On Tue, May 16, 2006 at 03:12:42PM -0500, Peng Gu wrote: > Hi all, > > For unknown reasons, it seems that pvfs2-1.4.0 is not happy with FC5. > So I directly build the CVS tree and now the pvfs is running. > > However, in my case, the pvfs2 became error prone. For example, even > mpi-tile-io test won't go through. > > Here is the errors I got: > > mpiexec -np 100 $MPITILEIO --collective --write_file --filename=/pvfs2/foo > # mpi-tile-io run on cse-wang04.unl.edu > # 100 process(es) available, 100 used > # filename: /pvfs2/foo > # collective I/O on > # 0 byte header > # 2500 x 400 element dataset, 32 bytes per element > # 25 x 4 tiles, each tile is 100 x 100 elements > # tiles overlap by 0 elements in X, 0 elements in Y > # total file size is ~30.00 Mbytes, 1 file(s) total. > [E 14:55:33.303022] invalid (unknown) I/O type specified > [E 14:55:33.303872] PVFS_isys_io call: Invalid argument > failed during MPI_File_(read or write) > [E 14:55:27.425927] invalid (unknown) I/O type specified > [E 14:55:27.426202] PVFS_isys_io call: Invalid argument > failed during MPI_File_(read or write) > [E 14:55:54.533111] invalid (unknown) I/O type specified > [E 14:55:54.533262] PVFS_isys_io call: Invalid argument > failed during MPI_File_(read or write) > [E 12:37:37.743449] invalid (unknown) I/O type specified > [E 12:37:37.745183] PVFS_isys_io call: Invalid argument > failed during MPI_File_(read or write) > > After that the program just hanged. > > Any ideas?
When you upgraded pvfs2 to CVS HEAD, did you also rebuild MPICH2 (at least the src/mpi/romio part, if nothing else)? After you rebuilt MPICH2, did you re-link mpi-tile-io ? I think if you do that, this error should go away. ==rob -- Rob Latham Mathematics and Computer Science Division A215 0178 EA2D B059 8CDF Argonne National Labs, IL USA B29D F333 664A 4280 315B _______________________________________________ Pvfs2-users mailing list [email protected] http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
