The stop_remaining_jobs command doesn't seem to work as intended.
Expected behaviour
Stop all running and pending jobs, do not submit new jobs, exit main loop.
Actual behaviour
Running and pending jobs are cancelled but the main loop is not exited, so new jobs will be submitted (at least when running hp_optimization, where new jobs are only submitted after running jobs finished.
Another funny thing I noticed: After calling it, I get a rather high number of failed jobs displayed in the "started execution" progress bar which then goes down over the next few seconds. So also something is wrongly counted there.
Anyway, if the purpose is to stop everything, do we actually need that command? Does it have any advantage over stopping with Ctrl+C?
The
stop_remaining_jobs
command doesn't seem to work as intended.Expected behaviour
Stop all running and pending jobs, do not submit new jobs, exit main loop.
Actual behaviour
Running and pending jobs are cancelled but the main loop is not exited, so new jobs will be submitted (at least when running
hp_optimization
, where new jobs are only submitted after running jobs finished.Another funny thing I noticed: After calling it, I get a rather high number of failed jobs displayed in the "started execution" progress bar which then goes down over the next few seconds. So also something is wrongly counted there.
Anyway, if the purpose is to stop everything, do we actually need that command? Does it have any advantage over stopping with Ctrl+C?
See also discussion here: https://github.com/martius-lab/cluster_utils/pull/114#discussion_r1661881859