Opened 16 years ago

Last modified 9 years ago

#88 new enhancement

IZ517: Need mechanism to force epilog to run

Reported by: svdavidson Owned by:
Priority: normal Milestone:
Component: sge Version: current
Severity: Keywords: execution
Cc:

Description

[Imported from gridengine issuezilla http://gridengine.sunsource.net/issues/show_bug.cgi?id=517]

        Issue #:      517              Platform:     All           Reporter: svdavidson (svdavidson)
       Component:     gridengine          OS:        All
     Subcomponent:    execution        Version:      current          CC:    None defined
        Status:       NEW              Priority:     P3
      Resolution:                     Issue type:    ENHANCEMENT
                                   Target milestone: ---
      Assigned to:    andreas (andreas)
      QA Contact:     pollinger
          URL:
       * Summary:     Need mechanism to force epilog to run
   Status whiteboard:
      Attachments:

     Issue 517 blocks:
   Votes for issue 517:


   Opened: Tue Apr 1 14:31:00 -0700 2003 
------------------------


A common use of the prolog/epilog is to allocate
and deallocate system resources for a job. Some
examples are IRIX cpusets and the IBM SP switch
table resources. The prolog/epilog is started with
an identical job environment (i.e. parameters,
env, stdout, stderr, ...) as the job itself. If
the job environment is not available (e.g. can't
open stdout file), the prolog/epilog fails to run.
However, because system resources are often being
cleaned up in an epilog, there needs to be a way
to force the epilog to run even if a typical job
environment is not available. For example, if the
stdout of a file is on a NFS file server which is
not available, and the job is killed, we still
need to run the epilog to clean up the resources
so the next job can use those resources.  There
needs to be some mechanism to allow the epilog to
run even if a "perfect" job environment is not
available.  I can see several possibilities for
implementing this. One would be to add an
execd_params (shepherd_params?) attribute called
FORCE_EPILOG and then use this attribute in the
shepherd to ignore certain errors.  Another
implementation might be to allow a flag or the
specification of output file location (e.g.
>/dev/null 2>&1) in the epilog specification in
the queue configuration. I believe this should
also apply to parallel environment start/stop scripts.

   ------- Additional comments from andreas Wed Apr 2 01:42:09 -0700 2003 -------
There is a related issue #368. The solution for this issue
should be "compatible" the solution of #368.

   ------- Additional comments from sgrell Mon Dec 12 03:02:55 -0700 2005 -------
Changed the subcomponent.

Stephan

Change History (0)

Note: See TracTickets for help on using tickets.