> >>> Thanks a lot, since we didnt read the same symptom on the other old CE box,
> >>> as well as new slc4 lcgCE. and thanks input from Jan Astalos that we add
> >>> another snip code into job manager to force process sleep if system load
> >>> over 20. seems this able to stablize the system load and we didnt receive
> >>> the same event right now.
> >> Can you provide that snippet?
> >
> > Jan provide the workaround they have, and force job manager to sleep
> > proportional to the loadavg. this is what i add after 'require
> > $manager_class':
> >
> > open (LOAD, "/proc/loadavg");
> > $loadavg = <LOAD>; $loadavg =~ s/ .*//;
> > close (LOAD);
> > sleep(int(20 * $loadavg));
>
> Welcome, but I have to second the advice to upgrade to lcg-CE 3.1 with
> globus-job-manager-marshal. It can significantly decrease the load on
> your CE.
Thanks Jan, i am finding another free slot to have new slc4 lcgCE 3.1
ready that i can smoohtly migrate the CE box to new r3.1, indeed, we never
find the load on new r31 CE but this could be less job load passing to the
GK to central batch pool. i am adding action for this anyway.
thanks
Br,
J
|