Ceph MCP Server

A Model Context Protocol (MCP) server that enables AI assistants to interact with Ceph storage clusters through natural language. This server provides a bridge between AI tools and your Ceph infrastructure, making storage management more accessible and intuitive.

🚀 Features

• Health Monitoring: Get comprehensive cluster health status and diagnostics
• Host Management: Monitor and manage cluster hosts and their services
• Detailed Analysis: Access detailed health checks for troubleshooting
• Secure Communication: Authenticated access to Ceph Manager API
• Structured Responses: AI-friendly output formatting for clear communication
• Async Architecture: Non-blocking operations for better performance

📋 Prerequisites

• Python 3.11 or higher
• UV package manager
• Access to a Ceph cluster with Manager API enabled
• Valid Ceph credentials with appropriate permissions

🛠️ Installation

1. Clone and setup the project:

# Create the project directory
mkdir ceph-mcp-server
cd ceph-mcp-server

# Initialize UV project
uv init --python 3.11

# Add dependencies
uv add mcp httpx pydantic python-dotenv structlog asyncio-mqtt
uv add --dev pytest pytest-asyncio black isort mypy ruff

2. Set up your environment:

# Copy the example environment file
cp .env.example .env

# Edit .env with your Ceph cluster details
nano .env

3. Configure your Ceph connection:

# .env file contents
CEPH_MANAGER_URL=https://192.16.0.31:8443
CEPH_USERNAME=admin
CEPH_PASSWORD=your_ceph_password
CEPH_SSL_VERIFY=false  # Set to true in production with proper certificates

🏃‍♂️ Quick Start

1. Start the MCP server:

uv run python -m ceph_mcp.server

2. Test the connection: The server will log its startup and any connection issues. Look for messages indicating successful connection to your Ceph cluster.

🔧 Configuration

Environment Variables

Security Considerations

• Production Usage: Always enable SSL verification (CEPH_SSL_VERIFY=true) in production
• Credentials: Store credentials securely and never commit them to version control
• Network Access: Ensure the MCP server can reach your Ceph Manager API endpoint
• Permissions: Use a dedicated Ceph user with minimal required permissions

🎯 Available Tools

The MCP server provides four main tools for AI assistants:

1. get_cluster_health

Get comprehensive cluster health status including overall health, warnings, and statistics.

Use cases:

• "How is my Ceph cluster doing?"
• "Are there any storage issues I should know about?"
• "What's the current status of my cluster?"

2. get_host_status

Retrieve information about all hosts in the cluster including online/offline status and service distribution.

Use cases:

• "Which hosts are online in my cluster?"
• "What services are running on each host?"
• "Are any hosts having problems?"

3. get_health_details

Get detailed health check information for troubleshooting specific issues.

Use cases:

• "What specific warnings does my cluster have?"
• "Give me detailed information about cluster errors"
• "Help me troubleshoot this storage issue"

4. get_host_details

Get comprehensive information about a specific host.

Parameters:

• hostname: The name of the host to examine

Use cases:

• "Tell me about host ceph-node-01"
• "What services are running on this specific host?"
• "Get detailed specs for this host"

📊 Example Interactions

Health Check

AI Assistant: "How is my Ceph cluster doing?"

Response: ✅ Cluster is healthy. All 3 hosts are online. OSDs: 12/12 up.
🟢 Overall Status: HEALTH_OK
🖥️  Hosts: 3/3 online
💾 OSDs: 12/12 up

Troubleshooting

AI Assistant: "What warnings does my cluster have?"

Response: 🟡 Cluster has 2 warning(s) requiring attention.
🟡 Warnings requiring attention:
   - OSD_NEARFULL: 1 osd(s) are getting full
   - POOL_BACKFILLFULL: 1 pool(s) are backfill full

🧪 Development

Running Tests

# Run all tests
uv run pytest

# Run with coverage
uv run pytest --cov=ceph_mcp

# Run specific test types
uv run pytest -m "not integration"  # Skip integration tests

Code Quality

# Format code
uv run black src/ tests/
uv run isort src/ tests/

# Lint code
uv run ruff check src/ tests/
uv run mypy src/

# All checks
uv run ruff check src/ tests/ && uv run mypy src/ && uv run pytest

Project Structure

ceph-mcp-server/
├── src/ceph_mcp/
│   ├── __init__.py          # Package initialization
│   ├── server.py            # Main MCP server
│   ├── api/
│   │   └── ceph_client.py   # Ceph API client
│   ├── config/
│   │   └── settings.py      # Configuration management
│   ├── handlers/
│   │   └── health_handlers.py # Request handlers
│   ├── models/
│   │   └── ceph_models.py   # Data models
│   └── utils/               # Utility functions
├── tests/                   # Test suite
├── .env.example            # Environment template
├── pyproject.toml          # Project configuration
└── README.md              # This file

🐛 Troubleshooting

Common Issues

1. Connection Refused

   • Check if Ceph Manager is running and accessible
   • Verify the URL and port in your configuration
   • Ensure network connectivity between MCP server and Ceph cluster

2. Authentication Failed

   • Verify username and password are correct
   • Check that the user has appropriate permissions
   • Ensure the Ceph user account is active

3. SSL Certificate Errors

   • For development: Set CEPH_SSL_VERIFY=false
   • For production: Use proper SSL certificates or specify CEPH_CERT_PATH

4. Permission Denied

   • Ensure the Ceph user has read permissions for health and host information
   • Check Ceph user capabilities: ceph auth get client.your-username

Debugging

Enable debug logging to get more detailed information:

LOG_LEVEL=DEBUG uv run python -m ceph_mcp.server

🤝 Contributing

1. Fork the repository
2. Create a feature branch: git checkout -b feature-name
3. Make your changes and add tests
4. Run the test suite: uv run pytest
5. Format code: uv run black src/ tests/
6. Submit a pull request

📄 License

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

🙏 Acknowledgments

• Ceph Storage - The distributed storage system
• Model Context Protocol - The protocol enabling AI integration
• Anthropic - For developing MCP and Claude

📞 Support

• Create an issue for bug reports or feature requests
• Check existing issues before creating new ones
• Provide detailed information about your environment when reporting issues