how to kill a process's child processes in python

kill_child_processes.py

#!/usr/bin/env python

import multiprocessing
import time
import subprocess, os, signal, sys

def test(s):
    while True:
        print s
        time.sleep(1.5)

def kill_child_processes(signum, frame):
    parent_id = os.getpid()
    ps_command = subprocess.Popen("ps -o pid --ppid %d --noheaders" % parent_id, shell=True, stdout=subprocess.PIPE)
    ps_output = ps_command.stdout.read()
    retcode = ps_command.wait()
    for pid_str in ps_output.strip().split("\n")[:-1]:
        os.kill(int(pid_str), signal.SIGTERM)
    sys.exit()

if __name__ == '__main__':
    pool = multiprocessing.Pool(processes=4)
    result = pool.map_async(test, 'abcd')
    signal.signal(signal.SIGTERM, kill_child_processes)
    result.get(10000)
    pool.close()
    pool.join()

I don't get it, why is the code for something so conceptually simple so complicated?

@brando90 you can use psutil and the following code to create a self-cleaning process:

from contextlib import contextmanager
import subprocess

import psutil


@contextmanager
def process(*args, **kwargs):
    proc = subprocess.Popen(*args, **kwargs)
    try:
        yield proc
    finally:
        for child in psutil.Process(proc.pid).children(recursive=True):
            child.kill()
        proc.kill()

Note that the *args, **kwargs means that process has the same API as subprocess.Popen 😄

Then you can do:

with process(['mycommand', 'myarg', '--myoption', 'myoptionvalue'], stdout=open('myfile.out.log', 'w')) as proc:
    print(proc.pid)
    0/0 # even though this throws an exception, the `process` contextmanager will *still* clean up the process correctly

@pcattori impressive usage of contextmanager and psutil👍

os.killpg(os.getpid(), signal.SIGTERM)

os.killpg(os.getpid(), signal.SIGTERM)

Best Solution! thanks!

@pcattori impressive usage of contextmanager and psutil 👍

I second that, thank you very much for sharing this, Pedro. We just added your code snippet to a little package we are currently gardening, to make it reusable by others, see pyveci/pueblo#39. We hope you don't have any objections about it. It was the last addition amongst a few updates happening last month, so it just become part of pueblo-0.0.4, also available on PyPI.