From: Reiner Dassing (dassing@wettzell.ifag.de)
Date: Thu Apr 07 2005 - 04:23:17 EDT
Hi managers!
The original request was:
----------------------------------------------------------------
After evaluation of the timeout messages of my cluster services I could break
it down to the defragment cron job:
when this jobs defragments one Advfs domain (user_domain)
the cluster services check routines are running into timeout which sometimes causes
a stop and a restart on another member of the cluster.
After deactivation of this defragcron everything went well.
Is there a possibility to run defragcron without hitting the performance of the cluster?
Now the details:
TruCluster V5.1 1885 alpha
#32 21 * * * /bin/cronrun && test -x /usr/sbin/defragcron && /usr/bin/nice -n 20 /usr/sbin/defragcron -T 60 -t 30 -l
/usr/adm/defragcron ... user_domain ...
(/bin/cronrun makes sure that defragcron is only run by exactly one member of the cluster)
-------------------------
Thany you very much to John L. Lanier, Tom Smith, Cris Adams, Martin Ronde Anderson!
The solution they provided is
to use vfast fpr True64 V5.1B or
to make sure that each defragment process runs on the member serving the AdvFS domain.
Tom Smith provided me a ksh script to manage exactly this approach.
Wenn I tested it; it works perfect. But one AdvFS domain seems to be so big that the whole
defragmentation process, used serially, needs all the cpu for about one hour.
The best solution, therefore, is to switch over to Tru64 V5.1B and use vfast.
--
Mit freundlichen Gruessen / With best regards
Reiner Dassing
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:50:17 EDT