I was trying to run a python script with GNU parallel. Everything seems to work, except for the atexit routine used inside the python script. It seems, after ctrl+c, parallel is killing the python process without giving python a chance to call the registered atexit routine. How to make parallel a bit nicer towards the child processes?
Here is an example to show the behaviour.
test_signal.py:
#!/usr/bin/env python3 import time import sys import atexit def cleanup(): print('cleanup called', flush=True) atexit.register(cleanup) time.sleep(60) print('completed process', sys.argv[1])
Tested with the command:
chmod +x test_signal.py ./test_signal.py 1 # this works as expected when using ^C parallel -j 4 ./test_signal.py {} ::: $(seq 1 12) # this one does not
Advertisement
Answer
GNU Parallel kills jobs as specified in --termseq
. It defaults to:
TERM,200,TERM,100,TERM,50,KILL,25
So SIGTERM, wait 200 ms, SIGTERM, wait 100 ms, SIGTERM, wait 50 ms, SIGKILL, wait 25 ms. If the process dies while waiting, GNU Parallel ignores the rest of the sequence – no need to kill a dead horse.
Change --termseq
to the signals you want sent and the time between them.
CTRL-C sends SIGINT, so this should work:
# Give 1 second to clean up, then kill parallel --termseq INT,1000,KILL,25 ...
Just be aware that you will not see the output unless you use --ungroup
.