On Thu, May 13, 1999 at 03:55:42PM +0000, Kent Polk wrote:
> I've been running an NFS mount from a Solaris 5.7 host to a NetBSD
> 1.3K #2: Fri Mar 19 15:47:24 CST 1999 (beta release) NFS client
> for a few months now with no problems until last week.
> 
> Reads and writes from the clinet to the NFS server began to take
> longer and longer, with typical 20k writes taking in excess of a
> minute to complete and reads often taking almost as long.

  Have you done any tcpdump logging yet?  If your network isn't
saturated, the only reason for such long delays that comes to
my mind is a _whole_ lot of retransmits.  tcpdump ought to be able
to help the More Knowledgeable Ones (and maybe the Less
Knowledgeable Ones like me! :) help you out.

  If you're not familiar with tcpdump, the following command should
do the trick, at least for a first pass.

tcpdump -l -vvv -s 10000 host marvin and host orac

  Replace "marvin" and "orac" with the names of your server and
client.  You may want to pipe that through "grep nfs" to
eliminate other traffic.  

  If you can put the full dump of a single write up on an ftp
site or http site, and reply to the list with the URL, those
interested can get all the gory details!

  Also, what are your NFS options -- rsize, wsize, intr, etc? 
Are these NFS v3 or v2?  Are you sure they are via UDP vs. TCP?

  Hope this helps!

  Brian