Is there any neater way I can accomplish the task (as in title)? Also, I'm not sure why asyncio.Task.all_tasks() explodes and whether this is bad or not.

from concurrent.futures import ProcessPoolExecutor
import asyncio
from contextlib import contextmanager
import random

def event_loop():
    loop = asyncio.get_event_loop()
    with ProcessPoolExecutor(3) as executor:
        yield loop

async def start_long_proc(delay=0, *, started_event):
    await asyncio.sleep(delay)
    cmd = 'while true; do date >> temp.log; sleep 0.1; done;'
    proc = await asyncio.create_subprocess_shell(cmd)
    return proc

async def periodic_monitor():
    while True:
        n = random.randrange(100)
        cmd = f'echo "{n}"'
        proc = await asyncio.create_subprocess_shell(
            cmd, stdout=asyncio.subprocess.PIPE)
        stdout, stderr = await proc.communicate()
        yield int(stdout.decode().rstrip())

async def f():
    started_event = asyncio.Event()

    # run Bash Job B
    long_proc = await start_long_proc(started_event=started_event)

    p_terminate = 10
    monitor_interval = 1

    async for n in periodic_monitor():  # periodically run Bash Job A
        if n < p_terminate:
            print('terminated and reschedule long proc')
            long_proc.terminate()  # kill Bash Job B
            long_proc = await start_long_proc(5, started_event=started_event)
        await asyncio.sleep(monitor_interval)

if __name__ == '__main__':
    with event_loop() as loop:
  • 1
    \$\begingroup\$ This seems to be an abstraction of your actual problem, no? \$\endgroup\$ – Daniel Dec 29 '17 at 14:54
  • \$\begingroup\$ @Coal_ Yes you're right :). It might not be an XY problem because I really need to run existing bash scripts... \$\endgroup\$ – pterodragon Dec 29 '17 at 15:32
  • \$\begingroup\$ Sorry for the late reply. In that case, this question should probably be considered off-topic anyway. \$\endgroup\$ – Daniel Dec 29 '17 at 22:21
  • \$\begingroup\$ Consider using a standard tool like Monit instead? \$\endgroup\$ – 200_success Dec 30 '17 at 14:58

Is there any neater way I can accomplish the task?

        long_proc.terminate()  # kill Bash Job B

Well, sending SIGTERM is effective, but since you're in control of what Job B runs you might prefer to have it do IPC so it gracefully shuts down after completing a processing phase, rather than being interrupted at some arbitrary point. Could be as simple as creating a sentinel /tmp file that B looks for.

A bigger issue is that you're not doing wait4() child, like this:


Instead you wrote a race with a 5-second delay to encourage the child to win. That long sleep implies wasted idle time. You could run the CPU flat out by synchronizing with exiting child.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.