ComfyUI on Linode RTX 4000 ADA Instance

Docker based ComfyUI deployment with Linode RTX 4000 ADA Instance, HTTPS reverse proxy, and GPU monitoring.

What is this project?

This project provides an simple automated deployment script for installing ComfyUI, a powerful and modular stable diffusion GUI.

script automatically install

• ComfyUI - Node-based interface for Stable Diffusion image generation
• Caddy - Automatic HTTPS reverse proxy with basic authentication
• GPU Monitor - Web-based nvtop interface for real-time GPU monitoring
• ComfyUI Manager - Pre-installed for easy custom node management

---

Prerequisites

1. NVIDIA GPU and Driver

You must have an NVIDIA GPU with proper drivers installed:

# Check if NVIDIA driver is installed
nvidia-smi

If the command fails or shows no GPU, install the NVIDIA driver first:

Ubuntu/Debian:

# Check available drivers
ubuntu-drivers devices

# Install recommended driver
sudo ubuntu-drivers autoinstall

# OR install specific version
# sudo apt install nvidia-driver-535

# Reboot after installation
sudo reboot

# Verify installation
nvidia-smi

Manual Installation: Visit NVIDIA Driver Downloads for your specific GPU model.

2. Recommended Cloud Instance

Linode RTX4000 Ada x1 Medium

• GPU: NVIDIA RTX 4000 Ada (20GB VRAM) x1
• CPU: 8 vCPUs
• RAM: 32GB
• Disk: 500TB
• OS: Ubuntu 22.04 or later

This instance provides good performance and cost balance for ComfyUI workflows. The 20GB VRAM can handle most Stable Diffusion models including latest model like Wan2.2 14B.

---

Setup Process

Step 1: Clone or Download the Project

# Clone to any directory you prefer
git clone https://github.com/takashito/linode-comfyui.git
cd linode-comfyui

Or download and extract the files to your preferred location.

Step 2: Install Docker and NVIDIA Container Toolkit

Run the Docker setup script:

sudo bash setup-docker.sh

This script will:

• Install Docker and Docker Compose
• Install NVIDIA Container Toolkit
• Configure Docker to use NVIDIA runtime
• Restart Docker daemon

Verify GPU access:

docker run --rm --gpus all nvidia/cuda:11.8.0-base-ubuntu22.04 nvidia-smi

You should see your GPU information displayed.

Step 3: Configure and Deploy ComfyUI

Run the ComfyUI setup script:

sudo bash setup-comfyui.sh

The script will interactively prompt you for:

1. Domain(s): Your domain name(s) for HTTPS access

   • Example: comfyui.example.com
   • Multiple domains: comfyui.example.com, alt.example.com

2. Username: Basic auth username for web access

   • Example: admin

3. Password: Basic auth password (hidden input)

   • Choose a strong password

The script will then:

• Generate a secure password hash
• Update the Caddyfile configuration
• Create necessary directories
• Build Docker images (this may take 3-5 minutes)
• Start all services
• Configure systemd for auto-start

Step 4: DNS Configuration

Before accessing your deployment, ensure your domain's DNS A record points to your server's IP address:

Type: A
Name: comfyui (or @ for root domain)
Value: YOUR_SERVER_IP
TTL: 3600

Wait for DNS propagation (usually 5-60 minutes).

Accessing ComfyUI

Once setup is complete:

• ComfyUI Interface: https://your-domain.com
• GPU Monitor: https://your-domain.com/monitor

Enter the username and password you configured during setup.

Directory Structure

.
├── README.md                    # This file
├── setup-docker.sh              # Docker & NVIDIA setup script
├── setup-comfyui.sh             # ComfyUI deployment script
├── docker-compose.yml           # Docker Compose configuration
├── Caddyfile                    # Caddy reverse proxy config (auto-generated)
├── Dockerfile.comfyui           # ComfyUI container definition
├── Dockerfile.gpu-monitor       # GPU monitor container definition
├── models/                      # Stable Diffusion models
│   ├── diffusion_models/        # Checkpoint files (.safetensors, .ckpt)
│   ├── vae/                     # VAE models
│   └── text_encoders/           # CLIP and text encoder models
├── input/                       # Input images
├── output/                      # Generated images
├── custom_nodes/                # ComfyUI custom nodes
└── user/                        # User configuration and workflows

---

Managing the Deployment

Service Control

# Check service status
sudo systemctl status comfyui-compose.service

# View logs
sudo journalctl -u comfyui-compose.service -f

# Restart services
sudo systemctl restart comfyui-compose.service

# Stop services
sudo systemctl stop comfyui-compose.service

# Start services
sudo systemctl start comfyui-compose.service

Docker Commands

# View running containers
docker ps

# View logs
docker compose logs -f

# Restart specific service
docker restart comfyui
docker restart caddy
docker restart gpu-monitor

# Access ComfyUI container shell
docker exec -it comfyui bash

# Stop all services
docker compose down

# Start all services
docker compose up -d

Installing Custom Nodes

Method 1: Using ComfyUI Manager (Recommended)

1. Access ComfyUI web interface
2. Click on "Manager" button
3. Browse and install custom nodes
4. Restart ComfyUI container: docker restart comfyui

Method 2: Manual Installation

# Access container
docker exec -it comfyui bash

# Navigate to custom nodes directory
cd /opt/ComfyUI/custom_nodes

# Clone the custom node repository
git clone https://github.com/author/custom-node-repo.git

# Exit container
exit

# Restart ComfyUI
docker restart comfyui

Adding Models

Place your model files in the appropriate directories:

# Stable Diffusion checkpoints
./models/diffusion_models/

# VAE models
./models/vae/

# Text encoders (CLIP, etc.)
./models/text_encoders/

Models are automatically mounted and available in ComfyUI.

---

Support and Resources

• ComfyUI Documentation
• ComfyUI Manager
• Caddy Documentation
• NVIDIA Container Toolkit

---

License

This deployment configuration is provided as-is. ComfyUI and other components have their own licenses.