How to Continuous Delivery

Disclaimer

No exp

Git flow

• complex
• merge hell
• maintainability cost
• protected environment
• QA before releases

Master only

• simple
• (almost) no conflicts
• fit with microservices
• base on automated tests
• crazy? ;)

Master only

• Commit often
• PR for code review and/or WIP
• Continuous Deployment to production

What could possibly go wrong?

Outage

Tips and tricks

Migrations

• atomic changes
• split full change across multiple migrations
• adjust code across multiple commits

http://blog.arkency.com/2014/04/the-biggest-obstacle-to-start-with-continuous-deployment/

No Migrations

• NoSQL
• Easy to scale
• Consider pros and cons

http://martinfowler.com/nosql.html

Complex features

When my team of three rebuilt GitHub Issues in 2014, we had shipped probably hundreds of tiny pull requests to production behind a feature flag that only we could see.

 

https://zachholman.com/posts/deploying-software

Zero Downtime

• two servers running code
• switch traffic to blue only
• deploy to green
• verify & repeat

http://martinfowler.com/bliki/BlueGreenDeployment.html

Staging

• run tests on staging without mocks - copy of production environment with dummy data
• automatically deploy to production

http://martinfowler.com/articles/continuousIntegration.html

Monitoring

• track errors and deployments
• alert when more errors after deployment
• automatically revert when peak is too high

• Continuous Delivery is battle tested
• There are tools
• There are patterns
• There are books and articles