How do I fix the root cause of a stale NFS file handle?

[Gary Cooper_1]

I've got 8 rx2660's running HP-UX 11.31. They were all commissioned last June.

One is an NFS server and the other 7 are NFS clients.

Approximately once a month (no obvious pattern), I get a stale NFS file handle reported in syslog on one machine only. Which machine varies, but to date there have only been 2 clients that have reported the error (again, no pattern, but almost taking it in turns).

By the time I get round to checking the error, it has righted itself.

As these are manufacturing systems, what I'd really like to do is get to the root cause of the problem rather than just resolve the immediate issue of the stale NFS file handle (which seems to 'fix' itself anyway).

So the question is what could be causing the stale NFS file handles? It's obviously not the server going down or all 7 clients would complain about it at the same time (as well as Production jumping up and down big time). My current thoughts are that it could be the odd dropped packet, but what would cause that? Some faulty networking component? Cable, NIC, switch port? Could there be anything else causing this intermittent problem?

I just thought I'd post this to see if I can get anyone else's take on what might be going on here.

All ideas welcome.

Many thanks,

Gary

[KishoreR]

Hi Gary,

I hope by this time ,you would have found some answer to this.

Just to explain what is Stale file handle...

with respect to NFS protocol, Stale file handle means that the filehandle that client is referring is no longer a valid one on  NFS server. 

In most of the cases, the reason of this is if one of the client is is doing write/read on particular file and the same file is deleted on the server by a different nfs client.

Most probably this could be the reason on your environment, since there is only one server and 7 clients are accessing the same share( I am assuming it is the same exported directory for all 7 clients).

this error is part of protocol where NFS server can send to client.Should not be an issue with cable,nic etc..

Regards,

Kishore

[Gary Cooper_1]

Hi Kishore,

Thanks for the suggestion. I guess this could have been the issue.

Best regards,

Gary