Does lpad defuse work on a queued firework running say on an LSF batch manager – will it kill the underlying batch job?
If not, can I get access to the batch job number for a qlaunch’ed firework.
Thanks,
Christine
Does lpad defuse work on a queued firework running say on an LSF batch manager – will it kill the underlying batch job?
If not, can I get access to the batch job number for a qlaunch’ed firework.
Thanks,
Christine
Hi Christine,
Unfortunately, FWS will not kill the underlying batch job. A “defuse” command will only prevent the job from running (being pulled) if it has not already started.
defuse_fws only operates on FWS with the following states: [‘DEFUSED’, ‘WAITING’, ‘READY’, ‘FIZZLED’, ‘PAUSED’]. So, if it is already RUNNING, you would have to:
kill the job yourself on the queue manager. If you need help getting the queue id and you submitted with reservation mode, you can use “lpad get_qids” to help you determine the qid for your job.
the job will now be stuck in a RUNNING state. The two ways to get this back are (i) rerun the fw (“lpad rerun_fws”) followed by defusing the fw if you want it defused (“lpad defuse_fws”) or (ii) use the “lpad detect_lostruns” command to either rerun or mark as FIZZLED any jobs that were abruptly killed (see docs / help for more info).
Best,
Anubhav
On Tuesday, August 8, 2017 at 10:50:50 AM UTC-7, Christine Sweeney wrote:
Does lpad defuse work on a queued firework running say on an LSF batch manager – will it kill the underlying batch job?
If not, can I get access to the batch job number for a qlaunch’ed firework.
Thanks,
Christine