Change is inevitable: versioning event-driven systems
summary
Shows a concrete pattern: split events into private granular events and stable public summary events to avoid functional coupling, use Avro/JSON schemas and a registry to enforce forward/backward/full/transitive compatibility, and perform breaking upgrades with dual publishing and a deprecation window (example given: 3–6 months). Demonstrates tooling: CloudEvents for headers, Azure Service Bus subscription-filter trick to drain in-flight messages, and the emerging XR Registry (RC4, runnable server, Docker) to validate on egress and emit change notifications.
AI-written summary · about
chapters (9)
- 0:00Introduction and welcome
- 3:04Change is inevitable in systems
- 12:57Designing better event types
- 23:18Public vs private events
- 30:20Schema versioning strategies
- 40:42Message headers and CloudEvents
- 46:00Breaking changes and upgrade strategy
- 51:46Schema registries and XRegistry
- 1:00:39Key takeaways and conclusion
