The below can be helpful once you have Orion access (to be done in INF305).

The following are all Unix commands (Linux or MacOS).

If you are a Windows user, please share your tips on what tools work best for you.

1. Remember to connect to the VPN first. Now you can log onto Orion:

   ssh <your login name>@login.orion.nmbu.no

2. On Orion, you have a folder called $SCRATCH that should be used for temporary content.
   For example, let us copy the spheres code with domain decomposition to that folder.
   
   On Orion:
   
   echo $SCRATCH
   
   This gives you the name of the folder, probably /mnt/SCRATCH/<your login name>

   Now from the local system:
   
   scp -r some-parallel-code <your login name>@login.orion.nmbu.no:/mnt/SCRATCH/<your login name>/
   
3. On Orion, check the list of available modules:

   module avail

   It is a very long list.
   For example, there is as of now a module "OpenMPI/4.1.5-GCC-12.3.0" which is the default for "OpenMPI".

4. Load the modules you would like to activate, for example:

   module load OpenMPI

5. You can always check the list of currently loaded modules using:

   module list
   
6. Compile the parallel code, for example:

   cd $SCRATCH/some-parallel-code
   make parallel-binary [or whatever your make target is for the parallel executable file]

7. We normally want to have the executable files outside the scratch folder, in the permanent storage, for example:

   mkdir ~/bin
   cp $SCRATCH/some-parallel-code/parallel-binary ~/bin/

8. Create a folder in scratch for your job, for example:

   mkdir $SCRATCH/some-test-job
   
   Copy your input data files into that scratch folder, or generate them directly there.
   Don't store large files elsewhere.

9. You need to create a bash script file to submit to the job scheduler Slurm.
   Details on this are found in the Orion documentation, and also in the documentation for Slurm.
   
   Let us assume that you created the bash script "some-test-job.qsub" on your machine locally.
   Now copy it via scp, or copy-paste its content using vim, whichever is simpler.
   
10. Now you can submit the job using "qsub" and check the queue status using "qstat":   

    qsub ./some-test-job.qsub
    qstat
    
11. Once the job is done, you will find its output in the scratch folder from which it was executed.

12. You will often want to copy the results back to your local system for further analysis.
    Recall that this needs to be done from your machine, not from Orion, for example as follows:
    
    scp -r <your login name>@login.orion.nmbu.no:/mnt/SCRATCH/<your login name>/some-test-job .