Subject: Re: Another recent NFS problem - actually atime problem
To: Manuel Bouyer <email@example.com>
From: Mark Davies <firstname.lastname@example.org>
Date: 07/08/2004 12:49:28
On Thu, 08 Jul 2004 06:43, Manuel Bouyer wrote:
> On Thu, Jul 08, 2004 at 01:26:50AM +1200, Mark Davies wrote:
> > Since updating an NFS server from 1.6ZH to 2.0F I've started seeing the
> > following problem:
> > >From Solaris 9 NFS clients, accessing some files now fail with the
> > > error:
> > "Value too large for defined data type" while those same files can be
> > accessed successfully from NetBSD and Tru64 clients.
> I'm seeing it from time to time too, using a NetBSD 1.6.2 server.
> I'm not sure it's a problem with the server, it may be caused by
> a solaris upgrade.
> Did you apply any solaris patch ?
No the Solaris boxes haven't been touched and its only happening with files on
the NFS server thats been updated to 2.0F, not those that are still at 1.6ZH
and its happening to quite a lot of files. (I have occassionally seen that
error on Solaris before but thats been for very large files - with holes -
which is not the problem here).
As I discovered in the tail part of my message the actual problem is that
something is setting the atime of the file(s) to a very bogus value and its
just that Solaris is complaining about that value while NetBSD and Tru64 are
happy to use it. If I "touch" the file on the server the atime comes right
and Solaris is happy to access it.
Still leaves the question what is breaking the atime's in the first place?