Last updated on Nov 15, 2024 3 min read

HPC Cluster Live Demo Guide

1. Setting Up Conda Environment

Install Miniconda from Source

Download the Miniconda installer:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

Make the installer executable:

chmod +x Miniconda3-latest-Linux-x86_64.sh

Run the installer:
```
./Miniconda3-latest-Linux-x86_64.sh
```
- Follow the on-screen prompts to complete the installation.

Add Miniconda to your PATH:

export PATH=~/miniconda3/bin:$PATH
source ~/.bashrc

Verify the installation:
```
conda --version
```

Create and Activate a Conda Environment

Create a new environment (e.g., myenv) with Python 3.8:
```
conda create -n myenv python=3.8
```
Activate the environment:
```
conda activate myenv
```

2. Setting Up Jupyter Notebook

Option 1: Using PowerShell or Tabby for SSH Tunneling

Install Jupyter Notebook within the Conda environment:
```
pip install notebook
```
Start Jupyter Notebook on the HPC cluster without a browser:
```
jupyter notebook --no-browser --port=8888
```
On your local machine, open PowerShell or Tabby and set up an SSH tunnel:
```
ssh -L 8888:localhost:8888 <your-username>@headnode01.hpc.gla.ac.uk
```
- Accept the fingerprint prompt and enter your password.
Open the Jupyter Notebook link in your local browser:
```
http://127.0.0.1:8888/tree?token=<your-token>
```

Option 2: Using Visual Studio Code

Open VS Code and install the Remote - SSH extension.
Use the Command Palette (Ctrl+Shift+P) to connect to the HPC cluster:
```
Remote-SSH: Connect to Host
```
Enter the hostname:
```
headnode01.hpc.gla.ac.uk
```
Open a terminal in VS Code and follow the steps to start the Jupyter Notebook server.

3. Basic Linux Commands

These basic Linux commands will help you navigate and manage files in your cluster environment:

Check Current Directory:
```
pwd
```
List Files and Directories:
```
ls
```
Change Directory:
```
cd <directory-name>
```
Create a Directory:
```
mkdir <directory-name>
```
Move or Rename Files:
```
mv <source-file> <destination>
```
Delete Files and Directories:
```
rm <file-name>
```
or
```
rm -rf <folder-name>
```
View File Contents:
```
cat <file-name>
```
Check Disk Usage:
```
du -sh
```

4. Submitting a Job on the HPC Cluster

Create a simple Slurm job script (e.g., job_example.slurm) for submitting a job:

#!/bin/bash
#SBATCH --job-name=example-task
#SBATCH --partition=gpu
#SBATCH --gres=gpu:1
#SBATCH --mem=64GB
#SBATCH --cpus-per-task=8
#SBATCH --time=00:10:00

module load python/3.8
python example_script.py

Submit the job:

sbatch job_example.slurm

Monitor the job:

squeue -u <your-username>

5. Starting an Interactive Session

If you want to run commands interactively:

srun --partition=gpu --gres=gpu:1 --mem=64GB --cpus-per-task=8 --pty bash

This document should guide you through the demo and help your audience follow along with each step.

6. Slurm Commands and Job Scheduling Basics

Common Slurm Parameters

Slurm parameters allow you to specify your resource requests. Below are the most commonly used ones:

Example Parameter	Description
`--job-name="Test-Job-1"`	Specify a name for the job.
`--nodes=2`	Number of servers to run the job on. Use only for parallel processing.
`--cpus-per-task=8`	Number of CPUs per process. Useful for multithreaded jobs.
`--gres=gpu:1`	Number of GPUs to request.
`--mem=16G`	Amount of memory per node. Default is megabytes unless a suffix is used.
`--partition=gpu`	Specify the partition to use (e.g., `cpu`, `gpu`).
`--time=00-01:00:00`	Total runtime limit for the job.
`--mail-user=<email>`	Email address for notifications (e.g., job start, end, failure).
`--mail-type=BEGIN,END,FAIL`	Types of notifications to send.

Interactive Jobs

Interactive jobs allow you to debug submission scripts, install software, or prepare environments. Use interactive jobs sparingly and for short durations.

Example command:

srun --job-name="Interactive-Job" --partition=gpu --gres=gpu:1 --mem=16G --time=02:00:00 --pty bash

Batch Jobs

Batch jobs are the recommended way to run jobs on the HPC cluster. Below is an example of a submission script (job_example.sh):

#!/bin/bash
#SBATCH --job-name="Test-Job-1"
#SBATCH --partition=gpu
#SBATCH --nodes=1
#SBATCH --cpus-per-task=8
#SBATCH --mem=16G
#SBATCH --time=00-05:00:00
#SBATCH --mail-user=firstname.lastname@glasgow.ac.uk
#SBATCH --mail-type=BEGIN,END,FAIL

module load python
cd ~/work/calculations/pythonscripts
python myPythonCalcs.py