python / python/cpython

File descriptor leak using multiprocessing.Queue

Open
#131,794 12 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

stdlib topic-multiprocessing type-bug
Dominant language
Python
Stars
77.2k
Forks
35.9k
PR merge metrics
PR metrics pending

Description

Bug report

Bug description:

"Too many open files" can occur with multiprocessing.Queue, even if the protocol used to work with the queue is correct.

The file descriptor leak can be demonstrated with two small programs:

File "ko.py"
import os
import multiprocessing

pid = os.getpid()
cmd = f'lsof -p {pid} | grep -i pipe | wc -l'
cmdfull = f'lsof -p {pid} | grep -i pipe'
queues = {}

for i in range(100):
    queues[i] = multiprocessing.Queue()
    os.system(cmd)
    queues[i].close()
    queues[i].join_thread()
    os.system(cmd)

os.system(cmdfull)
File "ok.py"
import os
import multiprocessing

pid = os.getpid()
cmd = f'lsof -p {pid} | grep -i pipe | wc -l'
cmdfull = f'lsof -p {pid} | grep -i pipe'
queues = {}

for i in range(100):
    queues[i] = multiprocessing.Queue()
    queues[i].put('Hello')
    os.system(cmd)
    queues[i].close()
    queues[i].join_thread()
    os.system(cmd)

os.system(cmdfull)

An empty queue that remains empty is not abnormal (e.g., an error queue when no errors occur).

The workaround is to create a class extending multiprocessing.Queue, adding a close method that inserts an extra message into the queue before performing the actual close.

This issue can be observed on Linux and macOS. However, it is not specific to a particular operating system.

CPython versions tested on:

3.12

Operating systems tested on:

Linux

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by running the ko.py and ok.py reproductions on a supported Linux or macOS system and compare their file-descriptor counts. Then inspect the multiprocessing.Queue close and join_thread paths; done means repeatedly closing empty queues no longer leaves the descriptors shown by lsof while preserving the working behavior demonstrated by ok.py.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
operating-systems
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.