[Swift-devel] [Bug 323] New: Report status for, and document, how to get Swift to submit jobs meeting a site's scheduler constraints

Tue Apr 5 11:51:16 CDT 2011

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=323

           Summary: Report status for, and document, how to get Swift to
                    submit jobs meeting a site's scheduler constraints
           Product: Swift
           Version: 1.0
          Platform: All
        OS/Version: All
            Status: ASSIGNED
          Severity: normal
          Priority: P3
         Component: Documentation
        AssignedTo: ketan at mcs.anl.gov
        ReportedBy: wilde at mcs.anl.gov

A user guide section on "How do I get Swift to submit jobs into a specific
queue, reservation, etc" would be useful.

This is the main reason that people initially deviate from coaster defaults
into more complex pool entries: to force jobs to fit into some site-imposed
constraint.

Ketan's experience reflected in the thread below, of not knowing why his job
didnt run, illustrates the need for both enhanced reporting and status from
Swift as well as documentation.

----- Original Message -----
> Ketan,
> 
> No, your job 64063 is not eligible to run in the development
> reservation. You requested 5 nodes (mppwidth = 120), and the
> development reservation is limited to 3.
> 
> If your job is less than 1 hour, and also less than 3 nodes (72 cores)
> than your job will run in this reservation.
> 
> Jason Hedden
> 
> On Apr 5, 2011, at 10:54 AM, Ketan Maheshwari wrote:
> 
> > Jason,
> >
> > Is it possible to verify if my request (walltime is 49 minutes) went
> > to this "Development Reservation" queue.
> >
> > Ketan
> >
> > On Mar 23, 2011, at 10:26 AM, Jason Hedden wrote:
> >
> >> I've enabled the development reservation on Beagle. Monday through
> >> Friday between 8:00AM and 5:00PM 3 nodes are reserved for jobs that
> >> request less than 1 hour walltime. Similarly to PADS, no additional
> >> job parameters are required, any job that fits these requirements
> >> will run in the reservation.
> >>
> >> Jason Hedden
> >>
> >> On Mar 22, 2011, at 4:26 PM, Daniel S. Katz wrote:
> >>
> >>> I agree, we definitely want at least some small jobs to be able to
> >>> run quickly most of the time, at least during business hours on
> >>> weekdays.
> >>>
> >>> On Mar 23, 2011, at 2:29 AM, Lorenzo Pesce wrote:
> >>>
> >>>> Seems reasonable and useful to me.
> >>>>
> >>>> On Mar 22, 2011, at 1:12 PM, Ti Leggett wrote:
> >>>>
> >>>>> -----BEGIN PGP SIGNED MESSAGE-----
> >>>>> Hash: SHA1
> >>>>>
> >>>>> I think that decision will be based on what kinds of jobs we
> >>>>> want to favor, so I imagine in a few months when we convene to
> >>>>> make the policy decision.
> >>>>>
> >>>>> On Mar 22, 2011, at 1:11 PM, Michael Wilde wrote:
> >>>>>
> >>>>>> No objection: it sounds like a good idea.
> >>>>>>
> >>>>>> I can't recall, but are we also going to have a development or
> >>>>>> fast queue, or is the scheduler just going to treat requests as
> >>>>>> if smaller requests get higher priority?
> >>>>>>
> >>>>>> - Mike
> >>>>>>
> >>>>>> ----- Original Message -----
> >>>>>>> -----BEGIN PGP SIGNED MESSAGE-----
> >>>>>>> Hash: SHA1
> >>>>>>>
> >>>>>>> This was just brought up by a user. There's currently no easy
> >>>>>>> way for
> >>>>>>> a user to test out their code, especially an MPI code other
> >>>>>>> than to
> >>>>>>> submit an interactive job and wait for it to finally run. What
> >>>>>>> we do
> >>>>>>> on other clusters is have a development reservation during
> >>>>>>> "business"
> >>>>>>> hours, say 8:00am to 5:00pm for jobs an hour or less. Anyone
> >>>>>>> object to
> >>>>>>> putting aside 3 nodes for these types of jobs for people to be
> >>>>>>> able to
> >>>>>>> test code quickly?
> >>>>>>> -----BEGIN PGP SIGNATURE-----
> >>>>>>> Version: GnuPG/MacGPG2 v2.0.14 (Darwin)
> >>>>>>>
> >>>>>>> iEYEARECAAYFAk2I5W0ACgkQ4RgdOxQVi0DkHQCfSCyc/hpAJ7prmphcK64EsrC9
> >>>>>>> 5y8AoJs1EsE5nSRHt22p9bkDlHalZsvL
> >>>>>>> =7e74
> >>>>>>> -----END PGP SIGNATURE-----
> >>>>>>> _______________________________________________
> >>>>>>> beagle-planning mailing list
> >>>>>>> beagle-planning at ci.uchicago.edu
> >>>>>>> http://mail.ci.uchicago.edu/mailman/listinfo/beagle-planning
> >>>>>>
> >>>>>> --
> >>>>>> Michael Wilde
> >>>>>> Computation Institute, University of Chicago
> >>>>>> Mathematics and Computer Science Division
> >>>>>> Argonne National Laboratory
> >>>>>>
> >>>>>
> >>>>> -----BEGIN PGP SIGNATURE-----
> >>>>> Version: GnuPG/MacGPG2 v2.0.14 (Darwin)
> >>>>>
> >>>>> iEYEARECAAYFAk2I5poACgkQ4RgdOxQVi0BsswCggHcK5CgdUjazD0eoBaWE7nAY
> >>>>> 6R8An1Wcz9e+C8pA5w9wpyJdaJ3hGg79
> >>>>> =YP0B
> >>>>> -----END PGP SIGNATURE-----
> >>>>> _______________________________________________
> >>>>> beagle-planning mailing list
> >>>>> beagle-planning at ci.uchicago.edu
> >>>>> http://mail.ci.uchicago.edu/mailman/listinfo/beagle-planning
> >>>>
> >>>> _______________________________________________
> >>>> beagle-planning mailing list
> >>>> beagle-planning at ci.uchicago.edu
> >>>> http://mail.ci.uchicago.edu/mailman/listinfo/beagle-planning
> >>>
> >>> --
> >>> Daniel S. Katz
> >>> University of Chicago
> >>> (773) 834-7186 (voice)
> >>> (773) 834-3700 (fax)
> >>> d.katz at ieee.org or dsk at ci.uchicago.edu
> >>> http://www.ci.uchicago.edu/~dsk/
> >>>
> >>>
> >>>
> >>> _______________________________________________
> >>> beagle-planning mailing list
> >>> beagle-planning at ci.uchicago.edu
> >>> http://mail.ci.uchicago.edu/mailman/listinfo/beagle-planning
> >>
> >> _______________________________________________
> >> beagle-planning mailing list
> >> beagle-planning at ci.uchicago.edu
> >> http://mail.ci.uchicago.edu/mailman/listinfo/beagle-planning
> >
> 
> _______________________________________________
> beagle-planning mailing list
> beagle-planning at ci.uchicago.edu
> http://mail.ci.uchicago.edu/mailman/listinfo/beagle-planning

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

-- 
Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the reporter.