Journey to Automated Releases
summary
Netflix UI team describes a concrete pipeline: split deployment builds per-app clusters and routes via keymaster so apps deploy independently, plus mandatory PR checks, auto-merge, and a localized-string blocker. They implement a Release ROBOT that tags/promotes builds, computes a reliability test score from reruns and known-issue integration, and automates Kennedy canaries using percentage variation baselines. Result: rollout cadence rises from weekly to three times a week with 90% automation and fewer rollbacks.
AI-written summary · about
chapters (10)
- 0:00Start
- 2:48Team Structure
- 4:35Split Deployment
- 8:57Confidence and Productivity
- 10:35Commit to Getting into the Release Branch
- 11:49Pre-Checking for String Localization
- 14:29Evaluating the Build Status
- 16:16Third Phase of the Release Cycle
- 22:17Promoting Changes To Release Branch
- 23:26Takeaways
related talks
54mNetflix’s Journey to Confident Automated Changes
45mNetflix’s Secret to Safe Automation at Scale
29mOur Journey to Fully Automated Weekly Releases
41mThe Pipeline-Driven Organization
29mFast, Safe, and Continuous -- Picking all 3 at Netflix
27mManaging OS Release Transitions at Netflix Scale
05mRelease management as a catalyst for CI/CD
58mBuilding a Culture of Continuous Delivery