From: Steve Starer (steve.starer@opco.com)
Date: Tue Dec 09 2003 - 09:07:23 EST
All,
I have a 420R with 4CPU and 2 GB of memory, running Solaris 8, MU 7, connected
to a Clariion CX-600 via a pair of Brocade switches. There is only one IO path
to the storage. Veritas Volume Manager manages the storage, with a single
large vxfs file system mounted that hosts pdf files. The box is a member of a
Veritas Cluster Server environment. It acts as an nfs server, and provides
this service to a web based application that can pull down pdf files from a
directory that is an nfs mount on the web server. Administrators and DBAs
get their home directories from the same exported file system. About a week
ago, the developers noticed that it was taking a very long time to get new
pdf files loaded. I recently started looking into the potential issues.
I've checked that the network is not an issue; ping times between the nfs host
and various clients are essentially zero, as they are on the same switch. No
dropped packets either. No collisions. I've run vmstat, and I see no evidence
of a memory shortfall, and the CPU is essentially idle. I run iostat, and I
see that there are bursts of activity between periods of idleness, all while I
am copying files from, for example /usr/local/info to my home directory, which
as mentioned is on the storage in question. I've run nfsstat, and I do not see
any errors, or hi percentage of bad calls, dupchecks, etc.
I've failed the Veritas Service Group over to the standby machine, and tested.
The second unit is configured as close as possible to the first. No change in
performance. I rebooted the first machine, and failed the VCS service group
back without any significant change in performance either.
I checked the cache on the Clariion. It's got High and low watermarks set to
80/60. EMC has suggested a change to 60/40, but I don't believe this is
significant, as the unit has always been configured this way, and there are
other hosts running OLTP and Decision Support database applications with
Sybase and Oracle using storage from this platform without issues.
Database backups use storage from this host without issue.
Absent any other evidence, and I freely admit there may be more exploration to
perform, I feel that this issue is related to nfs, and that we've outgrown our
default environment.
I have never tuned the box for nfs. Perhaps it's time? I am looking for
guidance, especially in what parameters to tune in the kernel. The sun docs
seem to focus on Solaris 6 and 7, and I'm not sure whether the tuning guide
available on line is appropriate for a Solaris 8 environment.
Thanks in advance to all.
regards,
Steve
This communication is for informational purposes only and nothing
herein should be construed as a solicitation, recommendation or
an offer to buy or sell any securities or product. The information
contained herein has been obtained from sources believed to be
reliable but we do not guarantee accuracy or completeness.
Oppenheimer & Co. Inc and its affiliated companies, their officers,
directors and employees may have a position in or, make a market
in any securities mentioned above and, may act as an investment
banker or advisor to such companies. Client account information
or transaction details do not supersede mailed confirmations or
account statements which are the only official records containing
this information. As a matter of policy, orders are not accepted
via e-mail or voice mail and no responsibility shall accrue
relating to any orders placed in this manner. If this communication
has been received in error, please delete or destroy immediately.
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:27:39 EDT