Continuous Improvement
summary
Built Etsy's CI that supports continuous delivery with 96 deployments in one day and uses feature flags so code lands dark then flips on; production monitoring aggregates about quarter million metrics so teams surface problems via dashboards and catch a major outage in 20 minutes and subtle micro‑outages within hours. QA shifts from end‑of‑line checking to exploratory testing, close to the code, pairing with ops and devs; organizational change needs engineering‑level oversight to make automated checks, monitoring and production testing effective.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:43The Canonical Agile Release Cycle
- 1:43The Continuous Delivery Environment
- 2:53How Continous Delivery Differs from "Canonical" Agile Process
- 5:26Observed Behavior Of Complex Systems
- 10:39Falsehoods About Quality Assurance
- 17:23Leveson's 3 Practices For Limiting Risk
- 19:00Characteristics of Resilient Software
- 22:06Exploratory Testing
- 26:26Anomaly detection is hard.
- 27:42Keep the feedback loop short.
- 30:35Questions?
related talks
38mBuilding a high quality product
33mSoftware development practices for companies that want to make an impact
24mEtsy
25mDevOps Handbook Experiments in Accelerating Delivery
12mConvergent Quality Engineering
30mRelease Early and Release Often: Reducing deployment friction
34mContinuous Integration: A Bittersweet Love Story
55mContinuous Deployment Workflows for Modern JavaScript
