IZ2892: need option to populate queue slot counts with execd core counts

[templedf]

[Imported from gridengine issuezilla ​http://gridengine.sunsource.net/issues/show_bug.cgi?id=2892]

        Issue #:      2892             Platform:     All           Reporter: templedf (templedf)
       Component:     gridengine          OS:        All
     Subcomponent:    clients          Version:      6.2u2            CC:    None defined
        Status:       NEW              Priority:     P4
      Resolution:                     Issue type:    ENHANCEMENT
                                   Target milestone: ---
      Assigned to:    roland (roland)
      QA Contact:     roland
          URL:
       * Summary:     need option to populate queue slot counts with execd core counts
   Status whiteboard:
      Attachments:

     Issue 2892 blocks:
   Votes for issue 2892:


   Opened: Tue Jan 27 12:43:00 -0700 2009 
------------------------


There should be an option with qconf and qmon to set the slot count for each
host in a cluster queue to the number of cores on that host.  In other words, it
should be possible to create the default all.q slot settings in any queue at any
time.  One solution is to copy the setting from all.q, but that only works if
all.q is still present and unmodified.  The more general solution is to write a
script like:

hostlist=`qconf -sq $queue | grep hostlist`

for entry in $hostlist; do
   if [ $entry startswith '@' ]; then
      hosts=$hosts `qconf -shgrp $entry | awk '{print $2}'`
   else
      hosts=$hosts $entry
   fi
done

for host in $hosts; do
   slots=`qhost -h $host | awk '(print $3}'`
   qconf -rattr queue slots $slots ${queue}@{$host}
done

(Some of the details are obviously missing from the above script.)

It would be nice, however, to be able to do this without having to write a
complicated script.

   ------- Additional comments from reuti Tue Jan 27 13:03:24 -0700 2009 -------
If you have already a hostgroup, why not attach the number of slots to it? Often you have 2-3-4 types of
machines in the cluster and need just these setting. Having it declared for each host would lead to a long
entry and less comfortable to edit IMO.