Mjolnir consist of both CPU and GPU nodes, each with very different CPU/GPU and memory setups. Please note the differences in architecture when you are booking your jobs as it will affect the efficiency of the running jobs. If you book more resources than you use, then you will limit your colleagues to compute as well. Its a balance that needs to be adjusted.
We have one entrance to Mjolnir called mjolnirgate. This is a DNS pointing to a physical server (mjolnirgate01fl) and several virtual servers serving as backup. No heavy computation is allowed at the resulting login node, neither physical nor virtual server. All computation has to be submitted to the SLURM queue. Only rsync, sftp, editing data, moving data, or similar, is allowed.
Users systematically running computation at Mjolnirgate will get their access revoked.