CODEWITHSUNDEEP
pythonmultiprocessingdaemon
linpingta
linpingta

Reputation: 2620

daemonic processes are not allowed to have children

I know it's a common question, and related question like this, but I want to ask for best way to fit my scenario because I haven't used celery now.

My service scenario will use multiprocessing.Process to create multi-campaign order, in each campaign order, it still use multiprocessing.Process to create multi-ad (campaign and ad are 1toM relationship).

As you know, if I set multi-process on both campaign and ad creation part, it will fail with "daemonic processes are not allowed to have children", I think celery may meet similar problem even though I haven't used it now.

My question is, what is the general way to solve this kind of problem? should I use celery still, or any way to work around it?

Thanks a lot

Upvotes: 13

Views: 10773

Answers (1)

Mithril
Mithril

Reputation: 13748

1. General Good Workaround

You should use a message queue to decouple

For example,

  1. main program, create task , push to queue_1

  2. multi-campaign worker get task from queue_1, process, and push some multi-ad task to queue_2

  3. multi-ad worker get task from queue_2 , process , done.

The logic is simple, easy to implement by yourself. Also there are some existing libs for such stuff, such as rq/celery.

2. Simple Workaround

If get AssertionError , use thread instead

def run_in_subprocess(func, *args, **kwargs):
    from multiprocessing import Process
    thread = Process(target=func, args=args, kwargs=kwargs)
    thread.daemon = True
    thread.start()
    return thread

def run_in_thread(func, *args, **kwargs):
    from threading import Thread
    thread = Thread(target=func, args=args, kwargs=kwargs)
    thread.daemon = True
    thread.start()
    return thread

def run_task(config):
    try:
        run_in_subprocess(xxxx_task, config)
    except AssertionError:
        print('daemonic processes are not allowed to have children, use thread')
        run_in_thread(xxxx_task, config)

I use this code in some demo app, but I don't recommend use in production.

Upvotes: 4

Related Questions