为什么multiprocessing.Process.join()会挂起?

El *_*uso 4 python multiprocessing python-2.7

我正在以这种方式使用多处理:

import multiprocessing as mp

def worker(thread_id, tasks, results):
    tmp_dir = 'temp_for_{}'.format(thread_id)
    os.makedirs(tmp_dir)
    try:
        while not tasks.empty():
            data = tasks.get()
            response = process_pdf(data, tmp_dir)
            results.put(response)
    except (KeyboardInterrupt, SystemExit):
        log.info('Interrupt signal received in thread %s.', thread_id)
    except Queue.Empty:
        pass
    except Exception:
        log.error("Unexpected error in %s", thread_id, exc_info=True)
    finally:
        shutil.rmtree(tmp_dir)
        log.info("Thread %s exit", thread_id)

if __name__ == "__main__":
    tasks, results = mp.Queue(), mp.Queue()
    for record in cursor.select(query):
        tasks.put(record)
    manager = mp.Manager()
    workers = [mp.Process(target=worker, args=(i, tasks, results)) for i in xrange(8)]
    for worker in workers:
        worker.start()
    try:
        for worker in workers:
            worker.join()
    except (KeyboardInterrupt, SystemExit):
        log.info('Interrupt signal received in main. Cleaning up main')
    finally:
        log.info('Got %s results. Saving', results.qsize())
        while not results.empty():
            cursor.update_one('documents', 'id', results.get())
        cursor.close()
Run Code Online (Sandbox Code Playgroud)

这是我运行此代码时的输出:

14:34:04 15/10 INFO: Thread 6 exit
14:34:04 15/10 INFO: Thread 7 exit
14:34:21 15/10 INFO: Thread 3 exit
14:34:24 15/10 INFO: Thread 2 exit
14:34:24 15/10 INFO: Thread 1 exit
14:34:29 15/10 INFO: Thread 5 exit
14:34:36 15/10 INFO: Thread 0 exit
14:35:37 15/10 INFO: Thread 4 exit
Run Code Online (Sandbox Code Playgroud)

然后我在等待一段时间没有进展后输入^ C,并获得此输出:

^C14:37:16 15/10 INFO: Interrupt signal received in main. Cleaning up main
14:37:16 15/10 INFO: Got 16 results. Saving
Run Code Online (Sandbox Code Playgroud)

我得到了所有线程的回溯:

Process Process-9:
Traceback (most recent call last):
  File "/usr/lib64/python2.7/multiprocessing/process.py", line 261, in _bootstrap
    util._exit_function()
  File "/usr/lib64/python2.7/multiprocessing/util.py", line 328, in _exit_function
    util._exit_function()
  File "/usr/lib64/python2.7/multiprocessing/util.py", line 274, in _run_finalizers
    finalizer()
  File "/usr/lib64/python2.7/multiprocessing/util.py", line 207, in __call__
    res = self._callback(*self._args, **self._kwargs)
  File "/usr/lib64/python2.7/multiprocessing/queues.py", line 218, in _finalize_join
    thread.join()
  File "/usr/lib64/python2.7/threading.py", line 952, in join
    thread.join()
  File "/usr/lib64/python2.7/threading.py", line 340, in wait
    waiter.acquire()
KeyboardInterrupt
Run Code Online (Sandbox Code Playgroud)

这为什么挂?如果它很重要,我可以添加process_pdf()运行一些子进程subprocess.Popen().

El *_*uso 6

非常感谢dano的暗示.修复此问题是使用Manager()以下命令创建队列:

manager = mp.Manager()
tasks, results = manager.Queue(), manager.Queue()
Run Code Online (Sandbox Code Playgroud)


将Tnx 编辑为ShadowRanger.看起来像2.7.10的调度中的例外,现在我们可以使用multiprocessing.Poolimap_unordered和不需要为简单的工作写代码墙 :)但我还没试过