事件驱动 CQRS 与事件存储架构

Architecture

CQRS 架构图，结合独立的命令和查询 API 与事件总线进行异步读模型同步，包括从领域事件构建反规范化视图的投影器。该模板演示了完整的 CQRS+ES 技术栈，写入通过领域验证，读取从优化的物化视图提供服务。适合读写工作负载具有根本不同扩展需求的高吞吐量系统。

下载 .fz 文件

完整 FlowZap 代码

WriteAPI { # Write API (Command)
n1: circle label:"POST/PUT Request"
n2: rectangle label:"Command Handler"
n3: rectangle label:"Validate Business Rules"
n4: rectangle label:"Persist to Write DB"
n5: rectangle label:"Publish Domain Event"
n1.handle(right) -> n2.handle(left)
n2.handle(right) -> n3.handle(left)
n3.handle(right) -> n4.handle(left) [label="Valid"]
n4.handle(right) -> n5.handle(left) [label="Committed"]
n5.handle(bottom) -> EventBus.n6.handle(top) [label="OrderPlaced"]
}
EventBus { # Event Bus (Kafka)
n6: rectangle label:"Receive Domain Event"
n7: rectangle label:"Route to Topic"
n8: rectangle label:"Guarantee Delivery"
n9: rectangle label:"Fan-Out to Consumers"
n6.handle(right) -> n7.handle(left)
n7.handle(right) -> n8.handle(left) [label="At-Least-Once"]
n8.handle(right) -> n9.handle(left)
n9.handle(bottom) -> ReadSide.n10.handle(top) [label="Consume"]
n9.handle(bottom) -> Projector.n14.handle(top) [label="Consume"]
}
ReadSide { # Read API (Query)
n10: rectangle label:"Query Handler"
n11: rectangle label:"Read from Read DB"
n12: rectangle label:"Transform to DTO"
n13: circle label:"GET Response"
n10.handle(right) -> n11.handle(left)
n11.handle(right) -> n12.handle(left) [label="Denormalized"]
n12.handle(right) -> n13.handle(left)
}
Projector { # Read Model Projector
n14: rectangle label:"Event Handler"
n15: rectangle label:"Update Read Model"
n16: rectangle label:"Rebuild Materialized View"
n17: diamond label:"Consistency Check"
n18: rectangle label:"Mark Projection Current"
n14.handle(right) -> n15.handle(left) [label="Transform"]
n15.handle(right) -> n16.handle(left) [label="Upsert"]
n16.handle(right) -> n17.handle(left)
n17.handle(right) -> n18.handle(left) [label="Consistent"]
n18.handle(top) -> ReadSide.n11.handle(bottom) [label="Updated"]
}

为什么需要这个工作流？

Systems with high read-to-write ratios suffer when the same data model serves both commands and queries. CQRS with an event store separates the write path (optimized for consistency) from the read path (optimized for speed), allowing each to scale independently while domain events keep them synchronized.

工作原理

Step 1: Write requests go through a command handler that validates business rules.
Step 2: Valid commands are persisted to the write database and a domain event is published.
Step 3: The event bus guarantees at-least-once delivery to all subscribed projectors.
Step 4: Projectors transform domain events into denormalized read models optimized for specific queries.
Step 5: The read API serves queries directly from the optimized read database.
Step 6: Consistency checks monitor lag between write and read models.

替代方案

A single database with read replicas is simpler but limits query optimization. GraphQL can reduce over-fetching without full CQRS. This template shows the complete CQRS+ES architecture for teams that need independent read/write scaling.

Key Facts

Template Name	事件驱动 CQRS 与事件存储架构
Category	Architecture
Steps	6 workflow steps
Format	FlowZap Code (.fz file)