0

I have been running MPI programs on my testbed with ssh without problems. But when I wanted to switch to rsh to avoid encryption and run a program with mpirun, there is no output. I inspected the traffic with Wireshark and there is a TCP packet with PUSH flag, where the data says:

bash: orted: command not found

Open MPI is installed in the same directory in both machines, and they both have Ubuntu 16.04. I set it so there is no password for rsh needed within the testbed from the other machines. I can run programs on the remote machine with rsh, but not mpirun. Any idea what the problem could be?

María M.
  • 21
  • 3
  • It's probably the case that logging via `rsh` does not set the `PATH` variable properly. Try adding the `--prefix=/openmpi/base` option to `mpiexec`, where `/openmpi/base` is the base directory of the remote Open MPI installation. If it is the same as the one on the local host, simply call `mpiexec` by the full path to it, which will activate the same behaviour as the `prefix` option. By the way, use of `rsh` is strongly discouraged in favour of SSH with public key authentication. – Hristo Iliev Nov 08 '16 at 12:12
  • thanks, this helped! I didn't choose to use `rsh`, but my boss wanted to see it working. I thought what would be useful is a Wireshark dissector to see the MPI packets contents, but I haven't been able to find anything good around. – María M. Nov 08 '16 at 15:05
  • `orted` is the Open MPI shepherd process. It is responsible for spawning and controlling MPI processes on the remote nodes. `mpiexec` launches that process via SSH or rsh, then passes it the job description. `orted` then proceeds further and launches whatever MPI ranks are scheduled for running on that host. If it fails to start, e.g., because not found in `$PATH`, nothing MPI-related will. – Hristo Iliev Nov 08 '16 at 23:24
  • And keep in mind that in most if not all MPI implementations the raw data will be transferred unencrypted independently of the exact startup mechanism used. – Georg Bisseling Nov 10 '16 at 16:04
  • yes, I noticed. Now that it is working I can get past it and keep using `ssh`. But is there somewhere where the fields of the messages and their possible values are described? – María M. Nov 11 '16 at 08:06

0 Answers0