Session description
In the rush to "break the monolith" through GraphQL Federation, organizations can go too far and end up with an explosion of subgraphs. At scale, this can lead to a different kind of pain: high operational overhead, reliability issues, and ultimately, a degraded graph quality. This session explores how to pivot from "splitting" to "merging" without impacting the customer.
The presentation will dive into the workflow developed by Brex to consolidate federated subgraphs safely and reliably, covering:
• The Why: Identifying the tipping point where service fragmentation hurts more than it helps.
• The Strategy: A zero-downtime workflow for merging services covering code migration, schema composition and traffic shifting
• Reliability: How to ensure schema integrity and 0 customer impact during the transition.
• Outcomes: How the consolidation improved our graph design and simplified our overall architecture.
Attendees will leave with a framework for evaluating when federation boundaries are hurting more than helping and a roadmap for executing a "subgraph smash" in their own federated infrastructure
PS: I’m okay with decreasing the scope for a lightning talk if necessary