2.6.0 NFS server giving 'stale NFS handle' errors

Been running a 2.4.21 NFS server with 2.4.21 and 2.6.0 clients without any
problems at all.

Switched server to 2.6.0 and now clients occasionally get "Stale NFS file
handle".

If i :

cd ~
find .

... it lists all of the files in my home directory, but occasionally i'll see
"Stale NFS file handle" scroll past. Problem is intermittent. Of the 8000+
files in my home directory, it might only produce 30 NFS errors, and each
time around it will be a slightly different list of files.

If i switch the server back to 2.4.21, problem disappears.

s wrote:

>
> Been running a 2.4.21 NFS server with 2.4.21 and 2.6.0 clients without any
> problems at all.
>
> Switched server to 2.6.0 and now clients occasionally get "Stale NFS file
> handle".
>
> If i :
>
> cd ~
> find .
>
> .. it lists all of the files in my home directory, but occasionally i'll
> see "Stale NFS file handle" scroll past. Problem is intermittent. Of the
> 8000+ files in my home directory, it might only produce 30 NFS errors, and
> each time around it will be a slightly different list of files.
>
> If i switch the server back to 2.4.21, problem disappears.

switching to TCP over NFS with 2.6.0 seems to have fixed the problem.

NFS/TCP/2.6 is noticably quicker than NFS/UDP/2.4.

s wrote:

> s wrote:
>
>>
>> Been running a 2.4.21 NFS server with 2.4.21 and 2.6.0 clients without
>> any problems at all.
>>
>> Switched server to 2.6.0 and now clients occasionally get "Stale NFS file
>> handle".
>>
>> If i :
>>
>> cd ~
>> find .
>>
>> .. it lists all of the files in my home directory, but occasionally i'll
>> see "Stale NFS file handle" scroll past. Problem is intermittent. Of the
>> 8000+ files in my home directory, it might only produce 30 NFS errors,
>> and each time around it will be a slightly different list of files.
>>
>> If i switch the server back to 2.4.21, problem disappears.
>
> switching to TCP over NFS with 2.6.0 seems to have fixed the problem.
>
> NFS/TCP/2.6 is noticably quicker than NFS/UDP/2.4.

spoke too soon .. TCP didn't solve the problem, now trying no_subtree_check