The getexample tool provide users on a HPC resource with a fast and easy way to begin using the HPC resource. Using getexample a directory of a working example is copied to the users directory that has typically a slurm script and an executable README script. The information in the files provide the user a working template to submit jobs to the resource manangement software. The users should be able to use these scripts to as a template that they can edit easily to run their jobs on the HPC resources.
To use getexample on the Pawsey Supercomputer is very easy! Simple load the getexample environment module.
module load getexample
The command to run getexample is getexample!
The getexample command by itself will show all the different examples available that are on that system. This is a list of directories that typically contain the executable README file and the slurm job script associated with that getexample.
To download an example the command is:
getexample name of example
This will copy the directory name of example to your current directory.
cd name of example
ls will list the contents.
To run the README the command is: ./README
Some useful SLURM VARIABLEs associated with SBATCH directives. How to set the SBATCH_ACCOUNT variable to default to the MYGROUP system variable local to the Pawsey Supercomputing Center.
#export SBATCH_ACCOUNT=echo $MYGROUP | awk 'BEGIN{FS="/"}{print $3}'
SLURM_SUBMIT_DIR is the directory from which the SBATCH command was run.
The slurm job scripts use some standard local variable to make the scripts consistent regardless of the job/example type.
JOBID=$SLURM_JOBID is used to so that we future proof the examples if we change resource manangers and it makes the scripts easier to read for new users.
SCRATCH is the name of filesystem ideally the parallel filesystem such as Lustre or GPFS where you will run your jobs. At the Pawsey Supercomputing Centre we allocate resources by projects and users are assign to these projects accordingly. To simplify the path naming foreach project/user on the SCRATCH filesystem we have created a local system variable called MYSCRATCH. Where MYSCRATCH is defined as /scratch/{project_id}/{user} this is key as it allows us to develop standard scripts for any user on any project. So we can provide better user support tools such as getexample.