nfs client performance

[David]

On Mon, 5 Mar 2007, Craig White wrote:

> On Mon, 2007-03-05 at 14:36 -0700, David wrote:
>> I'm in need of some help.  I have a rhel4 server mounting an nfs share on
>> a NetApp filer that is having some serious performance issues.  Not being
>> overly familiar with troublshooting nfs, I'm at a wall.  Here are the
>> details:
>>
>> Previous server:
>> Dell 2650, rhel3, kernel 2.4.something
>> - no issues
>>
>> Current server:
>> Dell 2850, rhel4, kernel 2.6.9
>> - gig network, e1000 module
>> - rpciod cpu usage is very high.
>>
>> here are my nfs mount options:
>> filer:/vol/exportvol /local/dir    nfs
>> tcp,noatime,rsize=32768,wsize=32768,nfsvers=3,hard,intr 0 0
> ----
> You don't give us enough details to do anything much more than guess but
> I am curious what is meant by rpciod cpu usage is very high thinking
> that it might have something to do with this issue.
>
> I don't often detail things such as noatime, rsize, wsize, hard (in
> fact, I haven't used hard nfs mounts as I seemed to recall some hanging
> issues.

Some more info (sorry its been awhile.  Y2k2DST seemed to get in the way.

When I mean high cpu usage, while running top (sorted for cpu usage), 
rpciod is often the top process.  Its not uncommon to see it hit > 50% cpu 
usage for many minutes at a time.

some other things I've noticed:

When rpciod hits to over 30% cpu usage for more than a few seconds, the 
servers context switches go from over 15k to around 6k.  Also, the number 
of processes waiting for cpu time increases, although it doesnt get above 
single digits.

That info is from watching top and 'vmstat 2'  at the same time.
The server in question is our Lotus Domino 7 server.  The nfs mounts are 
the domino server's data directories.

Any other comments/questions?

Thanks again,
David

>
> Craig
>
>

-- 
"I find your lack of faith disturbing."
--Darth Vader
---
  13:05:01 up  5:44,  2 users,  load average: 0.02, 0.05, 0.02