A computing cluster was set up to train detection algorithms. The cluster administration is based on slurm and ansible. The cluster is distributed over two computer rooms. Slurm has an automatic fail-over and thus increased reliability.