Numeca SLURM Batch-System
 

Hi Numeca Users,

recently i set up my Simulation of an radial fan and now i want to start using a cluster with more computational power. The System uses a batch System (SLURM) and i cant find a right script to run my jobs. Unfortunately the NUMECA User-Guide doesnt provide any suggestions for this kind of batch system. Has anyone some experience with it and can suggest me a solution?

Thanks for your help
Daniel

Dear Daniel,

are you familiar with the procedure to set up a parallel / parallel portioned calculation with the Task Manager (a module integrated in the FINE/Turbo environment)?

Hey,

yes i´m familiar with setting up this tasks.

Hello Daniel,

I have never worked with SLURM, but read about it now. The Task Manager won't help you here I am sure now.

What you basically need is a shell script to start your computation. This shell script is handed over to the batch server, which decides when your task is executed. I have two ideas for this.

1) I added a script file as an example. You need to rename it to myscript.sh, as it is not allowed to upload shell scripts here. You may need to read in the FINE/Turbo manual the chapter "Launch FINE™ in Batch". The syntax and parameters for setting up different computations are described in detail there.

Make your script executable (I am not sure if this step is necessary):
move to where the script is located
chmod +x myscript.sh

Hand this script over to the batch server:
sbatch myscript.sh

2)
Execute the command line (fine -niversion 102 and so on) from the myscript.sh directly in the shell. Four result files should be created then, one of them is "a script file to submit the computation on queuing system."

Maybe this is the better one to hand over to your batch server. I suppose this is a shell script, too and it should work like in 1): sbatch however_its_name_is.sh


I hope at least one of the methods work and I hope I got your problem?

======

Here I read about sbatch command:
https://rc.fas.harvard.edu/resources/running-jobs/