Often experiencing nfs server foo:/bar: not responding

[Bernd Walter]

On Sun, Jun 25, 2006 at 09:49:30AM +0000, Philippe CASIDY wrote:
> David O'Brien wrote:
> >I am getting these errors all the time now (now being -CURRENT newer than
> >Dec'05-Jan'06 time frame).  Are there some known issues in UDP or NFS
> >serving since then?  This is on a virtually zero loaded 100Mbit network.
> >Both the NFS server and client are FreeBSD-CURRENT systems.
> >
> >I can trivially trigger this on all my FreeBSD-CURRENT NFS clients,
> >simply by exiting Vim.  Did something change sometime in 2006 that would
> >affect the default NFS mounts?
> >
> 
> I experience this problem from now and then too.
> The problem was gone when I upgraded my server to from 5.1 to 
> 5.5-PRELEASE. Before that, I added an entry in the server's crontab to 
> launch a script that does :
> 
> #!/bin/sh
> /sbin/ifconfig re0 mtu 1490
> /bin/sleep 15
> /sbin/ifconfig re0 mtu 1500
> 
> I had then bad network performances but no more NFS lockdown.

This is likely a different story - re chips have broken hardware
checksum offloading.
All retransmission get the same defect until you force a different
pattern with the MTU change.
The feature was turned off by default later.

-- 
B.Walter                http://www.bwct.de      http://www.fizon.de
bernd at bwct.de           info at bwct.de            support at fizon.de