Saga 分布式事务架构

Architecture

分布式事务架构图，实现两阶段提交协议，事务协调器向参与服务发送准备和提交消息，任何投票失败时执行全局中止。该模板可视化经典的 2PC 协议，用于多个服务间需要强一致性的场景，展示准备、投票和提交/中止阶段。对于理解微服务架构中分布式共识权衡至关重要。

下载 .fz 文件

完整 FlowZap 代码

Coordinator { # Transaction Coordinator
n1: circle label:"Begin Distributed Transaction"
n2: rectangle label:"Assign Transaction ID"
n3: rectangle label:"Send Prepare to All"
n4: rectangle label:"Collect Votes"
n5: diamond label:"All Voted Commit?"
n6: rectangle label:"Send Global Commit"
n7: rectangle label:"Send Global Abort"
n8: circle label:"Transaction Complete"
n1.handle(right) -> n2.handle(left)
n2.handle(right) -> n3.handle(left) [label="TxID"]
n3.handle(bottom) -> ServiceA.n9.handle(top) [label="Prepare"]
n3.handle(bottom) -> ServiceB.n12.handle(top) [label="Prepare"]
n3.handle(bottom) -> ServiceC.n15.handle(top) [label="Prepare"]
n4.handle(right) -> n5.handle(left)
n5.handle(right) -> n6.handle(left) [label="Yes"]
n5.handle(bottom) -> n7.handle(top) [label="No"]
n6.handle(right) -> n8.handle(left)
n7.handle(right) -> n8.handle(left) [label="Rolled Back"]
n6.handle(bottom) -> ServiceA.n11.handle(top) [label="Commit"]
n6.handle(bottom) -> ServiceB.n14.handle(top) [label="Commit"]
n6.handle(bottom) -> ServiceC.n17.handle(top) [label="Commit"]
}
ServiceA { # Booking Service
n9: rectangle label:"Lock Resources"
n10: rectangle label:"Vote Commit"
n11: rectangle label:"Apply Changes"
n9.handle(right) -> n10.handle(left) [label="Prepared"]
n10.handle(top) -> Coordinator.n4.handle(bottom) [label="Vote"]
}
ServiceB { # Payment Service
n12: rectangle label:"Hold Funds"
n13: rectangle label:"Vote Commit"
n14: rectangle label:"Capture Payment"
n12.handle(right) -> n13.handle(left) [label="Prepared"]
n13.handle(top) -> Coordinator.n4.handle(bottom) [label="Vote"]
}
ServiceC { # Notification Service
n15: rectangle label:"Queue Notification"
n16: rectangle label:"Vote Commit"
n17: rectangle label:"Send Notification"
n15.handle(right) -> n16.handle(left) [label="Prepared"]
n16.handle(top) -> Coordinator.n4.handle(bottom) [label="Vote"]
}

为什么需要这个工作流？

When strong consistency is absolutely required across multiple services (e.g., financial transactions), eventual consistency is not acceptable. The two-phase commit protocol ensures all participants either commit or abort together, providing ACID-like guarantees across distributed services at the cost of availability during the prepare phase.

工作原理

Step 1: The transaction coordinator assigns a unique transaction ID and sends prepare messages to all participants.
Step 2: Each participant locks its resources and votes commit or abort.
Step 3: The coordinator collects all votes and makes a global decision.
Step 4: If all participants voted commit, a global commit message is sent.
Step 5: If any participant voted abort, a global abort message is sent to all.
Step 6: Participants apply or rollback changes based on the global decision.

替代方案

Saga patterns provide better availability but only eventual consistency. Consensus protocols like Raft/Paxos are used for state machine replication. This template shows the classic 2PC protocol for understanding distributed consensus trade-offs.

Key Facts

Template Name	Saga 分布式事务架构
Category	Architecture
Steps	6 workflow steps
Format	FlowZap Code (.fz file)