[PLUG] NFS : Does mounted file system type matter to client side?

[शंतनू]

On 04-Dec-2010, at 1:48 PM, Mayuresh wrote:

> Noticed a strange thing with NFS-NTFS combination:
> 
> There is an NTFS drive mounted read-only and shared read-only over NFS and
> mounted read-only on a client machine.
> 
> If a large file is transferred over this mount, the connection breaks
> midway with "stale NFS error".
> 
> Restoring the connection does not even require re-mount. You can just say
> cd to mounted directory to restore the connection and the file becomes
> visible again.
> 
> The amount of time elapsed or amount of data transferred till such error
> occurs varies a lot. (In fact the file does not have to be necessarily
> large. Sometimes the problem occurs with small files, too, though the
> likelihood of hitting the problem increases with large files.)
> 
> However if same large files are kept on an ext4 system then the transfer
> goes through fine between the same two nfs client and server machines.
> 
> In all these experiments, network quality was good. Both wired and
> wireless networks showed the same behavior.
> 
> Now one could conclude that there is a problem with NFS - NTFS combination
> somewhere. However, now if I use the same combination of NTFS drive and
> NFS server but with a different client machine then the whole file is
> transferred fully without any problems.
> 
> So, only with a certain nfs client machine and only when the mounted file
> system is NTFS there occurs a problem during transfer of files.
> 
> NFS client being sensitive to mounted drive type at server end defies the
> logic. Is there such dependency even possible from NFS architecture point
> of view?
> 
> Mayuresh.

<snip>
A filehandle becomes stale whenever the file or directory referenced by the handle is removed by another host, while your client still holds an active reference to the object. A typical example occurs when the current directory of a process, running on your client, is removed on the server (either by a process running on the server or on another  client).
</snip>

Is there any other machine which trying to access/modify the file which you are trying to copy?

NFS is stateless protocol. No need to change any thing at client end.
Since ext4 works and ntfs gives some issue, I suspect this issue is related to mounting the fs on the server.
Sharing following information could be useful to find out the problem.

- output of mount command (which will tell more about the mount options used)
- ntfs driver used for mounting the ntfs