Re: [HTCondor-users] [CondorLIGO] Getting job PID from condor

[Todd Tannenbaum <tannenba@xxxxxxxxxxx>]

Hi Duncan,

Besides Peter's idea, you can ask the condor_startd to publish an 
attribute "JobPid" into the slot ad by putting in the config of your 
execute machine :

  STARTD_JOB_ATTRS = $(STARTD_JOB_ATTRS) JobPid

and then of course a condor_reconfig.

Note that by default the startd only advertises when the state of the 
slot changes, or every UPDATE_INTERVAL seconds (which defaults to 5 
minutes).  So it may take 5 minutes before you see the JobPid attribute 
appear in your slot ad; if that matters you could do a condor_status 
-direct <slot-name> to see the pid immediately, and/or lower 
UPDATE_INTERVAL (although that will result in more traffic going to your 
collector... maybe not a big deal but depends on the size of your pool...).

Maybe we should add JobPid (and several other useful bits of info that 
the startd receives from the starter) to SYSTEM_STARTD_JOB_ATTRS by 
default?  Besides the pid, the starter is monitoring things like 
resident set size of the job, number of cores being utilized...

p.s. general interest HTCondor questions may benefit by going to 
htcondor-users email list... :)

regards,
Todd

On 2/17/2017 8:30 AM, Peter Francis Couvares wrote:
> Duncan,
>
> I donât believe Condor transmits the pid of your back to the submit
> machine by default (and note that there are often many pids associated
> with a job; but I assume you mean the âparent" job pid initially forked
> by the condor_starter on the execute machine).
>
> However, one good way to dramatically streamline your current technique
> might be to use condor_ssh_to_job and print the _CONDOR_JOB_PIDS
> environment variable.  I.e.:
>
> % condor_ssh_to_job <jobid> 'echo _CONDOR_JOB_PIDSâ
>
> -Peter
>
>
>
> > On Jan 25, 2017, at 9:29 PM, Duncan Meacher <duncan.meacher@xxxxxxx
> > <mailto:duncan.meacher@xxxxxxx>> wrote:
> >
> > Hi,
> >
> > I was just wondering if there is any way of easily obtaining the PID
> > of a job running on a compute node. I've looked through the
> > documentation as well as the output of condor_q -long but haven't been
> > able to see anything. At the moment I'm having to do a rather
> > complicated process of getting the condor ID from condor_q, searching
> > for the ip address within the dags nodes.log file, then ssh'ing into
> > that node to obtain it.
> >
> > Thanks, Duncan
> >
> > --
> > ==========================
> >
> > Duncan Meacher, PhD
> > Postdoctoral Researcher
> > Institute for Gravitation and the Cosmos
> > Department of Physics
> > Pennsylvania State University
> > 104 Davey Lab #040
> > University Park, PA 16802
> > Tel: +1 814 865 3243
> > ==========================
> > _______________________________________________
> > Condorligo mailing list
> > Condorligo@xxxxxxxxxx <mailto:Condorligo@xxxxxxxxxx>
> > http://lists.aei.mpg.de/cgi-bin/mailman/listinfo/condorligo
>
> --
> Peter F. Couvares
> LIGO Laboratory / Caltech
> peter.couvares@xxxxxxxx <mailto:peter.couvares@xxxxxxxx>
>
>
>
> _______________________________________________
> Condorligo mailing list
> Condorligo@xxxxxxxxxx
> http://lists.aei.mpg.de/cgi-bin/mailman/listinfo/condorligo


--
Todd Tannenbaum <tannenba@xxxxxxxxxxx> University of Wisconsin-Madison
Center for High Throughput Computing   Department of Computer Sciences
HTCondor Technical Lead                1210 W. Dayton St. Rm #4257
Phone: (608) 263-7132                  Madison, WI 53706-1685