Building a better bug trap with GitHub Actions

Manning Twitch stream Jan 12, 2022

Learn how to automate one of the two pieces of Continuous Delivery: always keeping your codebase in a releasable state. We’ll use GitHub actions to explore the lifecycle of a change, discover all the places where bugs can squirm their way in, and learn the best times to run your pipelines for the most effective bug-catching!

Overall goal: Always be in a releasable state (one of the 2 pieces of CD)

Go from a repo with no automation to a repo with:

• PR based automation
• Post-merge PR based automation
• Periodic automation

Setup automation to:

• Run unit tests

Demonstrate why you need:

• PR based automation - show getting the repo into a broken state
• Post-merge based - show example of conflicting changes making main broken
• Periodic automation - can highlight conflicting changes and catch flakes

What's still missing? (could explore if time)

• Demonstrating flakes being caught
• Build image & run system test
• Release automation (e.g. tag based)
• Linting

Outline

1. Repo with a service that gets rates, includes dockerfile
2. Show a PR that breaks the repo (1_bad_pr)
3. Add github action to run tests (2_add_pr_tests)
4. Show how this fixes the original PR problem
5. Show 2 conflicting PRs and how a bug isn’t caught (3_bad_pr_1, 3_bad_pr_2)
6. Add periodic action to highlight issue (4_periodic)
7. Undo conflicting PRs
8. Open PRs again
9. Show requiring branch to be up to date (docs)
10. Explain downsides (many PRs = many updates)
11. Undo requiring branch to be up to date
12. Show post merge action (5_post_merge)
13. Describe merge queues / merge trains (GitHub merge queue docs)

Building and running the image

docker build --tag coin-rates .
docker run -p 80:5000 coin-rates

Accessing in browser:

• http://localhost/?coin_type=catcoin
• http://localhost/?coin_type=dogcoin