On Mar 18, 2006, at 5:40 AM, Frank wrote:
XGRID_CONTROLLER_HOSTNAME and XGRID_CONTROLLER_PASSWORD are
properly set
up, Open-MPI 1.0.1 is installed on all machines (with the same
configure
options). When configured with --prefix=/usr/local/openmpi my app is
supplied to the xgrid controller and I can see that copy's of my
app are
"supplied" to the other machines, too - but the jobs hang, nothing
happens (user nobody has full access to the folder /usr/local/myapp
where my app is run). /usr/local/openmpi/bin and /usr/local/openmpi/
lib
are added to the variables PATH and DYLD_LIBRARY_PATH on every
machine,
too. I'm running into this situation no matter from which machine
my app
ist started. To the guys with openmpi and xgrid performing correct:
which configure options did you use? The firewall is told not block
any
internal traffic on the subnet. When not using the xgrid my app
performs
correct.
Has anyone any idea concerning this matter?
My first guess was going to be the firewall issue, but if you can run
without XGrid, that probably isn't the case. Could you try an XGrid
run with the -d option to mpirun? That will enable some debugging
output that should help determine what is going wrong.
Thanks,
Brian
--
Brian Barrett
Open MPI developer
http://www.open-mpi.org/
