Skip to content

The Login Node

🚫 The Golden Rule: Do NOT compile or run computations on the login node.

The login node is a shared resource with a hard limit of 1 CPU core and 1 GB of RAM per user. It is shared simultaneously by every user on the system. Running a compilation, a Python script that loads large datasets, or a Monte Carlo job — even briefly — degrades the experience for everyone and may result in your process being killed automatically by the system.

Think of the login node as the foyer of a library: a place to orient yourself, organize your materials, and submit requests. The reading rooms (compute nodes) are where the actual work gets done.

What the Login Node IS for

✅ Allowed ❌ Not Allowed
Editing source files (vim, nano, emacs) Compiling C++/Fortran code
Writing and submitting Slurm scripts Running Python scripts with heavy computation
Managing files (cp, mv, ls, tar) Loading large datasets or ROOT files
Checking job status (squeue, sacct) Running Mathematica kernels
Light scripting (small awk, grep, sed) Any make, cmake, pip install with builds
SSH port forwarding for Jupyter Running parallel processes with mpirun

What Happens if You Violate This?

The system enforces resource limits via cgroups. A process exceeding its CPU or memory allocation will be killed without warning. Repeated violations may result in a temporary suspension of your account. When in doubt, use an interactive session:

srun --pty bash 

This command will create a slurm job with an interactive bash shell on one of the computing node.