celery.apps.worker 源代码

"""Worker command-line program.

This module is the 'program-version' of :mod:`celery.worker`.

It does everything necessary to run that module
as an actual application, like installing signal handlers,
platform tweaks, and so on.
"""
import logging
import os
import platform as _platform
import sys
from datetime import datetime
from functools import partial

from billiard.common import REMAP_SIGTERM
from billiard.process import current_process
from kombu.utils.encoding import safe_str

from celery import VERSION_BANNER, platforms, signals
from celery.app import trace
from celery.loaders.app import AppLoader
from celery.platforms import EX_FAILURE, EX_OK, check_privileges, isatty
from celery.utils import static, term
from celery.utils.debug import cry
from celery.utils.imports import qualname
from celery.utils.log import get_logger, in_sighandler, set_in_sighandler
from celery.utils.text import pluralize
from celery.worker import WorkController

__all__ = ('Worker',)

logger = get_logger(__name__)
is_jython = sys.platform.startswith('java')
is_pypy = hasattr(sys, 'pypy_version_info')

ARTLINES = [
    ' --------------',
    '--- ***** -----',
    '-- ******* ----',
    '- *** --- * ---',
    '- ** ----------',
    '- ** ----------',
    '- ** ----------',
    '- ** ----------',
    '- *** --- * ---',
    '-- ******* ----',
    '--- ***** -----',
    ' --------------',
]

BANNER = """\
{hostname} v{version}

{platform} {timestamp}

[config]
.> app:         {app}
.> transport:   {conninfo}
.> results:     {results}
.> concurrency: {concurrency}
.> task events: {events}

[queues]
{queues}
"""

EXTRA_INFO_FMT = """
[tasks]
{tasks}
"""


def active_thread_count():
    from threading import enumerate
    return sum(1 for t in enumerate()
               if not t.name.startswith('Dummy-'))


def safe_say(msg, f=sys.__stderr__):
    if hasattr(f, 'fileno') and f.fileno() is not None:
        os.write(f.fileno(), f'\n{msg}\n'.encode())


[文档] class Worker(WorkController): """Worker as a program."""
[文档] def on_before_init(self, quiet=False, **kwargs): self.quiet = quiet trace.setup_worker_optimizations(self.app, self.hostname) # this signal can be used to set up configuration for # workers by name. signals.celeryd_init.send( sender=self.hostname, instance=self, conf=self.app.conf, options=kwargs, ) check_privileges(self.app.conf.accept_content)
[文档] def on_after_init(self, purge=False, no_color=None, redirect_stdouts=None, redirect_stdouts_level=None, **kwargs): self.redirect_stdouts = self.app.either( 'worker_redirect_stdouts', redirect_stdouts) self.redirect_stdouts_level = self.app.either( 'worker_redirect_stdouts_level', redirect_stdouts_level) super().setup_defaults(**kwargs) self.purge = purge self.no_color = no_color self._isatty = isatty(sys.stdout) self.colored = self.app.log.colored( self.logfile, enabled=not no_color if no_color is not None else no_color )
[文档] def on_init_blueprint(self): self._custom_logging = self.setup_logging() # apply task execution optimizations # -- This will finalize the app! trace.setup_worker_optimizations(self.app, self.hostname)
[文档] def on_start(self): app = self.app super().on_start() # this signal can be used to, for example, change queues after # the -Q option has been applied. signals.celeryd_after_setup.send( sender=self.hostname, instance=self, conf=app.conf, ) if self.purge: self.purge_messages() if not self.quiet: self.emit_banner() self.set_process_status('-active-') self.install_platform_tweaks(self) if not self._custom_logging and self.redirect_stdouts: app.log.redirect_stdouts(self.redirect_stdouts_level) # TODO: Remove the following code in Celery 6.0 # This qualifies as a hack for issue #6366. warn_deprecated = True config_source = app._config_source if isinstance(config_source, str): # Don't raise the warning when the settings originate from # django.conf:settings warn_deprecated = config_source.lower() not in [ 'django.conf:settings', ] if warn_deprecated: if app.conf.maybe_warn_deprecated_settings(): logger.warning( "Please run `celery upgrade settings path/to/settings.py` " "to avoid these warnings and to allow a smoother upgrade " "to Celery 6.0." )
[文档] def emit_banner(self): # Dump configuration to screen so we have some basic information # for when users sends bug reports. use_image = term.supports_images() if use_image: print(term.imgcat(static.logo())) print(safe_str(''.join([ str(self.colored.cyan( ' \n', self.startup_info(artlines=not use_image))), str(self.colored.reset(self.extra_info() or '')), ])), file=sys.__stdout__, flush=True)
[文档] def on_consumer_ready(self, consumer): signals.worker_ready.send(sender=consumer) logger.info('%s ready.', safe_str(self.hostname))
[文档] def setup_logging(self, colorize=None): if colorize is None and self.no_color is not None: colorize = not self.no_color return self.app.log.setup( self.loglevel, self.logfile, redirect_stdouts=False, colorize=colorize, hostname=self.hostname, )
[文档] def purge_messages(self): with self.app.connection_for_write() as connection: count = self.app.control.purge(connection=connection) if count: # pragma: no cover print(f"purge: Erased {count} {pluralize(count, 'message')} from the queue.\n", flush=True)
[文档] def tasklist(self, include_builtins=True, sep='\n', int_='celery.'): return sep.join( f' . {task}' for task in sorted(self.app.tasks) if (not task.startswith(int_) if not include_builtins else task) )
[文档] def extra_info(self): if self.loglevel is None: return if self.loglevel <= logging.INFO: include_builtins = self.loglevel <= logging.DEBUG tasklist = self.tasklist(include_builtins=include_builtins) return EXTRA_INFO_FMT.format(tasks=tasklist)
[文档] def startup_info(self, artlines=True): app = self.app concurrency = str(self.concurrency) appr = '{}:{:#x}'.format(app.main or '__main__', id(app)) if not isinstance(app.loader, AppLoader): loader = qualname(app.loader) if loader.startswith('celery.loaders'): # pragma: no cover loader = loader[14:] appr += f' ({loader})' if self.autoscale: max, min = self.autoscale concurrency = f'{{min={min}, max={max}}}' pool = self.pool_cls if not isinstance(pool, str): pool = pool.__module__ concurrency += f" ({pool.split('.')[-1]})" events = 'ON' if not self.task_events: events = 'OFF (enable -E to monitor tasks in this worker)' banner = BANNER.format( app=appr, hostname=safe_str(self.hostname), timestamp=datetime.now().replace(microsecond=0), version=VERSION_BANNER, conninfo=self.app.connection().as_uri(), results=self.app.backend.as_uri(), concurrency=concurrency, platform=safe_str(_platform.platform()), events=events, queues=app.amqp.queues.format(indent=0, indent_first=False), ).splitlines() # integrate the ASCII art. if artlines: for i, _ in enumerate(banner): try: banner[i] = ' '.join([ARTLINES[i], banner[i]]) except IndexError: banner[i] = ' ' * 16 + banner[i] return '\n'.join(banner) + '\n'
[文档] def install_platform_tweaks(self, worker): """Install platform specific tweaks and workarounds.""" if self.app.IS_macOS: self.macOS_proxy_detection_workaround() # Install signal handler so SIGHUP restarts the worker. if not self._isatty: # only install HUP handler if detached from terminal, # so closing the terminal window doesn't restart the worker # into the background. if self.app.IS_macOS: # macOS can't exec from a process using threads. # See https://github.com/celery/celery/issues#issue/152 install_HUP_not_supported_handler(worker) else: install_worker_restart_handler(worker) install_worker_term_handler(worker) install_worker_term_hard_handler(worker) install_worker_int_handler(worker) install_cry_handler() install_rdb_handler()
[文档] def macOS_proxy_detection_workaround(self): """See https://github.com/celery/celery/issues#issue/161.""" os.environ.setdefault('celery_dummy_proxy', 'set_by_celeryd')
[文档] def set_process_status(self, info): return platforms.set_mp_process_title( 'celeryd', info=f'{info} ({platforms.strargv(sys.argv)})', hostname=self.hostname, )
def _shutdown_handler(worker: Worker, sig='SIGTERM', how='Warm', callback=None, exitcode=EX_OK, verbose=True): """Install signal handler for warm/cold shutdown. The handler will run from the MainProcess. Args: worker (Worker): The worker that received the signal. sig (str, optional): The signal that was received. Defaults to 'TERM'. how (str, optional): The type of shutdown to perform. Defaults to 'Warm'. callback (Callable, optional): Signal handler. Defaults to None. exitcode (int, optional): The exit code to use. Defaults to EX_OK. verbose (bool, optional): Whether to print the type of shutdown. Defaults to True. """ def _handle_request(*args): with in_sighandler(): from celery.worker import state if current_process()._name == 'MainProcess': if callback: callback(worker) if verbose: safe_say(f'worker: {how} shutdown (MainProcess)', sys.__stdout__) signals.worker_shutting_down.send( sender=worker.hostname, sig=sig, how=how, exitcode=exitcode, ) setattr(state, {'Warm': 'should_stop', 'Cold': 'should_terminate'}[how], exitcode) _handle_request.__name__ = str(f'worker_{how}') platforms.signals[sig] = _handle_request def on_hard_shutdown(worker: Worker): """Signal handler for hard shutdown. The handler will terminate the worker immediately by force using the exit code ``EX_FAILURE``. In practice, you should never get here, as the standard shutdown process should be enough. This handler is only for the worst-case scenario, where the worker is stuck and cannot be terminated gracefully (e.g., spamming the Ctrl+C in the terminal to force the worker to terminate). Args: worker (Worker): The worker that received the signal. Raises: WorkerTerminate: This exception will be raised in the MainProcess to terminate the worker immediately. """ from celery.exceptions import WorkerTerminate raise WorkerTerminate(EX_FAILURE) def during_soft_shutdown(worker: Worker): """This signal handler is called when the worker is in the middle of the soft shutdown process. When the worker is in the soft shutdown process, it is waiting for tasks to finish. If the worker receives a SIGINT (Ctrl+C) or SIGQUIT signal (or possibly SIGTERM if REMAP_SIGTERM is set to "SIGQUIT"), the handler will cancels all unacked requests to allow the worker to terminate gracefully and replace the signal handler for SIGINT and SIGQUIT with the hard shutdown handler ``on_hard_shutdown`` to terminate the worker immediately by force next time the signal is received. It will give the worker once last chance to gracefully terminate (the cold shutdown), after canceling all unacked requests, before using the hard shutdown handler to terminate the worker forcefully. Args: worker (Worker): The worker that received the signal. """ # Replace the signal handler for SIGINT (Ctrl+C) and SIGQUIT (and possibly SIGTERM) # with the hard shutdown handler to terminate the worker immediately by force install_worker_term_hard_handler(worker, sig='SIGINT', callback=on_hard_shutdown, verbose=False) install_worker_term_hard_handler(worker, sig='SIGQUIT', callback=on_hard_shutdown) # Cancel all unacked requests and allow the worker to terminate naturally worker.consumer.cancel_all_unacked_requests() # We get here if the worker was in the middle of the soft (cold) shutdown process, # and the matching signal was received. This can typically happen when the worker is # waiting for tasks to finish, and the user decides to still cancel the running tasks. # We give the worker the last chance to gracefully terminate by letting the soft shutdown # waiting time to finish, which is running in the MainProcess from the previous signal handler call. safe_say('Waiting gracefully for cold shutdown to complete...', sys.__stdout__) def on_cold_shutdown(worker: Worker): """Signal handler for cold shutdown. Registered for SIGQUIT and SIGINT (Ctrl+C) signals. If REMAP_SIGTERM is set to "SIGQUIT", this handler will also be registered for SIGTERM. This handler will initiate the cold (and soft if enabled) shutdown procesdure for the worker. Worker running with N tasks: - SIGTERM: -The worker will initiate the warm shutdown process until all tasks are finished. Additional. SIGTERM signals will be ignored. SIGQUIT will transition to the cold shutdown process described below. - SIGQUIT: - The worker will initiate the cold shutdown process. - If the soft shutdown is enabled, the worker will wait for the tasks to finish up to the soft shutdown timeout (practically having a limited warm shutdown just before the cold shutdown). - Cancel all tasks (from the MainProcess) and allow the worker to complete the cold shutdown process gracefully. Caveats: - SIGINT (Ctrl+C) signal is defined to replace itself with the cold shutdown (SIGQUIT) after first use, and to emit a message to the user to hit Ctrl+C again to initiate the cold shutdown process. But, most important, it will also be caught in WorkController.start() to initiate the warm shutdown process. - SIGTERM will also be handled in WorkController.start() to initiate the warm shutdown process (the same). - If REMAP_SIGTERM is set to "SIGQUIT", the SIGTERM signal will be remapped to SIGQUIT, and the cold shutdown process will be initiated instead of the warm shutdown process using SIGTERM. - If SIGQUIT is received (also via SIGINT) during the cold/soft shutdown process, the handler will cancel all unacked requests but still wait for the soft shutdown process to finish before terminating the worker gracefully. The next time the signal is received though, the worker will terminate immediately by force. So, the purpose of this handler is to allow waiting for the soft shutdown timeout, then cancel all tasks from the MainProcess and let the WorkController.terminate() to terminate the worker naturally. If the soft shutdown is disabled, it will immediately cancel all tasks let the cold shutdown finish normally. Args: worker (Worker): The worker that received the signal. """ safe_say('worker: Hitting Ctrl+C again will terminate all running tasks!', sys.__stdout__) # Replace the signal handler for SIGINT (Ctrl+C) and SIGQUIT (and possibly SIGTERM) install_worker_term_hard_handler(worker, sig='SIGINT', callback=during_soft_shutdown) install_worker_term_hard_handler(worker, sig='SIGQUIT', callback=during_soft_shutdown) if REMAP_SIGTERM == "SIGQUIT": install_worker_term_hard_handler(worker, sig='SIGTERM', callback=during_soft_shutdown) # else, SIGTERM will print the _shutdown_handler's message and do nothing, every time it is received.. # Initiate soft shutdown process (if enabled and tasks are running) worker.wait_for_soft_shutdown() # Cancel all unacked requests and allow the worker to terminate naturally worker.consumer.cancel_all_unacked_requests() # Stop the pool to allow successful tasks call on_success() worker.consumer.pool.stop() # Allow SIGTERM to be remapped to SIGQUIT to initiate cold shutdown instead of warm shutdown using SIGTERM if REMAP_SIGTERM == "SIGQUIT": install_worker_term_handler = partial( _shutdown_handler, sig='SIGTERM', how='Cold', callback=on_cold_shutdown, exitcode=EX_FAILURE, ) else: install_worker_term_handler = partial( _shutdown_handler, sig='SIGTERM', how='Warm', ) if not is_jython: # pragma: no cover install_worker_term_hard_handler = partial( _shutdown_handler, sig='SIGQUIT', how='Cold', callback=on_cold_shutdown, exitcode=EX_FAILURE, ) else: # pragma: no cover install_worker_term_handler = \ install_worker_term_hard_handler = lambda *a, **kw: None def on_SIGINT(worker): safe_say('worker: Hitting Ctrl+C again will initiate cold shutdown, terminating all running tasks!', sys.__stdout__) install_worker_term_hard_handler(worker, sig='SIGINT', verbose=False) if not is_jython: # pragma: no cover install_worker_int_handler = partial( _shutdown_handler, sig='SIGINT', callback=on_SIGINT, exitcode=EX_FAILURE, ) else: # pragma: no cover def install_worker_int_handler(*args, **kwargs): pass def _reload_current_worker(): platforms.close_open_fds([ sys.__stdin__, sys.__stdout__, sys.__stderr__, ]) os.execv(sys.executable, [sys.executable] + sys.argv) def install_worker_restart_handler(worker, sig='SIGHUP'): def restart_worker_sig_handler(*args): """Signal handler restarting the current python program.""" set_in_sighandler(True) safe_say(f"Restarting celery worker ({' '.join(sys.argv)})", sys.__stdout__) import atexit atexit.register(_reload_current_worker) from celery.worker import state state.should_stop = EX_OK platforms.signals[sig] = restart_worker_sig_handler def install_cry_handler(sig='SIGUSR1'): # PyPy does not have sys._current_frames if is_pypy: # pragma: no cover return def cry_handler(*args): """Signal handler logging the stack-trace of all active threads.""" with in_sighandler(): safe_say(cry()) platforms.signals[sig] = cry_handler def install_rdb_handler(envvar='CELERY_RDBSIG', sig='SIGUSR2'): # pragma: no cover def rdb_handler(*args): """Signal handler setting a rdb breakpoint at the current frame.""" with in_sighandler(): from celery.contrib.rdb import _frame, set_trace # gevent does not pass standard signal handler args frame = args[1] if args else _frame().f_back set_trace(frame) if os.environ.get(envvar): platforms.signals[sig] = rdb_handler def install_HUP_not_supported_handler(worker, sig='SIGHUP'): def warn_on_HUP_handler(signum, frame): with in_sighandler(): safe_say('{sig} not supported: Restarting with {sig} is ' 'unstable on this platform!'.format(sig=sig)) platforms.signals[sig] = warn_on_HUP_handler