Reconciling Everything
summary
Explains reconciliation loops as an alternative to message queues for distributed systems: stateless workers read and incrementally update a central datastore rather than passing messages. Compares reconciliation loops to queues and other messaging approaches, then examines real-world behavior, benefits, and pitfalls from using reconciliation loops in the Takahē ActivityPub/Fediverse server.
AI-written summary · about
chapters (7)
- 0:00Start
- 4:56RPC Microservices
- 6:34Event-Driven (Message Passing)
- 10:52Reconciliation (Control) Loop
- 12:28Kubernetes Example
- 18:19Making a single Fediverse post
- 23:07Airflow Example
related talks
47mEverything Loosely Coupled - Advanced Integration Patterns and Their Trade-Offs
01h 02mAt Least Once
59mSucceeding at Reactive Architecture
42mIt's like 10,000 streams when what you need is a queue
46mLost in transaction? Strategies to deal with (in-)consistency in distributed systems
48mCo-ordinating data changes in distributed systems
57mEvent Driven Collaboration
36mLost In Transaction