multiprocessing hangs when Pool used within Process

[zzzeek]

BPO	5331

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2009-06-30.17:14:56.090>
created_at = <Date 2009-02-20.16:06:50.657>
labels = ['library', 'type-crash']
title = 'multiprocessing hangs when Pool used within Process'
updated_at = <Date 2009-06-30.17:14:56.088>
user = 'https://bugs.python.org/zzzeek'

bugs.python.org fields:

activity = <Date 2009-06-30.17:14:56.088>
actor = 'jnoller'
assignee = 'jnoller'
closed = True
closed_date = <Date 2009-06-30.17:14:56.090>
closer = 'jnoller'
components = ['Library (Lib)']
creation = <Date 2009-02-20.16:06:50.657>
creator = 'zzzeek'
dependencies = []
files = []
hgrepos = []
issue_num = 5331
keywords = []
message_count = 6.0
messages = ['82534', '82540', '89192', '89194', '89901', '89945']
nosy_count = 4.0
nosy_names = ['OG7', 'LambertDW', 'jnoller', 'zzzeek']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = 'crash'
url = 'https://bugs.python.org/issue5331'
versions = ['Python 2.6', 'Python 3.0']

[zzzeek]

this occurs for me running on Mac OSX Leopard. The equivalent code
using "processing" in python 2.5 works fine, which is how I found this
bug - my code hung when upgraded to 2.6. Basically initiating a
multiprocessing.Pool inside of multiprocessing.Process hangs the
application. Below is code which illustrates the issue on both py2.6
and py3.0:

    from multiprocessing import Pool, Process
    import sys

    def f(x):
        return x*x

    def go():
        pool = Pool(processes=4)              
        x = pool.map(f, [1, 2, 3, 4, 5, 6, 7])
        sys.stdout.write("result: %s\n" % x)

    def go2():
        x = map(f, [1,2,3,4,5,6,7])
        sys.stdout.write("result: %s\n" % x)

    # pool alone, fine
    go()

    # process alone, fine
    p = Process(target=go2)
    p.start()
    p.join()

    # use both, hangs
    p = Process(target=go)
    p.start()
    p.join()

[lambertdw]

Fails also on this system.

$ p3
Python 3.0.1 (r301:69556, Feb 13 2009, 23:52:55) 
[GCC 3.4.6 20060404 (Red Hat 3.4.6-3)] on linux2
(...)
>>> # use both hangs
(...)
>>> p.start()
>>> p.join()
(hmm la  do diii   laaa...)^C
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File
"/usr/users/lambert/bin/python/lib/python3.0/multiprocessing/process.py",
line 119, in join
    res = self._popen.wait(timeout)
  File
"/usr/users/lambert/bin/python/lib/python3.0/multiprocessing/forking.py",
line 118, in wait
    return self.poll(0)
  File
"/usr/users/lambert/bin/python/lib/python3.0/multiprocessing/forking.py",
line 107, in poll
    pid, sts = os.waitpid(self.pid, flag)
KeyboardInterrupt
>>>

[OG7]

Apparently the pool workers die all at once, just after the pool creates
them and before the pool is used.
I added a few lines to multiprocessing/pool.py to get the stack and the
exception backtrace.

        except (EOFError, IOError):
            import traceback
            debug(traceback.format_exc())
            debug(''.join(traceback.format_stack()))
            debug('worker got EOFError or IOError -- exiting')
            break

INFO::Rule dispatcher::multiprocessing::child process calling self.run()
DEBUG::Rule dispatcher::multiprocessing::created semlock with handle
3082559488
DEBUG::Rule dispatcher::multiprocessing::created semlock with handle
3082104832
DEBUG::Rule dispatcher::multiprocessing::created semlock with handle
3081826304
DEBUG::Rule dispatcher::multiprocessing::created semlock with handle
3081822208
INFO::PoolWorker-3:1::multiprocessing::child process calling self.run()
DEBUG::PoolWorker-3:1::multiprocessing::Traceback (most recent call last):
  File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/pool.py",
line 57, in worker
    task = get()
  File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/queues.py",
line 339, in get
    return recv()
IOError: [Errno 9] Bad file descriptor

DEBUG::PoolWorker-3:1::multiprocessing:: File
"/home/who/var/co/git-svn/what-base/correlator/bin/correlator", line 30,
in <module>
what.corr.actors.main.main()
File
"/home/who/var/co/git-svn/what-base/correlator/lib/what/corr/actors/main.py",
line 47, in main
rrp.start()
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/process.py",
line 109, in start
self._popen = Popen(self)
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/forking.py",
line 99, in __init__
code = process_obj._bootstrap()
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/process.py",
line 236, in _bootstrap
self.run()
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/process.py",
line 93, in run
self._target(*self._args, **self._kwargs)
File
"/home/who/var/co/git-svn/what-base/correlator/lib/what/corr/actors/rule_dispatcher.py",
line 26, in main
initargs=(manager.agg_msgs_queue, ))
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/init.py",
line 232, in Pool
return Pool(processes, initializer, initargs)
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/pool.py",
line 107, in __init__
w.start()
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/process.py",
line 109, in start
self._popen = Popen(self)
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/forking.py",
line 99, in __init__
code = process_obj._bootstrap()
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/process.py",
line 236, in _bootstrap
self.run()
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/process.py",
line 93, in run
self._target(*self._args, **self._kwargs)
File
"/home/who/.buildout/eggs/multiprocessing-2.6.1.1-py2.6-linux-i686.egg/multiprocessing/pool.py",
line 61, in worker
debug(''.join(traceback.format_stack()))

DEBUG::PoolWorker-3:1::multiprocessing::worker got EOFError or IOError
-- exiting
INFO::PoolWorker-3:1::multiprocessing::process shutting down
DEBUG::PoolWorker-3:1::multiprocessing::running all "atexit" finalizers
with priority >= 0

[OG7]

It seems the root cause is at http://bugs.python.org/issue5155 .
A workaround is to use a duplex Pipe in SimpleQueue.

[jnoller]

Patch attached to bpo-5313, please review

[jnoller]

Committed in r73708 on trunk