[pnfs] MDS exporting pNFS and non-pNFS file systems

[Dean Hildebrand]

In general, you want to set the file-system attribute fs_layout_type 
per-file system.  Currently, gpfs can only set this attribute globally 
using a gpfs-specific admin command. (there is a server export op to 
retrieve this attr from the file system called layout_type).  Using a 
mount option might make sense, or the exports.  We could also just add 
the fsid or mount point as an arg to our gpfs admin command.

Back with pvfs2 I also only supported global configuration using a 
/proc/pvfs/xxx file.

Hope that helps,
Dean

Muntz, Daniel wrote:
> What are people doing to support exporting pNFS and non-pNFS file
> systems from an MDS?  The information from EXCHANGE_ID is not
> per-file-system, so can't be used to determine if a particular file
> system is pNFS-enabled.  Based on talking with Ricardo and Mike, we were
> thinking perhaps GETATTR could be used.  E.g., GETATTR could return a
> layout type of 0 for entities that do not support pNFS.
>
> Perhaps the IBMers have some thoughts on this.  If you had a server
> exporting 2 GPFS file systems, and you wanted to export one with pNFS
> and one without, how would you want to handle this in the protocol?  For
> that matter, how would you want to handle it from a UI perspective?  For
> GPFS, an /etc/exports flag might make sense.  This wouldn't really work
> well for spNFS though.  For GPFS I assume you could switch back and
> forth between exporting with/without pNFS.  For spNFS, you're committed
> to exporting with pNFS (exporting without pNFS would result in access to
> the files on the MDS).
>
>   -Dan
> _______________________________________________
> pNFS mailing list
> pNFS at linux-nfs.org
> http://linux-nfs.org/cgi-bin/mailman/listinfo/pnfs
>