Opened 13 years ago

Last modified 9 years ago

#305 new defect

IZ1918: default_duration is wrongly assumed as job runtime even if queue_conf(5) s_rt/h_rt rlimit is smaller

Reported by: andreas Owned by:
Priority: low Milestone:
Component: sge Version: 6.0
Severity: Keywords: scheduling
Cc:

Description

[Imported from gridengine issuezilla http://gridengine.sunsource.net/issues/show_bug.cgi?id=1918]

        Issue #:      1918             Platform:     All      Reporter: andreas (andreas)
       Component:     gridengine          OS:        All
     Subcomponent:    scheduling       Version:      6.0         CC:    None defined
        Status:       NEW              Priority:     P4
      Resolution:                     Issue type:    DEFECT
                                   Target milestone: ---
      Assigned to:    sgrell (sgrell)
      QA Contact:     andreas
          URL:
       * Summary:     default_duration is wrongly assumed as job runtime even if queue_conf(5) s_rt/h_rt rlimit is smaller
   Status whiteboard:
      Attachments:

     Issue 1918 blocks:
   Votes for issue 1918:


   Opened: Fri Nov 25 03:02:00 -0700 2005 
------------------------


DESCRIPTION:
For jobs without a -l h_rt/s_rt runtime limit scheduler does not know the
duration and thus assumes sched_conf(5) default_duration as runtime. This is
correct. It is problematic however that once the job is put into the reservation
schedule a queue_conf(5) s_rt/h_rt rlimit is not used do adjust job runtime in
case it is smaller than default_duration.

   ------- Additional comments from andreas Fri Nov 25 03:03:25 -0700 2005 -------
The related "job should have finished" logging in schedd messages  file that
hints on such jobs has been disabled for 6.0u7 to prevent mass loggings.

   ------- Additional comments from andreas Wed Dec 20 09:27:44 -0700 2006 -------
SUGGESTED FIX:
For pending jobs with h_rt/s_rt specified the default_duration is valid only, if
no s_rt/h_rt limit was specified globally, host-wide or at queue-level. If
however such a limit is given it is possible and desirable to use effective the
minimum over all queue instances suited for the job.

Due to jobs with the same scheduling category matching with the same
queues/hosts it is possible to compute the h_rt/s_rt limit minimum once per
scheduling category and cache that number for further runs.

Change History (0)

Note: See TracTickets for help on using tickets.