[Lustre-discuss] Lustre and MPI-ROMIO, specifically ADIOI_Set_Lock and noncontiguous writing

[Andreas Dilger]

On Feb 22, 2008  08:47 -0500, Bill Wichser wrote:
> I'm trying to do noncontiguous writes on a Lustre filesystem using 
> MPI-ROMIO routines.  It is failing in ADIOI_Set_Lock with the message:
> 
> File locking failed in ADIOI_Set_lock. If the file system is NFS, you
> need to use NFS version 3, ensure that the lockd daemon is running on
> all the machines, and mount the directory with the 'noac' option (no
> attribute caching).
> 
> Now I've searched the web for some solution but only seem to come across 
> similar problems.  Is there some way to mount a lustre filesystem so 
> that it somehow does the right thing here and use the ufs kernel 
> interface?  NFS mounted filesystems work fine as they provide the file 
> locking mechanism apparently.

The default for Lustre is to _not_ support cluster-coherent flock locking,
because it can cause problems in some cases (depending on locking mode).

There are client mount options "-o localflock" which just pretends that flock
locking is working (it is using only the local node's flock locking), and
"-o flock" which enables real filesystem-wide flock.  You can pick between
the two, depending if you need real flock or not.

Also, there have been improvements for the Lustre ADIO driver done at ORNL.

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.