Submit Jobs to Leonardi

1. Concept:

• There are Head Nodes and Computing Nodes on supercomputer (Leonardi).

• You can drag/copy/paste/create directories/files on Head Nodes, BUT CANNOT run jobs on Head Nodes. You need to send jobs to the queue and they will run on the Computing Nodes which has the real power. Refer to section 2 for instruction how to run jobs.

• Run time for jobs needs to be approximated carefully. Shorter time enables your jobs to have higher queue priority, but it may kill the jobs before they finish. Longer time allows your job to run longer, i.e. better results, but may lower your priority.

• After the jobs are finished, you can transfer the output files back to your computer using FileZilla to view your results. Note: Only transfer the .csv file since the .e is heavy and takes much time to transfer.

• Important: you cannot run peacock on Computing Nodes. If you want to visualise your results, they need to be transferred from Leonardi to your local computer.

• More info: http://leonardi.unsw.wikispaces.net/file/view/leonardi-cluster.pdf

2. Instruction

• Connect to Leonardi using ssh command in your Terminal.

• To submit a job: qsub . Note: in this case is not your actual simulation, but a script to command Leonardi to run your simulation with specific computing information, such as 'wall time', 'number of cores', etc.

• To track status of jobs: squeue –u <username, usually zID>.

• ‘R’ means running. ‘C’ means cancelled/finished.

• To see when your job starts: squeue --start –j

• To delete a job: qdel <jobID; obtain from qsub/qstat>

• To watch job’s output in real time: tail –f

• For more info: https://github.com/pou036/redback/wiki/MOOSE-on-Leonardi@UNSW#user-content-how-to-use-leonardi