Unable to run openMPI from two machines

[sacharin1993]

Please submit all the information below so that we can understand the working environment that is the context for your question.

• If you have a problem building or installing Open MPI, be sure to read this.
• If you have a problem launching MPI or OpenSHMEM applications, be sure to read this.
• If you have a problem running MPI or OpenSHMEM applications (i.e., after launching them), be sure to read this.

Background information

What version of Open MPI are you using? (e.g., v4.1.6, v5.0.1, git branch name and hash, etc.)

It should be the latest version from openFoam installation (a 4 version) but I also built the latest version from your website (version5)
I probably have two version 4 and 5

Describe how Open MPI was installed (e.g., from a source/distribution tarball, from a git clone, from an operating system distribution package, etc.)

I have installed latest version of openFoam 2312 which comes with a openMPI version 4.
I also built the latest version 5 from your website

If you are building/installing from a git clone, please copy-n-paste the output from git submodule status.

Please describe the system on which you are running

• Operating system/version: Ubuntu 2204
• Computer hardware: MB=ASUS KRPA-U16 and
• AMD EPYC7713P on both machines
• Network type: I am using one ethernet port to connect the two machines

---

Details of the problem

I have two machines A and B with identical HW and SW. They seem to have no problems in ssh and sharing a folder (on A).
I can regularly run an example (as the hello_c.c from example folder) or a openfoam simulation in parallel using the 64 cores of a single machine with a command : ...$ mpirun -np 64 ./hello . Either on A and B machine.
If I try to run both machines as for example ...$ mpirun --hostfile /etc/hosts -np 168 ./hello the terminal hangs and no output is shown (error messages neither).
I am attaching some of the configurations of my system and the strace final part of the command ...$ strace mpirun --hostfile /etc/hosts -np 128 ./hello

Note: If you include verbatim output (or a code block), please use a GitHub Markdown code block like below:

shell$ mpirun -n 2 ./hello_world

documents.zip

[ggouaillardet]

/etc/hosts is not a valid Open MPI hostfile.

[sacharin1993]

Thanks for the reply but I am not sure I understand what you mean I should change.
In the past using a etc.hosts file as the ones I attached in the former message I managed to have openMPI run correctly with two nodes.
Even if I did not mention it, I have also tried -this time - using a "machines" text file to read nodes from A and B (just these names or IP and the names). If I just write A and have openMPI run with 64 cores $ mpirun --hostfile machines -np 64 ./hello , it works. If I change the name to B or A and B (and changing -np to 128 ) and launch mpirun from A , it does not work. Justs hungs as with the other situation (/etc/hosts)

[ggouaillardet]

make sure there is no firewall between both hosts (passwordless ssh is necessary but not enough). With Open MPI v4, try restricting to known interfaces, for example mpirun --mca btl_tcp_if_include eth0 --mca oob_tcp_if_include eth0 ...

[jsquyres]

@sacharin1993 Please see https://docs.open-mpi.org/en/v5.0.x/getting-help.html#for-problems-launching-mpi-or-openshmem-applications