zulip/zerver/management/commands/process_queue.py

135 lines
5.1 KiB
Python
Raw Normal View History

import logging
import os
import signal
import sys
import threading
from argparse import ArgumentParser
from collections.abc import Iterator
from contextlib import contextmanager
from types import FrameType
from typing import Any
import sentry_sdk
from django.conf import settings
from django.core.management.base import CommandError
from django.utils import autoreload
from typing_extensions import override
from zerver.lib.management import ZulipBaseCommand
from zerver.worker.queue_processors import get_active_worker_queues, get_worker
@contextmanager
def log_and_exit_if_exception(
logger: logging.Logger, queue_name: str, threaded: bool
) -> Iterator[None]:
try:
yield
except Exception:
logger.exception("Unhandled exception from queue: %s", queue_name, stack_info=True)
if threaded:
# Sending SIGUSR1 is the right way to exit - triggering the
# exit_with_three signal handler, causing exit and reload.
os.kill(os.getpid(), signal.SIGUSR1)
else:
sys.exit(1)
class Command(ZulipBaseCommand):
@override
def add_arguments(self, parser: ArgumentParser) -> None:
parser.add_argument("--queue_name", metavar="<queue name>", help="queue to process")
parser.add_argument(
"--worker_num", metavar="<worker number>", type=int, default=0, help="worker label"
)
parser.add_argument("--all", action="store_true", help="run all queues")
parser.add_argument(
"--multi_threaded",
nargs="+",
metavar="<list of queue name>",
required=False,
help="list of queue to process",
)
help = "Runs a queue processing worker"
@override
def handle(self, *args: Any, **options: Any) -> None:
logging.basicConfig()
logger = logging.getLogger("process_queue")
def exit_with_three(signal: int, frame: FrameType | None) -> None:
"""
This process is watched by Django's autoreload, so exiting
with status code 3 will cause this process to restart.
"""
logger.warning("SIGUSR1 received. Restarting this queue processor.")
sys.exit(3)
if not settings.USING_RABBITMQ:
# Make the warning silent when running the tests
if settings.TEST_SUITE:
logger.info("Not using RabbitMQ queue workers in the test suite.")
else:
logger.error("Cannot run a queue processor when USING_RABBITMQ is False!")
raise CommandError
def run_threaded_workers(queues: list[str], logger: logging.Logger) -> None:
cnt = 0
for queue_name in queues:
if not settings.DEVELOPMENT:
logger.info("launching queue worker thread %s", queue_name)
cnt += 1
td = ThreadedWorker(queue_name, logger)
td.start()
assert len(queues) == cnt
logger.info("%d queue worker threads were launched", cnt)
if options["all"]:
signal.signal(signal.SIGUSR1, exit_with_three)
autoreload.run_with_reloader(run_threaded_workers, get_active_worker_queues(), logger)
elif options["multi_threaded"]:
signal.signal(signal.SIGUSR1, exit_with_three)
queues = options["multi_threaded"]
autoreload.run_with_reloader(run_threaded_workers, queues, logger)
else:
queue_name = options["queue_name"]
worker_num = options["worker_num"]
def signal_handler(signal: int, frame: FrameType | None) -> None:
logger.info("Worker %d disconnecting from queue %s", worker_num, queue_name)
worker.stop()
sys.exit(0)
logger.info("Worker %d connecting to queue %s", worker_num, queue_name)
with log_and_exit_if_exception(logger, queue_name, threaded=False):
worker = get_worker(queue_name, worker_num=worker_num)
with sentry_sdk.isolation_scope() as scope:
scope.set_tag("queue_worker", queue_name)
scope.set_tag("worker_num", worker_num)
worker.setup()
signal.signal(signal.SIGTERM, signal_handler)
signal.signal(signal.SIGINT, signal_handler)
signal.signal(signal.SIGUSR1, signal_handler)
worker.start()
class ThreadedWorker(threading.Thread):
def __init__(self, queue_name: str, logger: logging.Logger) -> None:
threading.Thread.__init__(self)
self.logger = logger
self.queue_name = queue_name
@override
def run(self) -> None:
with (
sentry_sdk.isolation_scope() as scope,
log_and_exit_if_exception(self.logger, self.queue_name, threaded=True),
):
scope.set_tag("queue_worker", self.queue_name)
worker = get_worker(self.queue_name, threaded=True)
worker.setup()
logging.debug("starting consuming %s", self.queue_name)
worker.start()