NFS and Mac OS X Server

[Mike Mackovitch]

On Fri, Sep 14, 2007 at 11:22:00AM -0400, Lance Westerhoff wrote:
>
> The plot thickens!  So once I successfully delete the directory by logging 
> into the osx_server, anything further done to this now-non-existent 
> directory (such as trying to create a file/directory with the same name) 
> yields a "Stale NFS" or "File exists" error.  So from the beginning:
>
> [user at rh_client ~]$ rm -rf directory/
> rm: cannot remove directory `directory/': Device or resource busy

I can't say that I've ever seen that particular error.
It would be interesting to see if that error code is being
explicitly returned by the server (or if it's something the
client is coming up with).

> osx_server:~ user$ rm -rf directory/
> [user at rh_client ~]$ touch directory
> touch: setting times of `directory': Stale NFS file handle
> [user at rh_client ~]$ mkdir directory
> mkdir: cannot create directory `directory': File exists

That sounds like maybe a caching issue on the client.

It's normal to see a "stale NFS file handle" error when a client
knows about an object, it gets deleted on the server (locally or
by a different client), and the client tries to access that same
object after it's been deleted.  However, even in that case this
error condition should not exist.  The client should be dropping
it's stale reference (and getting a reference to any new object
if it's been replaced).

>> Thank you for any insights you can provide!

I would suggest contacting Apple about this issue - for example by
filing a bug (http://bugreport.apple.com) or contacting support if you
have a contract.

Please include information about what file system type is exported,
what export options are used, what mount options are used on the
clients.  tcpdump capture files (ala "tcpdump -ns0 -w file.pcap")
demonstrating the problem are always helpful.  Also, it can be
helpful to provide detailed info about how/when the directory/files
are being created.

HTH
--macko