A Story of Real Implementation of Shift Left
summary
Concrete claim: every pull request deploys to a temporary Kubernetes stack pre-merge so full automation runs and returns feedback under 30 minutes. Implementation uses EKS, GitHub Actions, Helm and Docker with local testing via Kind; tests include unit, Cypress and Lighthouse. At scale five apps with ~7,000 PRs show >4 hours saved per PR (about 14 engineer‑years) and fewer production incidents versus post‑merge testing. Stacks destroy on PR close and concurrent PRs run independently.
AI-written summary · about
chapters (12)
- 0:00Shift Left implementation
- 1:28Why shift left is critical
- 4:47Problems with CI/CD pipelines
- 6:58Concurrent PR execution with Kubernetes
- 9:14Main branch always production ready
- 11:45GitHub Actions PR validation
- 13:13Automating tests with GitHub triggers
- 15:26Reliable and dependable tests
- 16:52Measuring productivity gains
- 18:18Cost optimization in PR deployments
- 19:23Laptops aren't scalable for testing
- 20:07Q&A
related talks
16mShift Left, Not Off a Cliff: Why Your Testing Strategy Is Breaking Developers
46mTestOps: A Journey to Story-Based Releases
32mContinuous Integration and Deployment: Five Lessons Learned
36mSIG Testing Deep Dive
53mShift Left Testing for Continuous Collaboration
36mIntro: SIG Testing
34mAre You About to Break Prod? Acceptance Testing with Ephemeral Environments
57mAsync Code Reviews Are Choking Your Company’s Throughput