Useful Truba resources

Truba is a high performance computing platform which allows researchers to pursure research on algorithms or models which require tremendous amount of memory and computing resources. There are several tips and tricks to use Truba. In this post I am sharing the ones that I use mostly to refer for the future use. Most of the commands used are SLURM scripts.

• To check and make sure about the necessary edits about the command that I prepared I use nano to open the command file before running it.

• Most used SLURM commands for me are sbatch , squeue & scancel.

• Presentation about basic SLURM commands from University of Utah.a basic intro which helps a lot.

• Truba handbook - TR Official handbook

• Folder systems from Truba handbook - TR A very important note on this section: Managed to run ANTs on /truba_scratch using a very basic workaround. Originally ANTs runs on /tmp which has a limited memory and deletes the temporary files once the code runs. -figuring this out took a lot of time-

• UFL note on tmp it seems like this note can be applied to TRUBA as well.

• HPC DOCS more on /tmp

• Slurm user group discussion on /TMP this is an active group to ask questions and search for the answers when needed.

• Example script of TMPDIR useful to have a script to plan a working way and check the possible mistakes.

• SLURM –chdir alternative Now I am adapting my scripts accordingly via bash instead of SBATCH –chdir (took me a day to fix)

• Workload manager useful commands

• A Presentation about SLURM do not forget about exporting the TMPDIR here there are some tips

• General truba presentation - TR helps to understand TRUBA architecture.

• Slurm documentation this is the handbook to use as a reference when stuck.

• Princeton’s documentation about SLURM job priority even though this is specific to Princetons system SLURM tricks are useful when in need.

• SLURM commands & JOB REASON CODES handy one.

• Properties of truba clusters to check before sending the job Useful to consider job limitations.

• HPC workdir example gooseslurm is a slurm wrapper but its documentation contains general slurm and bash cheatsheat too.

• Bash strict mode Bash strict mode to check codes. NOTE: As I am informed recently TRUBA does not allow to use two clusters at the same time with their maximum cpu limits. Use barbun or hamsi not both at the same time to avoid AssocGroupCPULimit error on hamsi since hamsi is faster.

_{[^*] This content is updated on 31.03.23}