Beyond the Linear Trap: Scaling CI/CD with State-on-Demand
summary
Modern CI/CD stalls on linear end-to-end flows; claim: treating state as first-class lets tests start from any business state, cutting validation from ~100 minutes to ~1 minute and provisioning in microseconds. Implementation combines test data APIs, a central gateway with buffering and retries, orchestration (think Temporal/persistent workflows) and environment virtualization (snapshots/ layered storage) plus UI-automation fallbacks. Proven in banks and at Rohlik.cz with progressive adoption: generators per domain, a catalog for discovery, and observable buffering that makes parallelized end-to-end testing practical.
AI-written summary · about
chapters (11)
- 0:00CI/CD and testing challenges
- 4:35Personal background and community involvement
- 7:42Legacy test data problems
- 10:29On-demand test data need
- 13:59Test data API architecture
- 18:31Implementing progressive adoption
- 22:00Q&A on buffering and composition
- 27:16Security in test data APIs
- 30:23Overcoming team resistance
- 33:19Low testing priority workarounds
- 34:43Summary
related talks
40mSustainable Continuous Delivery: Patterns for rapid and reliable Releases
43mEffective CI/CD for Large Systems
25mCI/CD Delays: Losing Time, Money & Sanity
19mCI/CD Is Dead, Agents Need Continuous Compute and Computers
41mAll Things Continuous
42mTesting in a Continuous Delivery World
32mEvent-based Continuous Delivery: The End of the Monolithic Pipeline
24mOrchestrating automation tests for Continuous testing