Dependency Updates: Before and After
summary
Describes a concrete platform that automates dependency updates across 5,000 repos and 3,000 JVM services: a Nebula custom Gradle distribution bundles versioned JSON-based resolution rules (substitute/replacement/alignment) applied at resolve time; an Astrid dependency index maps every artifact version in production; a Rocket-powered nightly update pipeline plus targeted campaigns open PRs and auto-merge on green tests. Results include ~70% Java 21 adoption, 60% daily deployments, over 80% of engineers comfortable with automated changes, and zero-touch deployment of dependency versions in hours across thousands of repositories.
AI-written summary · about
chapters (8)
- 0:00Introduction to dependency management at Netflix
- 4:36Build opinions and custom Gradle distribution
- 18:52Artifact observability and dependency tracking with Astrid
- 23:41Distributed refactoring and automated dependency updates
- 27:51Continuous delivery and building deployment confidence
- 30:21Testing and validation strategies for engineering teams
- 36:57Future work and improving artifact observability
- 40:12Key takeaways and summary
related talks
36mThe future of dependency management, seen from 2022
41mThe future of dependency management, seen from 2022
37mKeep Your Dependencies in Check
45mNetflix’s Secret to Safe Automation at Scale
38mKnow thy neighbours: dependency management done right.
28mTowards Automated Updates of Software Dependencies
31mKeep your dependencies in check
29mDependency Management: Where the Fork are We?