Messaging: The fine line between awesome and awful
summary
A team replaces a retail monolith with messaging and finds throughput and UI consistency get worse, not better: request/response over queues adds hops, at least once delivery creates inevitable duplicates, and ordering failures occur roughly 3% of the time even if 97% look fine. Practical remedies: use delayed messages for SLA enforcement (Azure Service Bus, SQS, RabbitMQ delay), add OpenTelemetry for traces, make handlers idempotent, and use the outbox pattern or a Saga/orchestrator for multi-step consistency.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 6:28Monolith to distributed mess
- 11:25Messaging patterns
- 29:18Decoupling services
- 36:00UI inconsistency in async
- 45:54Out-of-order messages
- 47:00Troubleshooting failures
- 51:34Duplicate messages and idempotency
- 1:00:30Summary
related talks
53mMessaging: The fine line between awesome and awful
01h 01mThe Unsung Hero of Modern Software: Asynchronous Messaging
01h 02mMessaging Patterns for Modern Software Solutions
01h 01mBeyond Pub/Sub - Advanced Messaging Patterns
01h 03mChoosing the Right Messaging Solution for Your Architecture
41mForget about HTTP
01h 03mBeyond Pub/Sub - Advanced Messaging Patterns
38mMessaging Patterns to Transform Your Cloud Architecture