Opened 16 years ago

Last modified 11 years ago

#234 new defect

IZ1522: Jobs don't get time to initialize signal handler before notification signals are sent

Reported by: ernst Owned by:
Priority: normal Milestone:
Component: sge Version: 6.0u3
Severity: Keywords: execution


[Imported from gridengine issuezilla]

        Issue #:      1522             Platform:     All      Reporter: ernst (ernst)
       Component:     gridengine          OS:        All
     Subcomponent:    execution        Version:      6.0u3       CC:    None defined
        Status:       NEW              Priority:     P3
      Resolution:                     Issue type:    DEFECT
                                   Target milestone: ---
      Assigned to:    andreas (andreas)
      QA Contact:     pollinger
       * Summary:     Jobs don't get time to initialize signal handler before notification signals are sent
   Status whiteboard:

     Issue 1522 blocks:
   Votes for issue 1522:

   Opened: Wed Mar 23 08:52:00 -0700 2005 

If I use the same load parameter for Load and
Suspend Thresholds as well as the Load Adjustment
and I used -notify in qsub, it looks like what can
happen is that the scheduler will start a job,
raise the load by the Load Adjustment value, see
that this passes the Suspend Threshold and send
the USR1 signal immediately.  However, my script
has not been fully able to get started enough to
be able to ignore the USR1 signal and so it
terminates (USR1 default behavior in Linux).  What
I see is that my job leaves the Pending Jobs list
and immediately shows upon the Finished Jobs list

   ------- Additional comments from ernst Wed Mar 23 08:59:58 -0700 2005 -------
find additional information in #1440

   ------- Additional comments from dangruhn Thu Mar 31 07:00:09 -0700 2005 -------
A good fix for this may be to have USR1 and USR2 start out as ignored when a job
is started.  If the job wants the signals, it can then change signal handling to
catch and handle them.

   ------- Additional comments from sgrell Tue Dec 6 08:16:10 -0700 2005 -------
Changed the Subcomponent.


Change History (0)

Note: See TracTickets for help on using tickets.