2003-03-12 17:07:31

by Brent M. Clements

[permalink] [raw]
Subject: NFS Client turning parameters?

Do you guys have any suggested kernel tuning parameters for a nfs v3
linux client? My server is an hp/ux fileserver.

Thanks,
Brent Clements







-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open!
Get cracking and register here for some mind boggling fun and
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs


2003-03-12 21:53:19

by Brent M. Clements

[permalink] [raw]
Subject: 32k NFS blocksize not working

We have a linux server running 2.4.20 plus it has nfsv3 udp/tcp server
support compiled in.

We have a linux client running 2.4.20 plus it has nfsv3 udp/tcp client
support compiled.


If we specify a read/write size above 8192 on the client, and then mount
the nfs directory. It gives us the following on /proc/mounts


cluster:/usr/local/ /tmp/d nfs
rw,v3,rsize=8192,wsize=8192,hard,udp,lock,addr=cluster 0 0

Do we have to edit something in a kernel include file to allow NFSV3 to
mount above 8k?

Thanks,
Brent





-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open!
Get cracking and register here for some mind boggling fun and
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

2003-03-12 22:22:06

by Dan Brunner

[permalink] [raw]
Subject: Re: 32k NFS blocksize not working

I believe your running UDP.

Try running it over TCP, and see what you get...


Dan


On Wednesday, March 12, 2003, at 03:55 PM, [email protected] wrote:

> We have a linux server running 2.4.20 plus it has nfsv3 udp/tcp server
> support compiled in.
>
> We have a linux client running 2.4.20 plus it has nfsv3 udp/tcp client
> support compiled.
>
>
> If we specify a read/write size above 8192 on the client, and then
> mount
> the nfs directory. It gives us the following on /proc/mounts
>
>
> cluster:/usr/local/ /tmp/d nfs
> rw,v3,rsize=8192,wsize=8192,hard,udp,lock,addr=cluster 0 0
>
> Do we have to edit something in a kernel include file to allow NFSV3 to
> mount above 8k?
>
> Thanks,
> Brent
>
>
>
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by:Crypto Challenge is now open!
> Get cracking and register here for some mind boggling fun and
> the chance of winning an Apple iPod:
> http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
> _______________________________________________
> NFS maillist - [email protected]
> https://lists.sourceforge.net/lists/listinfo/nfs



-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open!
Get cracking and register here for some mind boggling fun and
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

2003-03-12 22:42:07

by Trond Myklebust

[permalink] [raw]
Subject: Re: 32k NFS blocksize not working

>>>>> " " == Brent Clements <[email protected]> writes:

> We have a linux server running 2.4.20 plus it has nfsv3 udp/tcp
> server support compiled in.

> We have a linux client running 2.4.20 plus it has nfsv3 udp/tcp
> client support compiled.


> If we specify a read/write size above 8192 on the client, and
> then mount the nfs directory. It gives us the following on
> /proc/mounts


> cluster:/usr/local/ /tmp/d nfs
> rw,v3,rsize=8192,wsize=8192,hard,udp,lock,addr=cluster 0 0

> Do we have to edit something in a kernel include file to allow
> NFSV3 to mount above 8k?

You'll have to find a server which supports it. Linux 2.4.x knfsd as
it stands, supports only 8k, and so the client will override any
attempt that you make to set either rsize or wsize > 8k.

If you are feeling brave, you might try to change the value of
NFSSVC_MAXBLKSIZE in include/linux/nfsd/const.h and recompiling the
server.

Cheers,
Trond


-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open!
Get cracking and register here for some mind boggling fun and
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

2003-03-17 13:36:21

by barrie_spence

[permalink] [raw]
Subject: RE: NFS Client turning parameters?

Maybe not quite the answer you were looking for, but have you seen the HP-UX tuning paper?

http://h21007.www2.hp.com/dspp/files/unprotected/devresource/Docs/Presentations/NFSperf.pdf

Tuning the HP server could make a big difference all round.

Barrie

-----Original Message-----
From: Brent Clements [mailto:[email protected]]
Sent: 12 March 2003 17:10
To: [email protected]
Subject: [NFS] NFS Client turning parameters?


Do you guys have any suggested kernel tuning parameters for a nfs v3
linux client? My server is an hp/ux fileserver.

Thanks,
Brent Clements







-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open!
Get cracking and register here for some mind boggling fun and
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs


-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open!
Get cracking and register here for some mind boggling fun and
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

2003-03-17 17:08:13

by Tom McNeal

[permalink] [raw]
Subject: RE: NFS Client turning parameters?

Hi -

Dave has expanded that paper into a book, published by Prentice-Hall. Look at
http://www.hp.com/hpbooks/prentice/ptr_0130428167.html

I hadn't included it in the list of references in the FAQ because it was
HP-UX specific, but it still has a lot of good info in it that might be
relevant. I'm not biased or anything, since I was one of the technical
reviewers for it. :-)

Tom

--
Tom McNeal
(650)906-0761(cell)
(650)964-8459(fax)
> Maybe not quite the answer you were looking for, but have you seen the HP-UX
> tuning paper?
>
> http://h21007.www2.hp.com/dspp/files/unprotected/devresource/Docs/Presentations/
> NFSperf.pdf
>
> Tuning the HP server could make a big difference all round.
>
> Barrie
>
> -----Original Message-----
> From: Brent Clements [mailto:[email protected]]
> Sent: 12 March 2003 17:10
> To: [email protected]
> Subject: [NFS] NFS Client turning parameters?
>
>
> Do you guys have any suggested kernel tuning parameters for a nfs v3
> linux client? My server is an hp/ux fileserver.
>
> Thanks,
> Brent Clements
>
>
>
>
>
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by:Crypto Challenge is now open!
> Get cracking and register here for some mind boggling fun and
> the chance of winning an Apple iPod:
> http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
> _______________________________________________
> NFS maillist - [email protected]
> https://lists.sourceforge.net/lists/listinfo/nfs
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by:Crypto Challenge is now open!
> Get cracking and register here for some mind boggling fun and
> the chance of winning an Apple iPod:
> http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
> _______________________________________________
> NFS maillist - [email protected]
> https://lists.sourceforge.net/lists/listinfo/nfs


-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open!
Get cracking and register here for some mind boggling fun and
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs