Skip to content

Is Cross-Service Consistency Guaranteed—or Just Hoped For?

Type: Structure, DeepDive Category: Async
Audience: Engineers working on service integration, microservice orchestration, or eventual consistency design

🔍 What This Perspective Covers

When one service updates state—
how do you ensure the others agree?

  • Event published ≠ event received
  • Downstream updates might arrive late, or not at all
  • State might “look” right, but be structurally divergent

⚠️ Failure Scenarios

  • Event loss: system A updates, B never sees the change
  • Out-of-order delivery → B overwrites A’s latest data
  • Conflict resolution policies undefined → inconsistent recovery
  • No alerting for downstream update drift

âś… Design for Real Consistency

  • Use event versioning and schema contracts
  • Track causal links: what triggered what?
  • Implement reconciliation jobs for long-term consistency
  • Prefer eventual over “pretend real-time” sync
  • Monitor consistency lag across services—not just delivery

⚠️ Key Shift

Don’t assume the system is consistent.
Design it to notice when it isn’t.

âť“ FAQ

  • Q: But we use event delivery guarantees. Isn’t that enough?
    A: Guarantees cover delivery, not correctness.

  • Q: Can’t we just trust retries?
    A: Not without idempotence and visibility.