GEOCEAN — SCI-UC Cluster Guide

How the GEOCEAN group works on the SCI-UC cluster (ui.sci.unican.es). This complements, and does not replace, the official docs at doc.sci.unican.es — for anything cluster-wide (backup policy, account requests, helpdesk), that site is the source of truth. These pages only cover what's specific to how GEOCEAN works.

Support: soporte.sci@unican.es

Access

  • Login node: ui.sci.unican.es (193.144.184.31)
  • Auth: ed25519 SSH key + a cluster password set on first login
  • The login node is not for computing (limits: 2GB/4GB virtual memory, 512 processes, 120 min CPU time per process). Use Slurm for anything heavier.
# generate a dedicated key for this cluster
ssh-keygen -t ed25519 -f ~/.ssh/id_ed25519-ui

# connect
ssh -i ~/.ssh/id_ed25519-ui <username>@ui.sci.unican.es

# first login only — set your cluster password
passwd

New member checklist:

  1. Request an account via the SCI-UC portal (ACCESS → Request Access).
  2. Get added to the geocean group by an admin.
  3. Set up your SSH key and password as above.
  4. Read the storage guide before copying any data over.

Storage (short version)

See Data & Storage for the full guide.

Path Use for
/nfs/home/geocean/$USER dotfiles, scripts, small configs, source code
/lustre/geocean/WORK/$USER active job I/O, intermediate output
/lustre/geocean/DATA shared, long-lived group datasets
/nfs/software/geocean/ group-built software, modulefiles
/scratch (per node, during a job) fast scratch, wiped after the job ends

Slurm partitions

GEOCEAN owns two partitions across eight nodes, geocean01–geocean08.

Partition Nodes Cores/node Mem/core Notes
geocean geocean{02-06} 48 2GB default, unlimited max general group queue
geocean_priority geocean{01,07-08} 24 2GB default, 256GB max reserved nodes — ask before defaulting here 
# quick one-off job
sbatch --partition=geocean -t 00:10:00 --wrap "hostname; sleep 30"

# interactive shell on the group partition
srun --partition=geocean --pty bash

#!/bin/bash
#SBATCH -J swan_run
#SBATCH --partition=geocean
#SBATCH -N 1
#SBATCH --ntasks=48
#SBATCH --hint=nomultithread
#SBATCH --time=1-00:00:00
#SBATCH --mem-per-cpu=3G
#SBATCH -o %x-%j.out
#SBATCH -e %x-%j.err

module use /nfs/software/geocean/modulefiles
module load delft3d      # or whichever group module you need

srun ./run_model.sh

Monitoring & control:

squeue -u $USER                                 # your jobs
seff <jobid>                                     # CPU/mem efficiency, after completion
scancel <jobid>                                  # cancel one
scancel -u $USER                                 # cancel all of yours
srun --jobid <jobid> --overlap --pty bash -i     # attach to a running job

Group software

Group-built tools live under /nfs/software/geocean/<name>, published as Lmod modules:

module use /nfs/software/geocean/modulefiles
module avail
module load delft3d

Nine models are currently published this way — xbeach, delft3d, swan, swash, schism, sfincs, snapwave, lisflood, riskscape. See Software Modules for how to load and run each one, and Model Development if you need to build a new version or add a model — that page also covers the shared spack/conda toolchains under /nfs/software/geocean/.

/nfs/software/geocean/ also still has old, pre-models/ hand-built directories with the same names as the models above (e.g. a bare /nfs/software/geocean/swan/, with no models/ in the path) — these are legacy and being cleaned up, see Model Development. Always go through module load <name>, not a path guess.

New group tools go under /nfs/software/geocean/<toolname>. If others will reuse it, add a modulefile under /nfs/software/geocean/modulefiles/ — see Model Development for the convention.