Neil Brown wrote:
> On Saturday July 12, j.dittmer@portrix.net wrote:
> 
>>Problem:
>>Accessing the nfs shares on the Server gives lots of 'nfs stale file 
>>handles', making it unusuable. A simple cp from nfs to nfs triggers it 
>>in a matter of seconds.
> 
> This makes me a bit suspicious of hardware, probably networking.  It
> really looks like data is getting corrupted between client and server.
> 
> The fact that two different servers behaved differently while both
> running the same kernel, sees to support the hardware theory.
> 
> Maybe if you could get a tcpdump (-s 1500 port 2049) on both the server and the
> client  I could have a look at the filehandles as see if I can see why
> they are 'stale', and whether it could be a hardware problem.
> 
> NeilBrown
>

It seems, that it takes some time after starting the server for this to occur.
I attached to tcpdump logs from server & client from an incidence. What a I 
did was a kernel compile in a nfs directory. Can the 48s timeshift between the 
two computers be a problem? (ntpdate failed on the client) - but it also 
happens with synchronized time (just tested).
It seems to occur more often if there a multiple clients accessing the nfs 
server, but after giving it enough time it'll also happen with just one client.
Client was 2.6.0-test1 this time, server 2.5.75-bk1.
Additionally I thought I would give the newest nfs utils a try. But afterwards 
I'm only able to mount exactly one share after starting the 
'nfs-kernel-server', then I have to restart it to again being able to mount 
one share - known problem?

Thanks,

Jan

-- 
Linux rubicon 2.5.75-mm1-jd10 #1 SMP Sat Jul 12 19:40:28 CEST 2003 i686