problems with voltaire mpi.

[Roy Dragseth]

Hi.

When I try to start an application linked with the voltaire infiniband 
using -comm=ib I get the following error:

Can't read SMP_ID

and if I use -comm=pmi i get 

Can't read MPIRUN_HOST

When I use the included mpirun from the voltaire stack it works as expected, 
also mpiexec works fine with mpich_chp4 applications.

Any obvious mistakes I'm making here?

Full output included below.

System info:

MPICH:
$ /usr/voltaire/mpi/bin/mpichversion
MPICH Version:          1.2.6
MPICH Release date:     $Date: 2004/09/20 12:35:29 $
MPICH Patches applied:  none
MPICH 
configure:        --with-device=vapi --with-arch=LINUX -prefix=/usr/voltaire/mpi.gcc.rsh --enable-sharedlib -lib=-Wl,-R/usr/voltaire/mpi.gcc.rsh/lib -Wl,-R/usr/voltaire/mpi.gcc.rsh/lib/shared -Wl,-R/usr/mellanox/lib -L/usr/mellanox/lib -L/usr/voltaire/mpi.gcc.rsh/lib -L/usr/voltaire/mpi.gcc.rsh/lib/shared -lmtl_common -lvapi -lmosal -lmpga -lpthread
MPICH Device:           vapi

OS:
RHEL 4, linux 2.6.9-34.

$ ../src/mpiexec-0.82/mpiexec -comm=ib  -v ./bandwidth.volt.x
mpiexec: resolve_exe: using absolute path "./bandwidth.volt.x".
Can't read SMP_ID
Can't read SMP_ID
mpiexec: process_start_event: evt 2 task 0 on compute-1-9.local.
mpiexec: process_start_event: evt 3 task 1 on compute-1-0.local.
mpiexec: All 2 tasks (spawn 0) started.
mpiexec: read_ib_startup_ports: waiting for checkin: 2 to accept, 0 to read.
mpiexec: process_obit_event: evt 4 task 0 on compute-1-9.local stat 1.
mpiexec: kill_tasks: killing all tasks.
mpiexec: wait_tasks: waiting for compute-1-0.local.
mpiexec: process_obit_event: evt 5 task 1 on compute-1-0.local stat 1.
mpiexec: process_kill_event: evt 6 task 1 on compute-1-0.local.
mpiexec: Warning: tasks 0-1 exited before completing MPI startup.

$ ../src/mpiexec-0.82/mpiexec -comm=pmi  -v ./bandwidth.volt.x
mpiexec: resolve_exe: using absolute path "./bandwidth.volt.x".
Can't read MPIRUN_HOST
Can't read MPIRUN_HOST
mpiexec: process_start_event: evt 2 task 0 on compute-1-9.local.
mpiexec: process_start_event: evt 3 task 1 on compute-1-0.local.
mpiexec: All 2 tasks (spawn 0) started.
mpiexec: wait_tasks: waiting for compute-1-9.local and 1 others.
mpiexec: process_obit_event: evt 4 task 0 on compute-1-9.local stat 1.
mpiexec: wait_tasks: waiting for compute-1-0.local.
mpiexec: process_obit_event: evt 5 task 1 on compute-1-0.local stat 1.
mpiexec: Warning: tasks 0-1 exited with status 1.

Best regards,
r.

-- 

  The Computer Center, University of Tromsø, N-9037 TROMSØ, Norway.
	      phone:+47 77 64 41 07, fax:+47 77 64 41 00
     Roy Dragseth, High Performance Computing System Administrator
	 Direct call: +47 77 64 62 56. email: royd at cc.uit.no