【问题标题】:python subprocess.Popen blocks in futex_ statefutex_ 状态下的 python subprocess.Popen 块
【发布时间】:2012-07-19 09:54:47
【问题描述】:

我正在尝试实现一些基本调度,以产生最多 n 个进程。一次打开并等待它们执行。为此我已经完成了:

CHECKING_INTERVAL = 10

class StandAloneClient(object):
    """
    Instead of communicating with a backend cluster, fire locally a new thread.
    """
    max_thread_nr = 4
    thread_pool = []
    pending_operations = Queue.Queue(0)

    class OperationExecutor(threading.Thread):

        def run(self):
            """
            Get the required data from the operation queue and launch the operation.
            """
            while True:
                launch_data = StandAloneClient.pending_operations.get()
                if launch_data != None:
                    operation_id = launch_data[0]
                    user_name_label = launch_data[1]
                    LOGGER.debug("Launching operation " + str(operation_id) + " with name " + str(user_name_label))
                    ## Create a new process for the new launched operation
                    oper = ['python', '-m', 'processRunner', str(operation_id), user_name_label]
                    launched_process = Popen(oper, stdout=PIPE, stdin=PIPE, stderr=PIPE)
#                    launched_process.wait()
#                    while launched_process.poll() is None:
#                        sleep(CHECKING_INTERVAL)
#                        LOGGER.debug("Operation id=%s is still running. Going to sleep for %s seconds."%(operation_id,
#                                                                                                         CHECKING_INTERVAL))
                    LOGGER.debug("===========================================================")
                    LOGGER.debug("Finished operation %s succesfully."%(operation_id,))


    def __init__(self):
        """
        If there are still empty spots create a new executor thread.
        """
        for _ in xrange(self.max_thread_nr - len(self.thread_pool)):
            new_executor = StandAloneClient.OperationExecutor()
            self.thread_pool.append(new_executor)
            new_executor.start()

    @staticmethod
    def execute(operation_id, user_name_label="Unknown"):
        """Start asynchronous operation locally"""
        StandAloneClient.pending_operations.put((operation_id, user_name_label))

我通过以下方式向队列添加操作:

StandAloneClient().execute(...)

现在我已经评论了我认为可能由于某种原因阻塞线程的部分。但即便如此,似乎没有一个孩子产生过。执行到最后,我检查了日志,processRunner.py 完成了它需要做的所有事情,直到最后,但是如果我执行ps -el|grep python,我仍然会看到所有进程:

    0 S  1000   755     1  5  80   0 - 548314 poll_s pts/0   00:00:13 python
0 S  1000  1198   755  4  80   0 - 280172 futex_ pts/0   00:00:09 python
0 S  1000  1201   755  4  80   0 - 280176 futex_ pts/0   00:00:09 python
0 S  1000  1206   755  4  80   0 - 280230 futex_ pts/0   00:00:09 python
0 S  1000  1215   755  4  80   0 - 280198 futex_ pts/0   00:00:09 python
0 S  1000  1216   755  4  80   0 - 281669 futex_ pts/0   00:00:09 python
0 S  1000  1221   755  4  80   0 - 280201 futex_ pts/0   00:00:09 python
0 S  1000  1231   755  4  80   0 - 281668 futex_ pts/0   00:00:09 python
0 S  1000  1240   755  4  80   0 - 280229 futex_ pts/0   00:00:09 python
0 S  1000  1257   755  4  80   0 - 280201 futex_ pts/0   00:00:09 python

我正在使用 Python 2.7.2 的 fedora 16 机器上尝试此操作。任何建议>

问候, 波格丹

【问题讨论】:

  • 为什么使用子进程而不是多进程在多个进程中运行Python代码?附注:您的代码对于这样一个简单的任务很复杂

标签: python multithreading subprocess


【解决方案1】:

您的线程不是守护进程。它们在主线程退出后继续存在。他们在pending_operations.get() 被永久屏蔽。

【讨论】:

    猜你喜欢
    • 2012-12-12
    • 2011-07-05
    • 1970-01-01
    • 1970-01-01
    • 1970-01-01
    • 2014-05-28
    • 2010-10-01
    • 2012-03-23
    • 2011-05-28
    相关资源
    最近更新 更多