multiprocessing signal defect

[NealBecker]

BPO	10239
Superseder	bpo-9205: Parent process hanging in multiprocessing if children terminate unexpectedly

^{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 2011-05-16.20:36:30.693>
created_at = <Date 2010-10-29.23:16:53.976>
labels = ['type-bug', 'library']
title = 'multiprocessing signal defect'
updated_at = <Date 2011-05-16.20:36:30.691>
user = 'https://bugs.python.org/NealBecker'

bugs.python.org fields:

activity = <Date 2011-05-16.20:36:30.691>
actor = 'neologix'
assignee = 'none'
closed = True
closed_date = <Date 2011-05-16.20:36:30.693>
closer = 'neologix'
components = ['Library (Lib)']
creation = <Date 2010-10-29.23:16:53.976>
creator = 'Neal.Becker'
dependencies = []
files = []
hgrepos = []
issue_num = 10239
keywords = []
message_count = 2.0
messages = ['119948', '136126']
nosy_count = 3.0
nosy_names = ['asksol', 'neologix', 'Neal.Becker']
pr_nums = []
priority = 'normal'
resolution = 'duplicate'
stage = None
status = 'closed'
superseder = '9205'
type = 'behavior'
url = 'https://bugs.python.org/issue10239'
versions = ['Python 2.6']

[NealBecker]

multiprocessing signal defect
From:
Neal Becker <ndbecker2@gmail.com>
Date:
Friday 29 October 2010 8:12:19 am
To:
python-list@python.org
Groups:
gmane.comp.python.general

Tested on:
python-2.6.4-27.fc13.x86_64
linux fedora 13 x86_64

Seems multiprocessing doesn't behave well with signals:

-----------

from multiprocessing import Pool
import time

def sleep (dummy):
    time.sleep (10)
    
if __name__ == '__main__':
    pool = Pool (processes=2)
    result = pool.map (sleep, range (4))
    

---

start it up
$ python test_multip.py 

----------------------
ps auxf | grep python
nbecker   6605  1.6  0.1 338192  6952 pts/1    Sl+  08:03   0:00  |       \_ 
python test_multip.py
nbecker   6606  0.0  0.1 186368  4760 pts/1    S+   08:03   0:00  |           
\_ python test_multip.py
nbecker   6607  0.0  0.1 186372  4740 pts/1    S+   08:03   0:00  |           
\_ python test_multip.py

kill 6607
 ps auxf | grep python
nbecker   6605  0.5  0.1 338192  6952 pts/1    Sl+  08:03   0:00  |       \_ 
python test_multip.py
nbecker   6606  0.0  0.1 186368  4760 pts/1    S+   08:03   0:00  |           
\_ python test_multip.py
nbecker   6607  0.0  0.0      0     0 pts/1    Z+   08:03   0:00  |           
\_ [python] <defunct>

 kill 6606
ps auxf | grep python
nbecker   6605  0.3  0.1 338192  6952 pts/1    Sl+  08:03   0:00  |       \_ 
python test_multip.py
nbecker   6606  0.0  0.0      0     0 pts/1    Z+   08:03   0:00  |           
\_ [python] <defunct>
nbecker   6607  0.0  0.0      0     0 pts/1    Z+   08:03   0:00  |           
\_ [python] <defunct>

Now we have 2 dead children and the parent is hung forever.

Isn't this a serious defect?

[neologix]

Closing as duplicate of issue bpo-9205.