On Fri, 4 Mar 2011, Dave Love wrote:
...
I've been meaning to dust it off and bring it up to date for 6.2 and where
the client spool is local to each compute node (but I arrange for the
messages file to end up in the central location anyway),
What's your recommended recipe if one has enough pressure on the spool
to make the local spooling worthwhile (which we definitely don't)?
* Create a $SGE_ROOT/$SGE_CELL/spool/<HOSTNAME> directory
* Replace the messages file in the local spool directory with a symlink
pointing to $SGE_ROOT/$SGE_CELL/spool/<HOSTNAME>/messages
The intention is to gain many of the performance advantages of a local
spool, without having to change how you debug job failures.
However, I'm not doing this yet as we're hitting a bug (which I keep
meaning to log in a bugtracker and/or try and fix - sound familiar?) where
lots of logging info relating to failed job cleanup is being continually
written - so I'm copying data across to the central spool area at hourly
intervals (with cron).
Mark
--
-----------------------------------------------------------------
Mark Dixon Email : [email protected]
HPC/Grid Systems Support Tel (int): 35429
Information Systems Services Tel (ext): +44(0)113 343 5429
University of Leeds, LS2 9JT, UK
-----------------------------------------------------------------
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
