Deployment Overview of CogVideoX-5b on Server¶

Prerequisites and Basic Requirements¶

To ensure the successful deployment and operation of the application, the following system requirements must be met:

Operating System: Ubuntu 22.04 (recommended for HWE kernel support).
Privileges: Root or sudo access is required for all installation steps.
Hardware Requirement: An NVIDIA GPU is mandatory for CUDA acceleration.
Network Ports:
443 (HTTPS) - External access via Nginx proxy.
7860 - Internal application port.

FQDN of the final panel on the hostkey.in domain¶

The application is accessible through a specific subdomain template based on the server ID.

Parameter	Value
Prefix	`cogvideo`
Domain	`hostkey.in`
Full template	`cogvideo{Server_ID_from_Invapi}.hostkey.in`

File and Directory Structure¶

The following directories are used for application data, configurations, and system services:

/opt/CogVideo: Main application directory containing the source code and Python virtual environment.
/opt/CogVideo/env: Python virtual environment.
/root/nginx: Configuration directory for the Nginx proxy container.
/data/nginx/user_conf.d: Directory containing Nginx site configurations.
/data/nginx/nginx-certbot.env: Environment file for Certbot configuration.

Application installation process¶

The application is installed using a combination of system-level package management, manual script execution, and service configuration:

System Preparation: The system is updated and upgraded via apt. The linux-generic-hwe-22.04 kernel and gcc compiler are installed to support specialized drivers.
NVIDIA Driver & CUDA Setup:
A CUDA keyring is added to the package manager.
cuda-toolkit and nvidia-cuda-toolkit are installed.
The system environment variables (PATH and LD_LIBRARY_PATH) are updated in .bashrc.
NVIDIA drivers are initialized, and nvidia-docker2 is installed to enable GPU support within Docker containers.
Application Source: The CogVideo repository is cloned from GitHub into /opt/CogVideo.
Python Environment Setup:
Python 3.10 and venv modules are installed.
A virtual environment is created at /opt/CogVideo/env.
Required dependencies are installed via pip, including specific versions of torch, torchvision, and torchaudio compiled for CUDA 11.8, as well as moviepy.
Service Configuration: A systemd service named cogvideo.service is created to manage the application process using the Gradio composite demo script.

Access Rights and Security¶

Firewall/Access: The application is exposed via an Nginx reverse proxy on port 443.
User Permissions: The CogVideo service runs under the root user within its working directory /opt/CogVideo.
Docker Security: NVIDIA container runtime is configured to allow the application to interface with host GPU hardware securely.

Docker Containers and Their Deployment¶

The deployment utilizes a single Nginx container managed via Docker Compose for SSL termination and reverse proxying.

Container Name	Image	Ports (Host:Container)	Volumes	Environment Variables	Restart Policy
`nginx`	`jonasal/nginx-certbot:latest`	`host_network`	`- nginx_secrets:/etc/letsencrypt` `- /data/nginx/user_conf.d:/etc/nginx/user_conf.d`	`[email protected]` (plus custom env file)	`unless-stopped`

Custom Scripts and Additional Setup¶

The installation process includes an automated initialization script (install_script.sh) located in /tmp during the setup phase. This script performs the following critical actions:

Updates system repositories.
Installs necessary build tools (GCC).
Configures NVIDIA CUDA repositories and toolkits.
Modifies shell configuration files to set up the CUDA environment path.
Removes the nouveau driver to allow the NVIDIA proprietary driver to load.

Application Update Instructions¶

To update the main application, follow these steps:

Update Source Code: Navigate to /opt/CogVideo and pull the latest changes from the repository.

Update Dependencies: Activate the virtual environment and upgrade packages:

source /opt/CogVideo/env/bin/activate
pip install -r requirements.txt

Restart Service: Apply changes by restarting the systemd service:
```
systemctl restart cogvideo
```

Location of configuration files and data¶

Nginx Configuration: /data/nginx/user_conf.d/{prefix}{server_id}.hostkey.in.conf
Docker Compose File: /root/nginx/compose.yml
Application Source: /opt/CogVideo

Available ports for connection¶

HTTPS (External): 443
Internal Application Port: 7860

Starting and Stopping the application¶

The application is managed as a systemd service:

Start the application: systemctl start cogvideo
Stop the application: systemctl stop cogvideo
Check status: systemctl status cogvideo
Enable on boot: systemctl enable cogvideo

Proxy Servers¶

The system uses an Nginx container (jonasal/nginx-certbot) acting as a reverse proxy. It handles SSL certificates via Certbot and routes incoming traffic from port 443 to the internal application running on http://127.0.0.1:7860.