Event-Driven Architecture Pocket Guide. Saga, CQRS, Outbox, and the Traps Nobody Warns You About
Gabriel Anhaia
Synopsis "Event-Driven Architecture Pocket Guide. Saga, CQRS, Outbox, and the Traps Nobody Warns You About"
Look at any business workflow and decide - on purpose - whether to model it request-response or event-driven, then pick the right patterns (outbox, saga, CQRS, event sourcing) without over-engineering or under-engineering.
You shipped request-response services for years. Now a product manager wants an event-driven one - or you inherited one already in production and it is groaning under its own complexity. Dual writes are losing messages. A saga compensates halfway and leaves orphaned state. Consumers break every time the event schema changes. Somebody added Kafka because "it scales," and now nobody can trace a request end to end. You need the patterns, the trade-offs, and the anti-patterns spelled out - fast, without wading through 600 pages of abstract theory.
Event-Driven Architecture Pocket Guide is the working engineer's reference for building event-driven systems that do not collapse under their own weight. Across 15 chapters and five parts, it covers the whole landscape as of 2026 - Kafka 4.0, Redpanda, Pulsar 4.1, NATS JetStream, RabbitMQ quorum queues, Debezium 3.x, Confluent and Buf schema registries, Temporal, AWS Step Functions - and shows exactly when each pattern earns its complexity budget and when it becomes the wrong tool.
What you will take away: • Commands vs events - the subtle tense difference that determines whether your system scales or calcifies, and where architectural drift in event-driven systems actually starts. • The outbox pattern done right - why dual writes lose data, how to implement the production-proven fix on Postgres / MySQL, and the "we tried two-phase commit" cautionary tale. • Saga, CQRS, event sourcing - orchestration vs choreography, when to split reads from writes, when the log IS the database, and the three common traps for each. • Broker trade-offs on six axes - throughput, latency, ordering, retention, cost, and operational burden - for Kafka, Pulsar, NATS, RabbitMQ, and Redpanda. • Twelve real system designs - e-commerce orders, Stripe-style payments, IoT telemetry, ride-share dispatch, video encoding, fraud detection, inventory, notifications, and more - each walked from requirements to chosen stack to exit criteria. • Anti-patterns that wreck production - events used as commands, the event-driven monolith, missing contract tests, no backpressure strategy, message-bus-as-database - with the recovery paths that actually work.
Primary code examples are in Java 21 (Spring Boot / Micronaut) because Kafka + Spring is the dominant production stack. Python 3.13 and Go 1.24 snippets carry the lighter examples. Schema samples cover Avro, JSON Schema, and Protobuf. Every version is locked to April 2026 and web-verified in the fact-check pass.
Who this book is for: backend and platform engineers being asked to build event-driven systems, tech leads choosing between EDA and request-response for a new system, and anyone who inherited an event-driven system and needs to keep it running without rewriting it. Baseline assumed: you have shipped at least one production service; you know SQL, HTTP, and what a queue is.
Other books in Pocket Guides for Developers (standalone, no reading order): • System Design Fundamentals • System Design Interviews • AI Agents Pocket Guide • Prompt Engineering Pocket Guide • Database Playbook • LLM Observability Pocket Guide • This book - Event-Driven Architecture Pocket Guide • RAG Pocket Guide
Companion handbook: Observability for LLM Applications (The AI Engineer's Library, Book 1).
