[Swift-devel] OSG hang

Michael Wilde wilde at mcs.anl.gov
Thu Jul 14 13:34:08 CDT 2011


make sleep 1234 the app call, and then probe the head node with a ps shell command via globus-job-run to see if the sleep is running.  Should run under the group account that the Engage VO is mapped to (run /usr/bin/id to find this)

In some cases, Globus can start the job but not repond back with completion status (ie if callback ports are not accessible) - hence the two _TCP_ env vars (see globus docs for this).

Im dropping out now and leaving you in Mihael's guidance for this.

- Mike

----- Original Message -----
> How do I check the distinction? The output file is not created, the
> log stops after the GridExec INFO message, and the Swift stdout just
> says "Progress: Submitted:1" several times.
> 
> On Jul 14, 2011, at 1:24 PM, Mihael Hategan wrote:
> 
> > On Thu, 2011-07-14 at 13:14 -0500, Jonathan Monette wrote:
> >> Hello,
> >>   I am trying to submit jobs from an Amazon VM to an OSG site. I
> >> have tried Engage VO sites with a proxy created by both
> >> grid-proxy-init and voms-proxy-init. When running swift it says it
> >> submitted the job but it looks like nothing is executing.
> >
> > Is nothing really executing or is it swift not saying anything was
> > done?
> >
> > The distinction is important because the latter may mean that the
> > service isn't telling swift that stuff was done (this is what I was
> > alluding to yesterday when I said that if GLOBUS_HOSTNAME isn't set
> > properly, lots of things will break).
> >
> >>  This happens when using fork, condor, and pbs. I can successfully
> >> execute the script from communicado to OSG using fork with barely
> >> any
> >> wait time.
> >>
> >>
> >> The files are located in ~jonmon/OSG.0000 on the ci machines.
> >>
> >>
> >> I have filed a bug in bugzilla.
> >> Bug 471 - OSG hangs from VM
> >>
> >>
> >
> >
> > _______________________________________________
> > Swift-devel mailing list
> > Swift-devel at ci.uchicago.edu
> > https://lists.ci.uchicago.edu/cgi-bin/mailman/listinfo/swift-devel
> 
> _______________________________________________
> Swift-devel mailing list
> Swift-devel at ci.uchicago.edu
> https://lists.ci.uchicago.edu/cgi-bin/mailman/listinfo/swift-devel

-- 
Michael Wilde
Computation Institute, University of Chicago
Mathematics and Computer Science Division
Argonne National Laboratory




More information about the Swift-devel mailing list