如何在 celery task.apply_async 中使用优先级

Question

我test在芹菜中有一个队列，我已经为它定义了一个任务：

@celery_app.task(queue='test', ignore_result=True)
def priority_test(priority):
    print(priority)

刚刚打印argument.I要设置priority其定义属性在这里进行appy_async。所以，我写了一个for loop这样的：

for i in range(100):
    priority_test.apply_async((i%10,), queue="test", priority=i%10)

我希望看到这样的结果：

[2017-12-26 17:21:37,309: WARNING/ForkPoolWorker-1] 10
[2017-12-26 17:21:37,311: WARNING/ForkPoolWorker-1] 10
[2017-12-26 17:21:37,314: WARNING/ForkPoolWorker-1] 10
[2017-12-26 17:21:37,317: WARNING/ForkPoolWorker-1] 9
[2017-12-26 17:21:37,319: WARNING/ForkPoolWorker-1] 9
[2017-12-26 17:21:37,321: WARNING/ForkPoolWorker-1] 9
[2017-12-26 17:21:37,323: WARNING/ForkPoolWorker-1] 8
[2017-12-26 17:21:37,326: WARNING/ForkPoolWorker-1] 8
[2017-12-26 17:21:37,329: WARNING/ForkPoolWorker-1] 8
[2017-12-26 17:21:37,332: WARNING/ForkPoolWorker-1] 7
[2017-12-26 17:21:37,334: WARNING/ForkPoolWorker-1] 7
[2017-12-26 17:21:37,336: WARNING/ForkPoolWorker-1] 7
[2017-12-26 17:21:37,341: WARNING/ForkPoolWorker-1] 6
[2017-12-26 17:21:37,344: WARNING/ForkPoolWorker-1] 6
[2017-12-26 17:21:37,346: WARNING/ForkPoolWorker-1] 6
[2017-12-26 17:21:37,349: WARNING/ForkPoolWorker-1] 5
[2017-12-26 17:21:37,351: WARNING/ForkPoolWorker-1] 5
[2017-12-26 17:21:37,353: WARNING/ForkPoolWorker-1] 5
[2017-12-26 17:21:37,355: WARNING/ForkPoolWorker-1] 4
[2017-12-26 17:21:37,358: WARNING/ForkPoolWorker-1] 4
[2017-12-26 17:21:37,360: WARNING/ForkPoolWorker-1] 4

意味着在彼此之后执行相同的优先级，但它以正常方式执行它们：

[2017-12-26 17:21:37,309: WARNING/ForkPoolWorker-1] 10
[2017-12-26 17:21:37,311: WARNING/ForkPoolWorker-1] 9
[2017-12-26 17:21:37,314: WARNING/ForkPoolWorker-1] 8
[2017-12-26 17:21:37,317: WARNING/ForkPoolWorker-1] 7
[2017-12-26 17:21:37,319: WARNING/ForkPoolWorker-1] 6
[2017-12-26 17:21:37,321: WARNING/ForkPoolWorker-1] 5
[2017-12-26 17:21:37,323: WARNING/ForkPoolWorker-1] 4
[2017-12-26 17:21:37,326: WARNING/ForkPoolWorker-1] 3
[2017-12-26 17:21:37,329: WARNING/ForkPoolWorker-1] 2
[2017-12-26 17:21:37,332: WARNING/ForkPoolWorker-1] 1
[2017-12-26 17:21:37,334: WARNING/ForkPoolWorker-1] 10
[2017-12-26 17:21:37,336: WARNING/ForkPoolWorker-1] 9
[2017-12-26 17:21:37,341: WARNING/ForkPoolWorker-1] 8
[2017-12-26 17:21:37,344: WARNING/ForkPoolWorker-1] 7
[2017-12-26 17:21:37,346: WARNING/ForkPoolWorker-1] 6
[2017-12-26 17:21:37,349: WARNING/ForkPoolWorker-1] 5
[2017-12-26 17:21:37,351: WARNING/ForkPoolWorker-1] 4
[2017-12-26 17:21:37,353: WARNING/ForkPoolWorker-1] 3
[2017-12-26 17:21:37,355: WARNING/ForkPoolWorker-1] 2
[2017-12-26 17:21:37,358: WARNING/ForkPoolWorker-1] 1
[2017-12-26 17:21:37,360: WARNING/ForkPoolWorker-1] 10
[2017-12-26 17:21:37,362: WARNING/ForkPoolWorker-1] 9
[2017-12-26 17:21:37,364: WARNING/ForkPoolWorker-1] 8
[2017-12-26 17:21:37,365: WARNING/ForkPoolWorker-1] 7
[2017-12-26 17:21:37,367: WARNING/ForkPoolWorker-1] 6
[2017-12-26 17:21:37,369: WARNING/ForkPoolWorker-1] 5
[2017-12-26 17:21:37,371: WARNING/ForkPoolWorker-1] 4
[2017-12-26 17:21:37,373: WARNING/ForkPoolWorker-1] 3
[2017-12-26 17:21:37,374: WARNING/ForkPoolWorker-1] 2
[2017-12-26 17:21:37,376: WARNING/ForkPoolWorker-1] 1

我应该如何priority在 celery 中使用 rabbitmq 以及priority上面文档中的属性是什么？

Answer 1

为了priority正常工作，您需要正确配置一些设置，并且至少需要 RabbitMQ 3.5.0 版本。

首先将x-max-priority您的队列设置为 10。来自文档：

from kombu import Exchange, Queue

app.conf.task_queues = [
    Queue('tasks', Exchange('tasks'), routing_key='tasks',
          queue_arguments={'x-max-priority': 10},
]

可以使用以下task_queue_max_priority设置为所有队列设置默认值：

app.conf.task_queue_max_priority = 10

然后配置以下设置：

CELERY_ACKS_LATE = True
CELERYD_PREFETCH_MULTIPLIER = 1

默认情况下，预取乘数为 4，在您的情况下，这将导致优先级为 10、9、8 和 7 的前 4 个任务在队列中出现其他任务之前被提取。该CELERY_ACKS_LATE设置将导致任务在执行后被确认。您可以尝试使用此设置来查看您喜欢的行为。