Continuous delivery
of a Python project

with Docker and Ansible

Joost Cassee

PyCon Sette, 16 April 2016

Live slides

http://cass.ee/pyconsette2

 

Code

https://github.com/jcassee/registronavale

Painless Software

django-analytical

 

django-geckoboard

 

django-localeurl

 

angular-hypermedia

joost@cassee.net

@jcassee

jcassee.com

• Develop code
• Fix bugs
   
• Make release artifact
• Fix mistakes in release artifact
   
• QA testing
• Fix bugs
• Make new release
• Deploy to acceptance environment
• Fix mistakes in acceptance environment changes
   
• Acceptance testing
• Fix bugs, make new release, deploy
   
• Deploy to production environment
• Fix mistakes in production environment changes

Development

Deployment

Treat infrastructure as code

• Use Ansible to describe the infrastructure
   
• Keep the description in a Git repository
   
• Host the repository on GitHub
   
• Use Travis CI to build the infrastructure
   
• Require GitHub pull requests for changes

Any pull request is

• deployed, for review by non-technical stakeholders
   
• merged when all stakeholders accept the change

• developers
• testers
• operations
• application managers
• user / customer support

Why Docker?

• Move high-risk operations into the build phase
   
• Make build process CI-independent

Why Ansible?