Dear All, I am using Rocks+openmpi+hdf5+pvfs2. The soft on the rocks+pvfs2 cluster will output hdf5 files after computing. However, when the output starts, it shows errors: [root@nanohv pvfs2]# ./hdf5_mpio DH-ey-001400.20.h5 Testing simple C MPIO program with 1 processes accessing file DH-ey-001400.20.h5 (Filename can be specified via program argument) Proc 0: hostname=nanohv.columbia.edu Proc 0: MPI_File_open failed (MPI_ERR_IO: input/output error) -------------------------------------------------------------------------- MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD with errorcode 1.
If run in a none shared folder on the main node, the program goes well. it
shows:
Proc 0: hostname=nanohv.columbia.edu
Proc 0: all tests passed
The following is the setting of the PATH and LD_LIBRARY_PATH on one of the
nodes (I don't know whether it is because the hdf5 program can not find
something from openmpi I/O. What will be needed when it is going to input and
output files?):
[root@compute-0-3 ~]# $PATH
-bash:
/usr/kerberos/sbin:/usr/kerberos/bin:/usr/java/latest/bin:/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin:/opt/rocks/bin:/opt/rocks/sbin:/opt/gm/bin:/opt/hdf5/bin:/opt/meep-mpi/bin:/opt/openmpi/bin:/opt/pvfs2/bin:/root/bin:
[root@compute-0-3 ~]# $LD_LIBRARY_PATH
-bash:
:/opt/gm/lib:/opt/hdf5/lib:/opt/meep-mpi/lib:/opt/openmpi/lib:/opt/pvfs2/lib:
No such file or directory
[root@compute-0-3 ~]# mount -t pvfs2
tcp://nanohv:3334/pvfs2-fs on /mnt/pvfs2 type pvfs2 (rw)
[root@compute-0-3 ~]# ompi_info | grep gm
MCA btl: gm (MCA v2.0, API v2.0, Component v1.4.1)
The attached "log.out" is obtained by "./configure --with-gm
--prefix=/opt/openmpi | tee log.out"
Can anyone give some suggestions what is the reason of the input/output error?
MANY THANKS!!!
Best,
Jiangjun
log.out
Description: Binary data
