Also note: on large clusters, you might want to guard the execution to only do it on mother superior (i.e. if PBS_JOBID is defined, to do it only when PBS_NODENUM and PBS_TASKNUM are both 0) if you have programs that use ssh for remote processes:
See:
https://lmod.readthedocs.io/en/latest/110_lmod_mpi_parallel_filesystem.html
But then again, sometimes not, if your remote processes rely on modules to be initialized to find what they need.
Dear Alexis,
Thank you for such a detailed reply ,
The situation is that my master node SSD’s have failed due to the firmware not updated for the HDD on the node , so unfortunately , i can’t check this , now , the thing I’ve noticed is that you’ve mentioned the installation of environment modules package .
Now the predicament i was in is that i used OpenHPC for installation of the cluster and there was a package called Lmod which was clashing with the environment modules package , on the OpenHPC forums .
https://groups.io/g/OpenHPC-users/topic/lmod_error_any_ideas/806056
Now I’m not really sure i did understand this or how did Adam Roe solved it , but the situation was that i couldn’t install environment modules whatsoever .
That’s probably because you have “modules” from another RPM. Unfortunately you’ll have to find out exactly where the files are and how it behaves, but as I demonstrated you can actually get inspiration from a recent environment-modules RPM on another machine on how to deal with it.
Dear Alexis ,
Yes I’ll check for this environment variables clashes or the module packages , I will update after I reinstall the nessecary packages .