I set up 2 node cluster and installed mpich by using intel compiler. OS is CentOS 6.4
Each node is connected by ethernet. They shared directories where the compiler and mpich are installed, by using nfs.
SSH co
nnection can be done without password.
By way of experiment, I compiled simple example of MPI_File_open with mpicc and executed by mpiexec on the nfs directory.
The code is
#include "mpi.h"
#include
int main( int argc, char *argv[] )
{
MPI_Fint handleA, handleB;
int rc;
int errs = 0;
int rank;
MPI_File cFile;
MPI_Init( &argc, &argv );
MPI_Comm_rank( MPI_COMM_WORLD, &rank );
rc = MPI_File_open( MPI_COMM_WORLD, "temp", MPI_MODE_RDWR | MPI_MODE_DELETE_ON_CLOSE | MPI_MODE_CREATE, MPI_INFO_NULL, &cFile );
if (rc) {
printf( "Unable to open file \"temp\"\n" );fflush(stdout);
}
else {
MPI_File_close( &cFile );
}
MPI_Finalize();
return 0;
}
This example just opens the file and closes it in parallel.
On the single machine, it worked well.
However, it didn't work when 2 machines' processor executed the code in parallel.
The error message:
Internal Error: invalid error code 209e0e (Ring ids do not match) in MPIR_Bcast_intra:1119
Fatal error in PMPI_Bcast: Other MPI error, error stack:
PMPI_Bcast(1478)......: MPI_Bcast(b uf=0x1deb980, count=1, MPI_CHAR, root=0, comm=0x84000004) failed
MPIR_Bcast_impl(1321).:
MPIR_Bcast_intra(1119):
Fatal error in PMPI_Bcast: Other MPI error, error stack:
PMPI_Bcast(1478)......: MPI_Bcast(b uf=0x1deb980, count=1, MPI_CHAR, root=0, comm=0x84000004) failed
MPIR_Bcast_impl(1321).:
MPIR_Bcast_intra(1119):
When I added the prefix nfs: to the filename, as "nfs:temp", it worked well.
However, I don't want to add that because I should execute very big code on the cluster.
It is very hard to modify the code. Would you let me know what the problem is and how the problem is solved?
The nfs options are
(client)
mount -t nfs -o noac,nfsvers=3 "server_directory" "mount_directory"
(server, /etc/exports)
sync,rw,no_root_squash
I shared /opt and /home/username
mpich configuration is
MPICH2 Version: 1.4.1p1
MPICH2 Release date: Thu Sep 1 13:53:02 CDT 2011
MPICH2 Device: ch3:nemesis
MPICH2 configure: --prefix=/home/master/lib/mpich/mpich2-1.4.1p1 CC=icc F77=ifort CXX=icpc FC=ifort --enable-romio --with-file-system=nfs+ufs --with-pm=hydra
MPICH2 CC: icc -O2
MPICH2 CXX: icpc -O2
MPICH2 F77: ifort -O2
MPICH2 FC: ifort -O2
MPICH2 Release date: Thu Sep 1 13:53:02 CDT 2011
MPICH2 Device: ch3:nemesis
MPICH2 configure: --prefix=/home/master/lib/mpich/mpich2-1.4.1p1 CC=icc F77=ifort CXX=icpc FC=ifort --enable-romio --with-file-system=nfs+ufs --with-pm=hydra
MPICH2 CC: icc -O2
MPICH2 CXX: icpc -O2
MPICH2 F77: ifort -O2
MPICH2 FC: ifort -O2
| ------------------------------------------------------ Jaeyong Jeong Department of Mechanical Engineering Pohang University of Science and Technology
|
_______________________________________________
discuss mailing list discuss@mpich.org
To manage subscription options or unsubscribe: