Pattern: Saga

Problem

How do implement transactions that span services?

Forces

• 2PC is not an option

Solution

Implementing each business transaction that spans multiple services is a saga. A saga is a sequence of local transactions.

Each local transaction updates the database and publishes a message or event to trigger the next local transaction in the saga. 

If a local transaction fails because it violates a business rule then the saga executes a series of compensating transactions that undo the changes that were made by the preceding local transactions.

There are two ways of coordinating sagas:

• Choreography - each local transaction publishes domain events that trigger local transactions in other services
• Orchestration - an orchestrator (object) tells the participants what local transactions to execute

Example: Choreography-based saga

An e-commerce application that uses this approach would create an order using a choreography-based saga that consists of the following steps:

1. The Order Service receives the POST /orders request and creates an Order in a PENDING state

1. It then emits an Order Created event

1. The Customer Service’s event handler attempts to reserve credit

1. It then emits an event indicating the outcome

1. The OrderService’s event handler either approves or rejects the Order

Example: Orchestration-based saga

An e-commerce application that uses this approach would create an order using an orchestration-based saga that consists of the following steps:

1. The Order Service receives the POST /orders request and creates the Create Order saga orchestrator

1. The saga orchestrator creates an Order in the PENDING state

1. It then sends a Reserve Credit command to the Customer Service

1. The Customer Service attempts to reserve credit

1. It then sends back a reply message indicating the outcome

1. The saga orchestrator either approves or rejects the Order

The following illustration shows how the saga pattern implements an order processing system by using AWS Step Functions. 

Each step (for example, “ProcessPayment”) also has separate steps to handle the success (for example, "UpdateCustomerAccount") or failure (for example, "SetOrderFailure") of the process.

Resulting context

This pattern has the following benefits:

• It enables an application to maintain data consistency across multiple services without using distributed transactions

• The programming model is more complex. For example, a developer must design compensating transactions that explicitly undo changes made earlier in a saga.

• The service sends back a response once the saga completes, e.g. once it receives an OrderApproved or OrderRejected event.

• The service sends back a response (e.g. containing the orderID) after initiating the saga and the client periodically polls (e.g. GET /orders/{orderID}) to determine the outcome

• The service sends back a response (e.g. containing the orderID) after initiating the saga, and then sends an event (e.g. WebSocket, webhook, etc) to the client once the saga completes.

This solution has the following drawbacks:

There are also the following issues to address:

• In order to be reliable, a service must atomically update its database and publish a message/event. It cannot use the traditional mechanism of a distributed transaction that spans the database and the message broker. Instead, it must use one of the patterns listed below.

• A client that initiates the saga, which an asynchronous flow, using a synchronous request (e.g. HTTP POST /orders) needs to be able to determine its outcome. There are several options, each with different trade-offs: