我已经写了一个Celery Task类：
myapp.tasks.py
from __future__ import absolute_import, unicode_literals
from .services.celery import app
from .services.command_service import CommandService
from exceptions.exceptions import *
from .models import Command
class CustomTask(app.Task):
def run(self, json_string, method_name, cmd_id: int):
command_obj = Command.objects.get(id=cmd_id) # type: Command
try:
val = eval('CommandService.{}(json_string={})'.format(method_name, json_string))
status, error = 200, None
except Exception as e:
auto_retry = command_obj.auto_retry
if auto_retry and isinstance(e, CustomError):
command_obj.retry_count += 1
command_obj.save()
return self.retry(countdown=CustomTask._backoff(command_obj.retry_count), exc=e)
elif auto_retry and isinstance(e, AnotherCustomError) and command_obj.retry_count == 0:
command_obj.retry_count += 1
command_obj.save()
print("RETRYING NOW FOR DEVICE CONNECTION ERROR. TRANSACTION: {} || IP: {}".format(command_obj.transaction_id,
command_obj.device_ip))
return self.retry(countdown=command_obj.retry_count*2, exc=e)
val = None
status, error = self._find_status_code(e)
return_dict = {"error": error, "status_code": status, "result": val}
return return_dict
@staticmethod
def _backoff(attempts):
return 2 ** attempts
@staticmethod
def _find_status_code(exception):
if isinstance(exception, APIException):
detail = exception.default_detail if exception.detail is None else exception.detail
return exception.status_code, detail
return 500, CustomTask._get_generic_exc_msg(exception)
@staticmethod
def _get_generic_exc_msg(exc: Exception):
s = ""
try:
for msg in exc.args:
s += msg + ". "
except Exception:
s = str(exc)
return s
CustomTask = app.register_task(CustomTask())
Celery App定义：
from __future__ import absolute_import, unicode_literals
import os
from celery import Celery, Task
from django.conf import settings
# set the default Django settings module for the 'celery' program.
os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'myapp.settings')
_celery_broker = settings.CELERY_BROKER <-- my broker is amqp://username:password@localhost:5672/myhost
app = Celery('myapp', broker=_celery_broker, backend='rpc://', include=['myapp.tasks', 'myapp.controllers'])
app.config_from_object('django.conf:settings', namespace='CELERY')
app.autodiscover_tasks(['myapp'])
app.conf.update(
result_expires=4800,
task_acks_late=True
)
我的 init .py本教程推荐：
from .celery import app as celery_app
__all__ = ['celery_app']
正在运行任务的控制器：
from __future__ import absolute_import, unicode_literals
from .services.log_service import LogRunner
from myapp.services.command_service import CommandService
from exceptions.exceptions import *
from myapp.services.celery import app
from myapp.services.tasks import MyTask
from .models import Command
class MyController:
def my_method(self, json_string):
<non-async set up stuff here>
cmd_obj = Command.objects.create(<stuff>) # type: Command
task_exec = MyTask.delay(json_string, MyController._method_name, cmd_obj.id)
cmd_obj.task_id = task_exec
try:
return_dict = task_exec.get()
except Exception as e:
self._logger.error("ERROR: IP: {} and transaction: {}. Error Type: {}, "
"Celery Error: {}".format(ip_addr, transaction_id, type(e), e))
status_code, error = self._find_status_code(e)
return_dict = {"error": error, "status_code": status_code, "result": None}
return return_dict
**这是我的问题：**
当我通过一个请求打一个视图，一个接一个地运行该Django控制器时，它可以正常运行。 完美 。
但是，我要访问的外部服务将为2个并发请求抛出一个错误（这是可以预期的-可以）。收到错误后，我将自动重试任务。
这是奇怪的部分
重试后，我在控制器中的.get()对于所有并发请求均停止工作。我的控制器就挂在那里！而且我知道芹菜实际上正在执行任务！这是芹菜运行的日志：
[2018-09-25 19:10:24,932: INFO/MainProcess] Received task: myapp.tasks.MyTask[bafd62b6-7e29-4c39-86ff-fe903d864c4f]
[2018-09-25 19:10:25,710: INFO/MainProcess] Received task: myapp.tasks.MyTask[8d3b4279-0b7e-48cf-b45d-0f1f89e213d4] <-- THIS WILL FAIL BUT THAT IS OK
[2018-09-25 19:10:25,794: ERROR/ForkPoolWorker-1] Could not connect to device with IP <some ip> at all. Retry Later plase
[2018-09-25 19:10:25,798: WARNING/ForkPoolWorker-1] RETRYING NOW FOR DEVICE CONNECTION ERROR. TRANSACTION: b_txn || IP: <some ip>
[2018-09-25 19:10:25,821: INFO/MainProcess] Received task: myapp.tasks.MyTask[8d3b4279-0b7e-48cf-b45d-0f1f89e213d4] ETA:[2018-09-25 19:10:27.799473+00:00]
[2018-09-25 19:10:25,823: INFO/ForkPoolWorker-1] Task myapp.tasks.MyTask[8d3b4279-0b7e-48cf-b45d-0f1f89e213d4] retry: Retry in 2s: AnotherCustomError('Could not connect to IP <some ip> at all.',)
[2018-09-25 19:10:27,400: INFO/ForkPoolWorker-2] executed command some command at IP <some ip>
[2018-09-25 19:10:27,418: INFO/ForkPoolWorker-2] Task myapp.tasks.MyTask[bafd62b6-7e29-4c39-86ff-fe903d864c4f] succeeded in 2.4829552830196917s: {'error': None, 'status_code': 200, 'result': True}
<some command output here from a successful run> **<-- belongs to task bafd62b6-7e29-4c39-86ff-fe903d864c4f**
[2018-09-25 19:10:31,058: INFO/ForkPoolWorker-2] executed some command at IP <some ip>
[2018-09-25 19:10:31,059: INFO/ForkPoolWorker-2] Task command_runner.tasks.MyTask[8d3b4279-0b7e-48cf-b45d-0f1f89e213d4] succeeded in 2.404364461021032s: {'error': None, 'status_code': 200, 'result': True}
<some command output here from a successful run> **<-- belongs to task 8d3b4279-0b7e-48cf-b45d-0f1f89e213d4 which errored and retried itself**
如您所见，该任务确实在芹菜上运行！！只是我在控制器中拥有的.get()无法备份这些结果-无论成功完成的任务如何或错误的任务。
通常，运行并发请求Error: "Received 0x50 while expecting 0xce"时出现错误。 那是什么？ 再次，奇怪的是，当一个接一个的请求而没有Django处理多个传入请求时，所有这些工作正常。虽然，我无法重试单个请求。
答案 0 :(得分：0)
如果使用more than once or after a celery restart，则RPC后端（get所等待的）被设计为失败。
结果只能被检索一次，并且只能由发起任务的客户端检索。两种不同的过程不能等待相同的结果。
默认情况下，消息是瞬态的（非持久性），因此，如果代理重新启动，结果将消失。您可以使用result_persistent设置将结果后端配置为发送持久性消息。
所以看起来正在发生的事情是该异常导致celery停止并中断了与调用控制器的rpc连接。根据您的用例，使用永久结果后端（例如redis或数据库）可能更有意义。