Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.nyc-ai.app/llms.txt

Use this file to discover all available pages before exploring further.

This walkthrough takes you from a fresh HPCC account to a running job in about ten minutes. You’ll log in, load a module, submit a minimal SLURM job, and read its output.
Prerequisites: An approved HPCC account. If you don’t have one, start with Accounts & access.

​
Step 1: Log in

From your laptop, SSH to the Chizen gateway and then onto the cluster you want to use (Arrow is the default for most users): 
ssh <your_username>@chizen.csi.cuny.edu
ssh <your_username>@arrow
If you’re off-campus and connections fail, confirm you’re reaching an externally accessible gateway (Chizen or Karle only). All other systems require you to go through one of those first.

​
Step 2: Check your storage

Home and scratch are separate. Home is small and durable; scratch is larger and temporary. Check both: 
df -h ~                    # your home directory (/global/u/<username>)
df -h /scratch/$USER       # your scratch workspace
The default home quota is 50 GB / 10,000 files. Jobs must run out of scratch; don’t launch from your home directory. See Storage & quotas for full details.

​
Step 3: Find a module

The HPCC uses the LMOD environment module system on Arrow. Browse what’s available: 
module avail            # list all available modules
module spider python    # search for Python variants and their dependencies
module list             # show what's currently loaded
Load the module you want: 
module load Python/3.10.4       # exact version string may differ; check module avail
See Software & modules for the full workflow.

​
Step 4: Write your first SLURM script

Move to your scratch directory and create hello.sh: 
cd /scratch/$USER
mkdir -p hello-slurm && cd hello-slurm
hello.sh
#!/bin/bash
#SBATCH --job-name=hello
#SBATCH --output=hello.out
#SBATCH --error=hello.err
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=1
#SBATCH --mem-per-cpu=1G
#SBATCH --time=00:01:00

cd $SLURM_SUBMIT_DIR
echo "Hello from $(hostname) - job $SLURM_JOB_ID"
sleep 5
echo "Done."
The minimal template above omits --qos and --partition, which real HPCC jobs typically require (for example --qos=qoschem --partition=partchem). Ask your PI or the HPC Helpline which QOS / partition applies to your project before submitting production work.

​
Step 5: Submit and monitor

sbatch hello.sh          # submit; prints the job ID
squeue -u $USER          # see your jobs in the queue
scancel <jobid>          # cancel if needed
Once the job completes, the hello.out file in your working directory will contain the output. 
cat hello.out

​
What next

More SLURM examples

OpenMP, MPI, hybrid, GPU, and array templates.

Software & modules

Load compilers, MPI libraries, Python, Julia, and more.

Storage & transfers

Move data in and out with SFTP, Globus, or iRODS.

Policies

Rules every HPCC user needs to follow.