Of start Method

def start(self, override_pool_size=None):
    … 
    if self._started:
        LOG.warning(_LW('Restarting a MessageHandlingServer is inherently '
                        'racy. It is deprecated, and will become a noop '
                        'in a future release of oslo.messaging. If you '
                        'need to restart MessageHandlingServer you should '
                        'instantiate a new object.'))
    self._started = True

    executor_opts = {}

    if self.executor_type in ("threading", "eventlet"):
        executor_opts["max_workers"] = (
            override_pool_size or self.conf.executor_thread_pool_size
        )
    self._work_executor = self._executor_cls(**executor_opts)

    try:
        self.listener = self._create_listener()
    except driver_base.TransportDriverError as ex:
        raise ServerListenError(self.target, ex)

    # HACK(sileht): We temporary pass the executor to the rabbit
    # listener to fix a race with the deprecated blocking executor.
    # We do this hack because this is need only for 'synchronous'
    # executor like blocking. And this one is deprecated. Making
    # driver working in an sync and an async way is complicated
    # and blocking have 0% tests coverage.
    if hasattr(self.listener, '_poll_style_listener'):
        l = self.listener._poll_style_listener
        if hasattr(l, "_message_operations_handler"):
            l._message_operations_handler._executor = (
                self.executor_type)

    self.listener.start(self._on_incoming)

The above object initializes another EventletExecutor( from  oslo/messaging/_executors/impl_eventlet.py) Type of excuete object , It calls  self.listener.poll()  monitor rpc request , Every time a request is received , Create a process to handle the request .

class EventletExecutor(base.ExecutorBase):
    ......

    def start(self):
        if self._thread is not None:
            return

        @excutils.forever_retry_uncaught_exceptions
        def _executor_thread():
            try:
                while True:
                    incoming = self.listener.poll()
                    spawn_with(ctxt=self.dispatcher(incoming),
                               pool=self._greenpool)
            except greenlet.GreenletExit:
                return

        self._thread = eventlet.spawn(_executor_thread)

official account ： Skilled people with culture