There is only intel mpi installed on the machines. I don't know whether this can be solved with impi. Any more help?
2013/7/3 Rob Latham <[email protected]> > On Mon, Jul 01, 2013 at 03:36:57PM +0800, xihuang sun wrote: > > I installed orangefs2.8.7 on ten nodes and each of them acted as io node > > and meta node. then pvfs2-client is running on mdc node and node{1-10}. > > mount point is /mnt/orangefs. > > now when i am running mpi programs, they crashed and said > > > > > > 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). > > > > But the question is that I installed orangefs2.8.7 on local file system > and > > I am sure about it , where the NFS thing come from? > > That's a ROMIO message. ROMIO, an implementation of MPI-IO used > nearly everyhwere, can select file system specific functions, or fall > back to a general all-purpose file system. > > The logic for selecting which collection of "file system routines" to > use can be set two ways: > > - automatically: romio will stat the file or it's parent directory and > use that information to choose from available file systems. > > - manually: one can prefix the file (the whole file path) with > 'pvfs2:' and ROMIO will use the PVFS2 routines. > > Nice thing about this approach is ROMIO uses the "system interface" > and will bypass the kernel. > > It may be the case that your MPI implementation was not built with > PVFS support, in which case ROMIO will try to use general-purpose unix > routines on PVFS. As Kevin says, one of those is fcntl() to lock file > regions, which is not supported through the OrangeFS VFS interface. > > -- > Rob Latham > Mathematics and Computer Science Division > Argonne National Lab, IL USA >
_______________________________________________ Pvfs2-users mailing list [email protected] http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
