捕获键盘中断以阻止 Python 多处理工作人员处理队列

Question

从 stackoverflow 上的几篇文章中，我创建了这段代码。

场景

我想要一个 multiprocessing.queue 让几个工作人员“监听”

在键盘中断的情况下，主进程不应再将新项目放入队列中，并在哨兵对象的帮助下，应优雅地停止工作。

问题

我使用的当前版本的问题

signal.signal(signal.SIGINT, signal.SIG_IGN) 

忽略 Ctrl + C 是它也被主进程忽略。

有什么想法吗？ 我需要使用多处理工作池吗？一些例子表明我可能不得不这样做。那我还能用队列吗？

from multiprocessing import Pool, Process,Queue
import time
import signal
# http://docs.python.org/3.1/library/multiprocessing.html#multiprocessing.Queue
# http://docs.python.org/3.1/library/multiprocessing.html#multiprocessing.Process


class Worker(Process):
    def __init__(self, queue,ident):
        super(Worker, self).__init__()
        # Ignore Signals
        signal.signal(signal.SIGINT, signal.SIG_IGN)
        self.queue= queue
        self.idstr= str(ident)
        print "Ident" + self.idstr

    def run(self):
        print 'Worker started'
        # do some initialization here

        print 'Computing things!'
        for data in iter( self.queue.get, None ):
            print "#" + self.idstr + " : " + str(data)
            time.sleep(5)
            print "#" + self.idstr + "Queue Size: " + str(self.queue.qsize())

        print "Worker Done"

#### Main ####
request_queue = Queue(10)

for i in range(4):
    Worker( request_queue,i ).start()

try:
    for data in range(1000000):
        request_queue.put( data )
        #print "Queue Size: " + str(request_queue.qsize())
        # Sentinel objects to allow clean shutdown: 1 per worker.
    for i in range(4):
        request_queue.put( None ) 

except KeyboardInterrupt:
    print "Caught KeyboardInterrupt, terminating workers"
    while  request_queue.empty()==False:
         request_queue.get()
    request_queue.put( None )    

Answer 1

根据您的解决方案（这很好），我添加了额外的保护层，以防主代码无响应并且用户取消了两次：

global STOP

import os, signal
def signal_handler(sig, frame):
    global STOP
    if STOP:
        signal.signal(signal.SIGINT, signal.SIG_IGN)
        os.kill(os.getpid(), signal.SIGTERM)
    STOP = True
signal.signal(signal.SIGINT, signal_handler)

Answer 2

我想我找到了解决办法。我仍然不喜欢我从 main 获得 SIGINT 1 次，从 Worker 获得 4 次 SIGINT，但也许我必须忍受这种情况。

1. 我为中断信号指定了一个信号处理程序。

2. 收到第一个 Sig INT 后，我会忽略更多 SIG Int 信号

3. 我将停止标志切换为 TRUE

4. 我打破了队列插入循环

5. 我调用 stop 函数来清除队列并插入停止标记

   from multiprocessing import Pool, Process,Queue
   import time
   import signal
   # http://docs.python.org/3.1/library/multiprocessing.html#multiprocessing.Queue
   # http://docs.python.org/3.1/library/multiprocessing.html#multiprocessing.Process
   
   # Stop Flag for loop
   stop = False
   
   # Define SIGINT
   def signal_handler(sig, frame):
       print 'You pressed Ctrl+C!'
       global stop
       stop = True
       # Ignore more Ctrl+C
       signal.signal(signal.SIGINT, signal.SIG_IGN) 
   
   signal.signal(signal.SIGINT, signal_handler)
   
   def stopSentinel(request_queue):
       print "CTRL Stop Queue and insert None"
   
       # Empty Existing Queue
       while  request_queue.empty()==False:
           request_queue.get()
   
       # Put One None for each Worker
       for i in range(4):
           request_queue.put( None ) 
   
   
   class Worker(Process):
       def __init__(self, queue,ident):
           super(Worker, self).__init__()
   
           self.queue= queue
           self.idstr= str(ident)
           print "Ident" + self.idstr
   
       def run(self):
           print 'Worker started'
           # do some initialization here
   
           print 'Computing things!'
           for data in iter( self.queue.get, None ):
               print "#" + self.idstr + " : " + str(data)
               time.sleep(5)
               print "#" + self.idstr + "Queue Size: " + str(self.queue.qsize())
   
           print "Worker Done"
   
   
   
   #### Main #####
   request_queue = Queue(10)
   
   for i in range(4):
       Worker( request_queue, i ).start()
   
   #### Fill Queue with Data ####
   for data in range(1000000):
       request_queue.put( data )
       #print "Queue Size: " + str(request_queue.qsize())
       # Sentinel objects to allow clean shutdown: 1 per worker.
   
       # Check for Stop
       print "Check Breakout"
       if stop == True:
           print "Stop Break"
           break
   
   if stop == True:
       stopSentinel(request_queue)
   else:       
       print "Normal Stop" 
       for i in range(4):
           request_queue.put( None )