[lug] Stale NFS data

[Ed Hill]

On Fri, 2005-07-08 at 15:43 -0600, John Hernandez wrote:
> We have 4 RH8.0 NFS clients with ro mounts to a RH8.0 server
> 2.4.20-28.8smp.  The files on the server are changed once every 3 hours
> by a remote rsync.  Following the rsync job, some clients still see the
> old version of the file, even though it appears the timestamp has
> changed.  The only way to get the clients back up to date is to remount
> the filesystem.
> 
> Has anyone seen this before?  We tried the 'noac' option to mount_nfs,
> but the resulting loads are too high and performance is dismal.

Hi John,

We've seen all sorts of NFS problems over the past few years which is
probably a direct consequence of the large-ish numbers of both machines
and Red Hat and Fedora versions (RH7.x--9, RHEL2--3, FC1--3) that we use
or have previously used.

On a positive note, recent 2.6.x kernels appear to be more stable when
used as both NFS servers and clients.  If you search through the LKML,
you'll find that many and, in some cases, serious NFS bugs have been
discovered and fixed since the releases of RH8 and RH9.

In general terms, three things that might help are:

 - try to reduce the load (network traffic) on the NFS server(s)
 - upgrade to a newer kernel (eg. current 2.6.x)
 - try using an automounter on the clients with a timeout 
     of a few hours

good luck,
Ed

-- 
Edward H. Hill III, PhD
office:  MIT Dept. of EAPS;  Rm 54-1424;  77 Massachusetts Ave.
             Cambridge, MA 02139-4307
emails:  eh3 at mit.edu                ed at eh3.com
URLs:    http://web.mit.edu/eh3/    http://eh3.com/
phone:   617-253-0098
fax:     617-253-4464