Hi everyone,
I load module openmpi/4.0.2. However, when I use the “mpirun” command to run my code, an error message came out that “error while loading shared libraries: libpmi.so.0: cannot open shared object file: No such file or directory”. Do you know what is “libpmi.so.0”? Should it be “libmpi.so.0”?
What’s worse, even when I type basic commands like “ompi_info”, “mpirun --version”, I will get his error message. I appreciate any help. Thank you.
Zheming
I thought I figured out what happened. I need to load pmix first. After pmix module is loaded , I can run mpirun successfully. Anyway, thanks.
Yes, that is correct- the pmix module needs to be loaded alongside the openmpi module. Thank you for pointing this out- we will make a change to the ‘module load’ behavior of the openmpi module to have it load its dependent pmix module automatically.
Also, for openmpi jobs, we recommend using ‘srun --pmi=pmix_v2’ instead of ‘mpirun’ to launch MPI tasks.
Please take a look at our MPI documentation:
Link: https://carc.usc.edu/user-information/user-guides/software-and-programming/mpi