NFS problems from hpux 11.00 to NetApp filler

[Richard Groeneveld]

Hello,

I have a hpux 11.00 machine who haves about 23 nfs mounts.
When i am moving a lot of data on the nfs mounts then the system is generating so much network traffic that everybody get network problems.
Do i need more boids.....

[Mohanasundaram_1]

What kind of network problems? What is the syslog.log indicating?

How about the nfsstat output? How are you measuring/checking network traffic? are you using any tools?

provide us the nfsstat with -c , -m ,-r , -n options.

Cheers,
Mohan.

[Todd Whitcher]

There is a NFS perf. White paper on this site.


http://www.docs.hp.com/hpux/netcom/index.html#NFS%20Services

There are some netapps tuning guides here, basically taken from the Author of the HP White paper. The patches on this page are pretty old, check the itrc for later patches.

http://www.netapp.com/tech_library/3146.html#3.2.

For NFS, you need to check the stats and do some perf tests. Try the dd command examples shown in the White Paper or the Netapps site.

[Dave Olker]

Hello Richard,

There are many factors that could influence your NFS performance. The number of biods is one of them, but there are numerous others, including:

* Number/Type of network interface you're using
* Kernel configuration on the client
* Number/type of CPUs on the client
* Amount of buffer cache on the client
* NFS mount options you're using (i.e. UDP or TCP, rsize, wsize, etc.)
* Applications accessing NFS filesystems
* Response from the NFS server
* NFS server configuration
* Other network activity causing packet drops
* etc.

How certain are you that this is a client problem and not a server problem? When your client experiences the "network problems" are any non-NFS applications affected (i.e. ftp, telnet, rcp, etc.)? When your client sees the problem, are other NFS clients using the same server affected?

As Mohanasundaram suggested, a good place to start would be the nfsstat output. Depending upon what that shows, you might need to collect output from netstat, lanadmin, etc. Also, the Response Center can provide you with a data collection script for the client called "nfsshow", which collects all kinds of useful data from the client's perspective.

Are you also working with Network Appliance on this issue, or are you somehow certain that the server is not at fault? If so, how?

Regards,

Dave