Often experiencing nfs server foo:/bar: not responding

[Steve Kargl]

On Tue, Jun 20, 2006 at 03:31:15PM -0700, Mohan Srinivasan wrote:
> 
> --- "Matthew D. Fuller" <fullermd at over-yonder.net> wrote:
> > On Tue, Jun 20, 2006 at 01:45:07PM -0700 I heard the voice of
> > David O'Brien, and lo! it spake thus:
> > > 
> > > Note that mounting NFS /home with -T "fixes" the problem.
> > 
> > Hm, that's a point.  I have an early-June -CURRENT as a NFS client on
> > a late-Nov.  -CURRENT NFS server, and haven't had any troubles.  But I
> > do use TCP mounts.
> > 
> 
> I am not able to reproduce the problems you report :(
> 
> 1) On the latest -current NFS client, and against both NetApp filers and
> FreeBSD 4.11 servers, mounting NFS/UDP,
> 
> I copied a .tgz image of the freebsd source tree to an NFS mount, untarred 
> it over NFS and built the kernel over NFS.
> 
> That works fine. I don't see any "server not responding messages".
> 
> 2) Then I did the same with a FreeBSD4.11 client and the FreeBSD -current
> server - copying the .tgz file over NFS and untarring it over NFS, to see
> if might be a -current NFS server issue. That works fine too.
> 
> Unfortunately, I am unable to mount -current against -current here.
> 
> I have a mix of bge and fxp cards.

I've seen the problem described by David on RELENG_6.

You can see the network topology at
http://troutmask.apl.washington.edu/~kargl/hpc.html
The NFS server is on node10 (bge1) and /home is exported to node11
through node15.  If I mount node10:/usr/src on node15 a "make installworld"
will result in a never ending stream of "server not responding messages".

If you have some ideas for tests, I repeat the exercise.

-- 
Steve