hang and/or leaked processes with multiprocessing.Pool(...).imap(...)

[asottile]

BPO	35629
Nosy	@pitrou, @vstinner, @asottile, @pablogsal, @remilapeyre
Superseder	bpo-35378: multiprocessing.Pool.imaps iterators do not maintain alive the multiprocessing.Pool objects

^{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 2019-01-04.19:44:32.948>
created_at = <Date 2018-12-31.17:28:33.909>
labels = ['3.7', '3.8', 'type-bug', 'library']
title = 'hang and/or leaked processes with multiprocessing.Pool(...).imap(...)'
updated_at = <Date 2019-01-04.23:02:09.008>
user = 'https://github.com/asottile'

bugs.python.org fields:

activity = <Date 2019-01-04.23:02:09.008>
actor = 'vstinner'
assignee = 'none'
closed = True
closed_date = <Date 2019-01-04.19:44:32.948>
closer = 'pitrou'
components = ['Library (Lib)']
creation = <Date 2018-12-31.17:28:33.909>
creator = 'Anthony Sottile'
dependencies = []
files = []
hgrepos = []
issue_num = 35629
keywords = []
message_count = 7.0
messages = ['332825', '332841', '332899', '332995', '333013', '333014', '333019']
nosy_count = 5.0
nosy_names = ['pitrou', 'vstinner', 'Anthony Sottile', 'pablogsal', 'remi.lapeyre']
pr_nums = []
priority = 'normal'
resolution = 'duplicate'
stage = 'resolved'
status = 'closed'
superseder = '35378'
type = 'behavior'
url = 'https://bugs.python.org/issue35629'
versions = ['Python 3.7', 'Python 3.8']

[asottile]

This simple program causes a hang / leaked processes (easiest to run in an interactive shell):

import multiprocessing
tuple(multiprocessing.Pool(4).imap(print, (1, 2, 3)))

$ python3.6
Python 3.6.7 (default, Oct 22 2018, 11:32:17) 
[GCC 8.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import multiprocessing
>>> tuple(multiprocessing.Pool(4).imap(print, (1, 2, 3)))
1
2
3
<<<hang>>>
^CProcess ForkPoolWorker-1:
Traceback (most recent call last):
Process ForkPoolWorker-2:
Process ForkPoolWorker-3:
Process ForkPoolWorker-4:
  File "/usr/lib/python3.6/multiprocessing/pool.py", line 746, in next
    item = self._items.popleft()
IndexError: pop from an empty deque

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.6/multiprocessing/pool.py", line 750, in next
    self._cond.wait(timeout)
  File "/usr/lib/python3.6/threading.py", line 295, in wait
    waiter.acquire()
KeyboardInterrupt


$ python3.7
Python 3.7.2 (default, Dec 25 2018, 03:50:46) 
[GCC 7.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import multiprocessing
>>> tuple(multiprocessing.Pool(4).imap(print, (1, 2, 3)))
1
2
3
(None, None, None)
>>> 
KeyboardInterrupt
Process ForkPoolWorker-3:
Process ForkPoolWorker-1:
Process ForkPoolWorker-2:
Process ForkPoolWorker-4:
>>> Traceback (most recent call last):
  File "/usr/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
    self.run()
  File "/usr/lib/python3.7/multiprocessing/process.py", line 99, in run
    self._target(*self._args, **self._kwargs)
Traceback (most recent call last):
  File "/usr/lib/python3.7/multiprocessing/pool.py", line 110, in worker
    task = get()
  File "/usr/lib/python3.7/multiprocessing/queues.py", line 351, in get
    with self._rlock:
  File "/usr/lib/python3.7/multiprocessing/synchronize.py", line 95, in __enter__
    return self._semlock.__enter__()
KeyboardInterrupt
  File "/usr/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
    self.run()
  File "/usr/lib/python3.7/multiprocessing/process.py", line 99, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/lib/python3.7/multiprocessing/pool.py", line 110, in worker
    task = get()
  File "/usr/lib/python3.7/multiprocessing/queues.py", line 351, in get
    with self._rlock:
  File "/usr/lib/python3.7/multiprocessing/synchronize.py", line 95, in __enter__
    return self._semlock.__enter__()
KeyboardInterrupt
Traceback (most recent call last):
  File "/usr/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
    self.run()
  File "/usr/lib/python3.7/multiprocessing/process.py", line 99, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/lib/python3.7/multiprocessing/pool.py", line 110, in worker
    task = get()
  File "/usr/lib/python3.7/multiprocessing/queues.py", line 351, in get
    with self._rlock:
  File "/usr/lib/python3.7/multiprocessing/synchronize.py", line 95, in __enter__
    return self._semlock.__enter__()
KeyboardInterrupt
Traceback (most recent call last):
  File "/usr/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
    self.run()
  File "/usr/lib/python3.7/multiprocessing/process.py", line 99, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/lib/python3.7/multiprocessing/pool.py", line 110, in worker
    task = get()
  File "/usr/lib/python3.7/multiprocessing/queues.py", line 352, in get
    res = self._reader.recv_bytes()
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 216, in recv_bytes
    buf = self._recv_bytes(maxlength)
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 407, in _recv_bytes
    buf = self._recv(4)
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 379, in _recv
    chunk = read(handle, remaining)
KeyboardInterrupt

(python3.8 shows the same behaviour as python3.7)
$ ./python --version --version
Python 3.8.0a0 (heads/master:ede0b6fae2, Dec 31 2018, 09:19:17)
[GCC 7.3.0]

python2.7 also has similar behaviour.

I'm told this more reliably hangs on windows, though I don't have windows on hand.

I've "fixed" my code to explicitly open / close the pool:

with contextlib.closing(multiprocessing.Pool(jobs)) as pool:
    tuple(pool.imap(...))

I suspect a refcounting / gc bug

[remilapeyre]

Weirdly enough, it works with iPython:

    $ ipython3
    Python 3.7.1 (default, Nov  6 2018, 18:49:54)
    Type 'copyright', 'credits' or 'license' for more information
    IPython 6.4.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: import multiprocessing

In [2]: tuple(multiprocessing.Pool(4).imap(print, (1, 2, 3)))
...:
1
2
3
Out[2]: (None, None, None)

In [3]:

[remilapeyre]

I believe that this is similar to https://bugs.python.org/issue35378 on which @pablogsal is working.

You were right, the issue steems from a refcount bug. Until the resolution you can avoid the issue by explictly keeping a reference on the pool:

>>> import multiprocessing
>>> d = multiprocessing.Pool(4)
>>> tuple(d.imap(print, (1, 2, 3)))
1
2
3
(None, None, None)
>>>

[pitrou]

Indeed, looks like a duplicate.

[vstinner]

I suggest you to write:

with multiprocessing.Pool(4) as pool: result = tuple(pool.imap(print, (1, 2, 3)))

On Python 3.8, your example will now log a resource warning since you don't close/terminate explicitly the pool.

[asottile]

If you see the bottom of my issue, I've suggested (nearly) the same thing -- though I require python2.x compatibility so I'm using contextlib.closing

[vstinner]

I'm using contextlib.closing

Oh, I missed that: good!