Batch system

From Etp
Jump to: navigation, search

SLURM Batch System

Since Jan 2015 all nodes are setup for Slurm. The old SGE system has been decomissioned.

For Slurm submission you need to be logged in to one of the login nodes.

For job submission and job query it is important to specify our part of the Linux cluster via option --clusters lcg (similar to queue names in other batch systems).

Basic commands are similar to other batch systems:

  • Submitting jobs: sbatch <job-script>, e.g.
uh351ac@lxe14:~/gduck> sbatch
Submitted batch job 346029 
  • job scripts are basically shell scripts with a few extra lines for Slurm specific setup, here is a simple example script
  • some basic setup stuff:
    • #SBATCH --clusters=lcg # and
    • #SBATCH --partition=lcg_serial # are needed to identify proper par of cluster for LMU/ATLAS.
    • #SBATCH --qos=lcg_add # should be set for extra priority
  • squeue --clusters=lcg -u $USER gives information on your running jobs
  • scancel <job-id> kills a job

More info you can find in LRZ Slurm doc or Slurm home page.

And a more useful job-script running Root ntuple analysis

See also

  • Grid, Athena, Root: How to setup Athena on login nodes lxe16 / lxe31. (Setting up Athena from /cvmfs/ doesn't work. If you get errors like ERROR:root:code for hash md5 was not found., take a look there for help.)