Skip to content
hpcnotes
Notes
  • LineShine, an exaflop without GPUs
  • Working on a shared cluster
    • Login nodes are not compute
    • Environment modules are a contract
    • Anatomy of a Slurm job that actually finishes
    • From login node to first GPU kernel
Thoughts
  • What a note can carry
Loading…
2026.09.08

Working on a shared cluster

A cluster is a waiting room, a contract for software, a scheduler, and only then an accelerator. These notes stay in that order.
The login image is not a workstation. The modulefile is the stack you actually have. A job that finishes is mostly walltime, cwd, and the launcher. A GPU run starts with a device you can see, not the full application.
The four-rank cartoon is the shape of an srun, not a screenshot from a centre.
Four ranks on a mesh, each with a neighbour exchange
Four ranks on a mesh, each with a neighbour exchange
Slurm's sbatch is the usual door. The first page is login-node etiquette.
Login nodes are not compute........................................................................................................................................................................................................1
Environment modules are a contract........................................................................................................................................................................................................2
Anatomy of a Slurm job that actually finishes........................................................................................................................................................................................................3
From login node to first GPU kernel........................................................................................................................................................................................................4
Login nodes are not compute........................................................................................................................................................................................................Next
#batch#slurm