using MPI

[att0009]

I have fought for hours now to get PeriLab to work with MPI. It is PeriLab 1.5.2 , and the MPI on my machine is openmpi-4.1.1
Can someone assure me that it should indeed work? Is it possible I need an older, or different openmpi?

When I follow the "Getting Started" instructions from "https://perihub.github.io/PeriLab.jl/stable/man/getting_started/" exactly, I never see an error, the job just freezes at 0% and never progresses. The output files stay empty. I have tried various combinations of flags, and creating Julia project directories, and have received various combinations of error messages that I could potentially document in an "issue", but, before I go to that effort, I would greatly appreciate being assured that it has worked for someone else. Thank you,

[JTHesse]

Hi @att0009, I can confirm that MPI is working. For every commit we are testing different MPI scenarios for julia 1.11 and 1.12.
Can you let us know your current os and julia version and also try the debug mode:
mpiexecjl -n 2 julia -e 'using PeriLab; PeriLab.main("examples/DCB/DCBmodel.yaml"; debug=true)'

[att0009]

I'm embarrassed to realize, yesterday, I was testing "Dogbone.yaml" the whole time.
The "DCBmodel.yaml" works perfectly, but not the others I have tried. I attach the log files of the Dogbone.yaml and Disc_impact.yaml cases, which freeze at "Info: Run Verlet Solver" at 0%. To clarify, Dogbone and Disc_impact Exodus files are created, and correctly show the grid, but only show time=0. Whereas the exodus file for non-MPI includes many timesteps viewable in Paraview.

julia version 1.12.5
NAME="Red Hat Enterprise Linux"
VERSION="9.7 (Plow)"

Thank you for all the work you've put into this. I've used PD significantly at my current position, am well aware how much effort it is to build Peridigm, and appreciate the effort to make PD more accessible.

Disk_impact_2026_03_27_13_09_57_2.0.log
Disk_impact_2026_03_27_13_09_57_2.1.log
Dogbone_2026_03_27_13_23_22_2.0.log
Dogbone_2026_03_27_13_23_22_2.1.log