Unhappy Path & Dealing with Bad Events
summary
Lessons from putting event systems into 24x7 insurance and wealth platforms: design specific events with an envelope that includes correlation id and enough reference data to replay, avoid single-threaded consumers (they hit scalability and availability limits), partition by entity ID to retain ordering, use dead letter queue plus a quarantine of later events to prevent out-of-order processing, implement back pressure (503s, throttling), monitor DLQ depth, and log traces so a mass email-bounce bug affecting millions can be detected and replayed.
AI-written summary · about
chapters (4)
related talks
31mDesigning a mission critical event system
55mSo You Want to Maintain a Reliable Event Driven System
55mAvoiding mistakes with events, one event at a time
52mSo You Want to Build An Event Driven System?
49mSo You Want to Maintain a Reliable Event-Driven System
55mAn Introduction to Enterprise Design Events
47mTop 10 Event-Driven Architecture Pitfalls
49mAvoiding mistakes with events, one at a time