Content ¶

The author shares experience with Docker since its early announcement, offering guidance to new users. The seven key recommendations are:

1. Dense packing vs. microservers ¶

Docker excels at consolidating many isolated services on single servers, though configuration management tools like Ansible may suit other deployment approaches.

2. Architecture as appliances ¶

Avoid treating containers like virtual machines. Follow Unix philosophy by creating single-purpose appliances (PostGIS, uWSGI, MapServer, GeoServer each in separate containers) rather than bundling multiple services.

3. Stateless containers ¶

Design containers for destruction and redeployment without data loss. Invest in robust images rather than individual containers. Mount databases and uploaded files as host volumes; store no generated data internally.

4. Use orchestration tools ¶

Leverage Docker Compose (formerly Fig) with YAML configuration files instead of building custom orchestration. Combine with Makefiles for simplified administration.

5. Image publishing and versioning ¶

Publish images to hub.docker.com with automated builds triggered by repository pushes and upstream updates. Tag versions for tracking stable configurations.

6. Layered image construction ¶

Build progressively (Ubuntu base → add Python → add Django) to minimize configuration per image and maximize layer sharing across services.

7. Multi-host architecture planning ¶

Plan carefully for multi-node services, as Docker lacks robust built-in networking. Consider virtual switches or independent host operation.