[GE users] prolog problem

David Pankhurst david.pankhurst at materials.oxford.ac.uk
Tue Jun 1 14:09:42 BST 2004


Dear SGE users,

I have implemented a prolog script to copy files to local disk and execute
there in order to minimize NFS usage.  I include the script below.  
Under most circumstances it works fine but I've noticed a problem.  
Specifically, when a user kills their job using qdel during the execution
of the prolog script it returns error code 137 (I looked this up and it
means parent ID of reporting process given kill -9 signal).  The problem
is that Grid Engine marks the queue in error state because of this and
hence the queue accepts no further jobs until I reset it.  Has this
problem been reported before and, if not, can anyone think of an elegant
way around this, i.e. some way of making SGE trap 137 errors from
prolog/epilog/starter and exiting cleanly.

Many thanks,

Dave Pankhurst



#!/usr/local/bin/tcsh
if ( $?RUNTMP ) then
   source /usr/local/sge/bin/set_variables
   if ( -es $srcdir/include.dirs ) then
      /usr/local/bin/rsync -az --include-from $srcdir/include.dirs \
                --exclude "*/" $srcdir/ $rundir
   else
      /usr/local/bin/rsync -az --exclude "*/" $srcdir/ $rundir
   endif
   #
   # Print some useful information to stdout
   #
   echo 
"********************************************************************************"
   echo ""
   echo "Setting up to run on local disk..."
   echo ""
   if ( $?runtmpzero ) then
      echo    "Source directory not specified"
      echo -n "Using current working directory: "
   else
      echo -n "Source dir specified: "
   endif
   echo $srcdir
   echo ""
   echo "Remote machine:   " $HOSTNAME
   echo "Working directory:" $rundir
   echo ""
   if ( -es $srcdir/include.dirs ) then
      echo "Included subdirectories from include.dirs file:"
      cat $srcdir/include.dirs
      echo ""
   endif
   echo 
"********************************************************************************"
endif
exit 0


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list