-
Jeremy Verrier authoredJeremy Verrier authored
Getting Started with the Meso@LR cluster
About
This is a training tutorial to lean how to use the cluster Meso@LR.
Useful links
- Meso@LR - Landing page for getting started with Meso@LR cluster.
- CheatSheet Slurm - TLDR for slurm commands.
- Meso@LR user documentation - Documentation made by the admins of Meso@LR
- Meso@LR admin mail - Where to ask if you need help.
- CNRS HPC related mailing-list - Where to ask if you need MORE help. Big mailing list of developpers, engineers and everything centered about HPC.
How to open an acount on Meso@LR
Return both forms filled and signed to Meso@LR Mail
How this training works
You should first clone the repository when you login to Meso@LR by executing this command =>
git clone https://forgemia.inra.fr/formationcalcul2022/formation-slurm.gitCloning the repo will just copy all the contents you see on your screen to folders on Meso@LR. If you don't know how to connect to Meso@LR, please do Exercise 1 first.
Multiple folders are available each one containing a different exercice. You can do all of them or you can pick the ones you think you need the most.
Let's compute!
IMPORTANT: You should run your jobs out of the folder named "scratch" on your account (which is a link to /lustre/$USER). The filesystem is a very fast one and provide vast amounts of storage. Do not run jobs out of your home, mainly because it won't work (when running a job on the cluster, everything expect "scratch" is mounted as read-only). These filesystems are slow and should only be used for backing-up the files that you produce on "scratch". Your /home directory on the cluster should only be used for storing source code, executables, Conda environments, singularity images and small data sets.
Glossary
jobs # A job is a program launched via SLURM.
nodes # A compute server
login-nodes # A server used to login on Meso@LRAdditionnal ressources
Authors
HPC engineer at INRAE
Other contributors
HPC support team at INRAE
Special thanks to
Forgemia team