OpenShift DPF Automation

Complete automation framework for deploying NVIDIA DPF (DPU Platform Framework) on Red Hat OpenShift clusters with NVIDIA BlueField-3 DPUs.

🚀 Quick Deployment

One command does everything:

make all

This handles the complete deployment lifecycle:

• Creates OpenShift clusters using Red Hat Assisted Installer
• Deploys NVIDIA DPF operator with all prerequisites
• Sets up DPU-accelerated networking with OVN-Kubernetes
• Provisions worker nodes automatically via Bare Metal Operator

📋 Prerequisites

System Requirements

• Host: RHEL 8/9, 64GB+ RAM, 16+ CPU cores
• Hardware: NVIDIA BlueField-3 DPUs on worker nodes
• Network: Internet access, management and high-speed networks

Required Tools

• OpenShift CLI (oc)
• Red Hat Assisted Installer CLI (aicli)
• Helm 3.x
• Standard tools: jq, git, curl

Required Credentials

• Red Hat Pull Secret: Download from Red Hat Console
• Red Hat Offline Token: Generate at cloud.redhat.com/openshift/token
• NVIDIA NGC API Key: Create at NGC Portal → Account → Setup

🏃 Quick Start

1. Clone and Setup

git clone https://github.com/rh-ecosystem-edge/openshift-dpf.git
cd openshift-dpf
cp .env.example .env

2. Configure Credentials

# Add Red Hat offline token
mkdir -p ~/.aicli
echo "YOUR_OFFLINE_TOKEN" > ~/.aicli/offlinetoken.txt

# Add OpenShift pull secret (downloaded from Red Hat)
cp ~/Downloads/openshift-pull-secret.json openshift_pull.json

# Create NGC pull secret
cat > pull-secret.txt << 'EOF'
{
  "auths": {
    "nvcr.io": {
      "username": "$oauthtoken",
      "password": "YOUR_NGC_API_KEY",
      "auth": "BASE64_ENCODED_CREDENTIALS"
    }
  }
}
EOF

3. Configure Deployment

# Edit .env file with your settings
nano .env

# Essential settings:
CLUSTER_NAME=my-dpf-cluster
BASE_DOMAIN=example.com
VM_COUNT=1                    # 1=SNO, 3+=Multi-node
DPF_VERSION=v25.7.1

4. Deploy Everything

make all

⏱️ Takes 2-3 hours - fully automated, no user interaction needed

Environment Configuration (.env Generation)

All scripts and Make targets read configuration from a single .env file at the repo root. This file is generated — never edit the source files in ci/ for your local setup.

Source files (in ci/)

File	Role
ci/env.defaults	Default values for every optional variable. User environment variables always overwrite these.
ci/env.required	Variables that have no default and must be provided. Generation fails immediately if these are not set in user environment.
ci/env.template	The canonical set variables used by the scripts .env and in what order.

How to generate

1. Export your required variables (and any optional overrides):

   export CLUSTER_NAME=my-cluster
   export BASE_DOMAIN=example.com
   export API_VIP=10.1.150.100
   export INGRESS_VIP=10.1.150.101
   export DPU_HOST_CIDR=10.0.110.0/24
   export BFB_URL=https://content.mellanox.com/BlueField/...

   To keep overrides reusable, put them in a personal file (e.g. user.env) and source it first:

   source user.env
   make generate-env  

2. Run the generator:

   make generate-env                # creates .env (fails if .env already exists)
   make generate-env FORCE=true     # overwrites an existing .env

What happens under the hood

1. ci/env.defaults is sourced — sets defaults for every variable, but does not overwrite anything already exported in your shell.
2. ci/env.required is sourced — aborts with an error if any required variable is still unset.
3. envsubst renders ci/env.template into .env, substituting every ${VAR} with its resolved value.

Result

A flat KEY=VALUE file at the repo root (.env) containing the merged result of your overrides + the defaults. This is consumed by make and scripts/env.sh at runtime.

Validation

make validate-env-files

Checks that every variable in ci/env.defaults has a corresponding entry in ci/env.template so nothing is silently dropped, and report template-only variables that have no default.

📖 Documentation

Guide	Purpose
Getting Started	Step-by-step setup guide
Configuration	Environment variables
Worker Provisioning	Add physical worker nodes
Troubleshooting	Fix common issues
External storage	SKIP_DEPLOY_STORAGE and storage requirements

⚙️ Deployment Types

Single-Node OpenShift (SNO)

Perfect for development and edge computing:

VM_COUNT=1
RAM=32768                     # 32GB recommended
make all

Multi-Node Production

For high-availability environments:

VM_COUNT=3
API_VIP=10.1.150.100         # Required for multi-node
INGRESS_VIP=10.1.150.101     # Required for multi-node
make all

Production with Workers

Full DPU acceleration with worker nodes:

WORKER_COUNT=2               # Number of physical worker nodes
# Configure WORKER_*_BMC_IP, WORKER_*_BMC_USER, etc.
make all

VMs with static IPs

To assign fixed IPs to cluster VMs (e.g. for predictable addressing or firewall rules), set VM_STATIC_IP=true and provide these required variables in .env:

Variable	Description
VM_EXT_IPS	Comma-separated list of static IPs, one per VM (must have at least VM_COUNT entries). Example: 10.8.2.110,10.8.2.111,10.8.2.112
VM_EXT_PL	Prefix length for the subnet (e.g. 24 for /24)
VM_GW	Default gateway IP
VM_DNS	DNS server IP(s), comma-separated

Example for 3 VMs:

VM_STATIC_IP=true
VM_EXT_IPS=10.8.2.110,10.8.2.111,10.8.2.112
VM_EXT_PL=24
VM_GW=10.8.2.1
VM_DNS=8.8.8.8,4.4.4.4

Optional: PRIMARY_IFACE (default enp1s0) — interface name used for static config on the VMs.

🛠️ Key Commands

Complete Deployment

make all                     # Full automated deployment

Individual Steps (optional)

make create-cluster          # Create OpenShift cluster
make deploy-dpf             # Deploy DPF operator
make add-worker-nodes       # Add worker nodes

Management

make worker-status          # Check worker status
make run-dpf-sanity        # Health checks
make clean-all             # Complete cleanup

🎯 Supported Versions

• OpenShift: 4.20 (only supported version)
• DPF: v25.7+ (production), v25.4+ (legacy support)
• Hardware: NVIDIA BlueField-3 DPUs on Dell/HPE/Supermicro servers

🤝 Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

💬 Support

• Issues: Report problems in GitHub Issues
• Documentation: Complete user guides in docs/user-guide/
• Community: Join discussions in repository discussions

---

Get started with your first deployment: Getting Started Guide