Distributed Transactions with Saga Pattern

Prerequisites

To understand this topic fully, familiarity with transactions and microservices is recommended.

Here are some of my previous blog posts that will be useful:

• Maintain Database Consistency With Transactions
• Microservice Orchestration vs. Choreography

Introduction

In distributed systems, workflows often span multiple microservices.

Let’s consider the example of a travel application that I have used in my previous blog posts. This application is responsible for coordinating between an airline service, a hotel service, and a car rental service. This allows users to book all essential services for their vacation in one place.

Example Workflow

1. Reserve airline tickets
2. Reserve a hotel
3. Reserve a rental car

In a monolithic application, with all data in a single database, achieving a transaction is straightforward because everything is handled within the same database context.

However, in a distributed system, each service (airline, hotel, car rental) operates independently, with its own database. Ensuring that all services either succeed or fail as a unit becomes much more challenging!

Problem

To illustrate this concern, imagine the following scenario:

1. ✔️ Airline reservation is successful.

2. ✔️ Hotel reservation is successful.

3. ❌ Rental car reservation fails.

Now, the user is left with partial reservations, an inconsistent state where only some services are booked, leading to a poor user experience.

Solution: The Saga Pattern

The Saga pattern helps by treating the entire workflow as a series of steps, each with an associated compensating action (a rollback step to undo previous work if something goes wrong).

Recall the example:

1. ✔️ Airline reservation is successful.

2. ✔️ Hotel reservation is successful.

3. ❌ Rental car reservation fails.

Using the Saga Pattern, we can roll back previous steps:

↩️ Cancel hotel reservation

↩️ Cancel airline reservation

As illustred in the image above, the operation behaves atomically: either all services are booked, or none are, ensuring users will not be left with partial bookings.

Conclusion

• The Saga pattern is an approach to achieve transactions in a distributed system.
• This is achieved by each step in a workflow having an associated compensating action, i.e. rollback to undo previous work if a failure is encountered.
• This improves the reliability and data consistency of a system.

References

• Saga distributed transactions pattern
• Pattern: Saga