The Saga Pattern

What Is the Saga Pattern?

A saga consists of multiple local transactions, each executed by a different service. After each completes, an event or message triggers the next. If any step fails, previously-completed steps are compensated via rollback actions, rather than undone via centralized rollback.

This approach suits systems with:

• Independent databases per service
• Long-running workflows
• Need for eventual consistency without service-wide locking

Saga Coordination Styles

1. Choreography (Event-Driven)
   Each service listens for events (e.g. “OrderCreated”) and autonomously acts, emitting the next event.
   Pros: Decoupled, minimal central coordination
   Cons: Harder to trace and debug complex flows

2. Orchestration (Coordinator)
   A central orchestrator directs each step, sending commands and invoking compensations if needed (e.g., “reserveCredit”, “refundPayment”).

Pros: Easier monitoring, more control over flow
Cons: Orchestrator becomes complex, single point to scale

Benefits & Tradeoffs

Benefits

• Avoids distributed transactions
• Service autonomy & resilience
• High availability & eventual consistency

Tradeoffs

• Must design compensating logic manually
• Saga complexity increases with steps
• Non-isolation can lead to anomalies

Key Concepts & Patterns

• Compensating Transactions: Undo operations—e.g. refund payment, restock items
• Pivot Step: A point of no return after which steps can’t be compensated
• Saga Log / State Machine: Tracks progress, supports retries and recovery after crashes
• Idempotency: Essential for handlers to avoid double execution during retries

Implementation Tips

• Prefer orchestration for complex sagas, choreography for simple flows
• Use asynchronous messaging (Kafka, RabbitMQ) for decoupled flows
• Ensure atomic write-and-publish, via outbox or transaction logs
• Implement saga recovery: persisting state allows safe retries

Real-world Use Cases

• E-commerce workflows (order → pay → ship)
• Trip planning (flight, hotel, car rental), cancel all if one fails
• Financial processes (loan approval, disbursement) with rollback

Tools & Frameworks

AWS Step Functions: Serverless saga orchestration

By using the Saga Pattern, you embrace resilience, scalability, and eventual consistency across your microservices—managing complex distributed workflows without locking to a single global transaction.