Hi all, I have an (cross platform) NFS problem which is perplexing me, and
am hoping that people could give me some pointers on:

Scenario:

Sol 9 (112234-07) on a dual-proc P3 is the server; sharing filesystems
along these lines:;

share -F nfs -o rw=.stephnet,log /export/files

Only the one client, (Linux, with NFSv3 support inc. over TCP) which
mounts the filesystems rw.

I'm able to read files from the client at wire speed more or less
(server->100Mbit switch<->100Mbit hub<->client) however writes to the
exported filesystem result in a catatonic server after about 10MB is
written. Intermittent pings, TCP sessions collapse (no SSH, http or
anything else for that matter really) and the server appears hung. It's
local but inaccessible so I've not been on the console to see if it's
behaving properly but as far as the network is concerned the machine is
dead. The only way to remedy the situation is to force an umount of the
filesystem on the client. Some seconds later the server reappears again.

I initially thought network issues, but I'm able to write files to the
filesystems again at more or less line speed via scp or to the same
filesystems from Windows clients that access via Samba.

Syslog isn't reporting anything strange on either machine, I've set up NFS
logging on the server but haven't yet had a chance to pour over the logs.
nfsstat on the server doesn't look odd to my eyes, but I'm no expert on
NFS so here it is:

Version 3: (3248724 calls)
null getattr setattr lookup access readlink
9 0% 366278 11% 36 0% 270818 8% 142039 4% 41 0%
read write create mkdir symlink mknod
2414602 74% 25896 0% 163 0% 20 0% 0 0% 0 0%
remove rmdir rename link readdir readdirplus
268 0% 5 0% 10 0% 0 0% 14321 0% 1641 0%
fsstat fsinfo pathconf commit
5836 0% 5768 0% 0 0% 973 0%

Does anyone have any pointers/suggestions, I'm thinking it probably would
be useful for me to be on the console looking at the stats as I try and
cripple the machine, but is there anything obvious I've missed?

Cheers,

Steph

Steph Gosling <stephanos@lineone.net> writes:

>Syslog isn't reporting anything strange on either machine, I've set up NFS
>logging on the server but haven't yet had a chance to pour over the logs.
>nfsstat on the server doesn't look odd to my eyes, but I'm no expert on
>NFS so here it is:


And the system continues to run just fine? (No panic or something)?

Have you installed the NFS patches: 113987-09 114973-02?

You can't logging on the system when this happens?

What type of ethernet hardware does it have?

Casper
--
Expressed in this posting are my opinions. They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.

On Wed, 11 Feb 2004 09:02:35 +0000, Casper H. S. Dik wrote:

> And the system continues to run just fine? (No panic or something)?

Locally I have no idea, it appears back on the network maybe a minute
after the exported fs has been unmounted on the client. It doesn't panic
and appears as if nothing is wrong

> Have you installed the NFS patches: 113987-09 114973-02?

Not unless they're in a patch-cluster. Will have a play and get back to
you.

> You can't logging on the system when this happens?

Like I say, I've not been on the console when this has happened, but I
can't get onto the system ssh wise. At a quite moment I'll get up there
and try doing some large writes. Anything I should look for? oddities in
iostat and netstat?

> What type of ethernet hardware does it have?

Intel EEPro 10/100

> Casper

Thanks for the help,

Steph