HPC Cluster

An Ansible-powered project that automates the creation of a fully functional HPC cluster — from bare nodes to parallel-ready in minutes.

Requirements

• Proper hostnames and networking in place
• Ansible (controller machine)
• SSH access to all nodes
• Python available on managed hosts

Architecture Overview

This setup is designed around a multi-node architecture managed via Ansible, consisting of a single master node and multiple worker nodes.

				       USER
				 	     |
				  ┌──────▼──────┐
				  │    MASTER   │
  				  └──────┬──────┘
					     |
			 ┌───────────┼───────────┐
  			 │           │           │
		┌────▼────┐ ┌────▼────┐ ┌────▼────┐
		│ Worker1 │ │ Worker2 │ │ Worker3 │
		└─────────┘ └─────────┘ └─────────┘

Offline workers friendly setup

Node Roles

Master node

• Acts as the control node

• Has internet access

• Runs Ansible and orchestrates configuration and deployment

• Connects to all worker nodes over an internal network

Worker nodes

• Have no direct internet access

• Are reachable only through the internal network

• Receive all configuration, packages, and updates via the master node

Supported Operating System Setups

The Ansible roles are written to support two deployment scenarios:

1. Mixed RHEL-based Setup

Master: RHEL 9

Workers: Rocky Linux 9

2. Debian-based Setup

Master: Debian 12

Workers: Debian 12

Roles

This project is organized around Ansible roles, each responsible for one concern.

Feature	RedHat	Debian
nginx server	✓	✓
Mirror server	✓	✓
Mirror client	✓	✓
DNS server	✓	✓
DNS client	✓	✓
NFS server	✓	✓
NFS client	✓	✓
Slurmd	✓	✓
Slurmctl	✓	✓
Slurmdb
Kubernetes cluster	✓	✓
Xcat
LDAP server
LDAP client

---