GraphQL Subscriptions in Production Is Easy Isnt It?
summary
Describes moving GraphQL subscriptions from localhost to production with concrete choices and gotchas: uses graph over sse for an SSO log-streaming feature so cookie auth and firewalls work, explains one subscription = one HTTP connection (HTTP/2 helps), benchmarks design tradeoffs versus graph over websocket for lower latency, uses redis pubsub as the cross-instance event source, experiments with shared web workers to avoid duplicate tabs, and relies on randomized exponential backoff for reconnect storms; recommends polling (30s–60s) when realtime isn’t required.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:38Understanding subscriptions
- 1:27Transport protocol overview
- 2:33Websockets versus SSE
- 5:14Benefits of SSE
- 6:30Choosing the right protocol
- 7:04Practical case study
- 8:49Backend implementation
- 9:48Scaling challenges
- 10:54Operational considerations
- 12:28When not to use subscriptions
- 13:47Conclusion
related talks
34mGraphQL Subscriptions - Scaling real-time in the real world
34mGraphQL real-time subscriptions
30mBest Practices for GraphQL and GraphQL Subscriptions at Scale
20mScaling GraphQL on AWS: Production Architecture for High-Volume Data Systems
18mCrafting compelling real-time web experiences with GraphQL
01h 04mBuilding real-time API's with GraphQL
39mScaling event-driven systems with Node and GraphQL subscriptions
58mStreaming Subscriptions in GraphQL