我有一个多处理程序,它启动了一组进程。这些程序的run()
部分是这样的:
def run(self):
self.prep()
self._prn('Started with PID {pid}'.format(pid=self.pid))
while True:
job = self._queue.get()
if job is None: break
self.do_work(**job)
self._prn('Received spindown signal.')
(_prn
方法只是简单地添加了Process的名称,如下所示:)
def _prn(self, *args, **kwargs):
print('{nam}({pid}):'.format(nam=self.name, pid=self.pid), end='')
print(*args, **kwargs)
通过将dict推入队列来发送作业参数。
完成后,我将几个None
推入队列,我看到Process-es全部打印出来Received spindown signal.
。
然而,由于莫名其妙的原因,即使我看到一个进程发出了Received spindown signal
消息,他们仍然活着! (也就是说,他们的.is_alive()
方法会一直返回True
)
这是我使用的投票代码:
while len(workers) > 0:
still_alives = []
for w in workers:
if w.is_alive():
still_alives.append(w)
else:
prn(' {nam} trying to join...'.format(nam=w.name))
w.join()
prn(' {nam} joined.'.format(nam=w.name))
prn(' Still alive:', ' '.join((w.name for w in still_alives)))
workers = still_alives
time.sleep(10)
这是我看到的输出的摘录:
Main: Still alive: Querier-0 Querier-2 Querier-3 Querier-4 Querier-6 Querier-7 Querier-9 Querier-10
Querier-9:Received spindown signal.
Querier-4:Received spindown signal.
Main:: Querier-4 trying to join...
Main:: Querier-4 joined.
Main:: Querier-9 trying to join...
Main:: Querier-9 joined.
Main:: Still alive: Querier-0 Querier-2 Querier-3 Querier-6 Querier-7 Querier-10
Querier-3:Received spindown signal.
Querier-10:Received spindown signal.
Querier-6:Received spindown signal.
Main:: Still alive: Querier-0 Querier-2 Querier-3 Querier-6 Querier-7 Querier-10
Main:: Still alive: Querier-0 Querier-2 Querier-3 Querier-6 Querier-7 Querier-10
Querier-0:Received spindown signal.
Main:: Still alive: Querier-0 Querier-2 Querier-3 Querier-6 Querier-7 Querier-10
Main:: Still alive: Querier-0 Querier-2 Querier-3 Querier-6 Querier-7 Querier-10
如您所见,Querier-4
和-9
正常死亡。但Querier-3
,-10
,-6
和-0
从未死亡!
这里发生了什么?