Adopting Actors... an epic tale of loss and learning
summary
Migration from stateless Java to Akka/Scala runs on ~3,000 nodes and grows from 3.5M to ~25M tasks/month, revealing concrete anti‑patterns: a movie star actor that holds queues and agent state causes bottlenecks and split brains; duplicated state between actors requires a single source of truth; passing responsibility between actors demands its own actor. Actors start by stashing until they receive a snapshot, receive() is atomic but cannot nest atomics, and resilience testing uses chaos marmoset to inject failures rather than masking them with long-term try/catch.
AI-written summary · about
chapters (6)
- 0:00Introduction to Cloud Master
- 3:33Migrating to the actor model
- 8:07Anti-pattern: Movie star
- 11:55Anti-pattern: Split brain
- 18:01The problem with robustness
- 20:47Resilience and final lessons
related talks
54mScala concurrency patterns with Actors
01h 32mFrom Spring + Java to Spring + Akka - A Journey of Discovery
33mThe Actor Model, Behind the Scenes
51mUp up and Out: Scaling Software with Akka
37mBeyond Fault Tolerance
48mThe Zen of Akka
50mThe Zen of Akka
40mREST no more - Using Actors for the Internet of LEGO...