Rolling-out like a Rock-Star
summary
Spotify builds an in-house remote configuration system (replacing ABBA) that delivers over a thousand feature flags daily across Android iOS web and C++ shared layers. Developers define typed flags in properties.yaml, generate safe immutable Java/C++ classes via code generation, and publish from CI to a remote-config admin; runtime uses an activate and fetch model to ensure consistent UX but accepts propagation delays of minutes to days. The system scales beyond Firebase limits and is live with company-wide adoption.
AI-written summary · about
chapters (16)
- 0:00Introduction and speaker intro
- 1:17History of ABBA framework
- 4:12Disclaimers and scope
- 7:04Design principles safety ubiquity simplicity
- 10:18Architecture overview actors
- 12:26Backstage and data scientist role
- 13:30Developer build phase properties
- 17:28Code generation and developer setup
- 19:58Publish phase and CI checks
- 23:14Resolve phase runtime SDK
- 24:39Fetch strategy activate and fetch
- 27:51Lessons learned correctness
- 30:40Experiments vs rollouts differences
- 32:05Propagation time and mitigation
- 35:02Buy vs build decision
- 37:37Q&A session
related talks
31mRolling-out like a Rock-Star
51mWorkshop: How to Use Powerful Feature Flags and Controlled Rollouts
27mScaling your mobile app release process
59mFaster & less risky releases with feature flags
01h 15mExperimentation at scale: Fuel your application’s growth through A/B testing
25mReleasing apps faster with feature flags
32mMaster your software delivery with feature flags
13mHow to ship stable features with Firebase