-
for info about cluster see documentation
-
especially, focus on section
Slurm partitions
, here we are interested inslim
-
log in to server (my username is pospisil):
- list the available modules (to see what can be loaded - compilers, MATLABs, etc..)
module avail
- see the status of server
sinfo
- see who is computing what
squeue
- load git module (because default is not working)
module load git/2.13.0-gcc-7.1.0
- clone repo with sample code
git clone https://github.com/lukaspospisil/matlab_on_server_test.git
- run interactive session on
1 node, 1 process, 20 cores (all), slim partition, for 30 minutes, all available memory
salloc -N 1 -n 1 --mincpus 20 --partition=slim --time 0:30:00 --mem=64040 --exclusive
- load MATLAB module
module load matlab
- check loaded modules
module list
- test running the application from interactive session
cd ~/matlab_on_server_test
matlab -nodesktop -r "myparameter=10;runme;exit"
- check if testing output file has been created
cat myoutput.txt
- close interactive session
exit
- see batch script with the definition of the job (
%j
in file denotes the unique job id)
cat myexample.batch
- send batchscript to slurm
sbatch myexample.batch
- see if we are in queue
squeue
(but maybe now there will be nothing since the "problem" is already computed)
- maybe we made a mistake and we want to "kill" the process
scancel [job_id]