사가 오케스트레이터 설계

In an Orchestration Saga, a dedicated service, called the Orchestrator, takes charge of managing the entire distributed transaction.

Think of it as a conductor in an orchestra. Instead of individual musicians (services) reacting to each other, the conductor (orchestrator) tells each musician when to play their part.

Orchestration offers several advantages, especially for complex business flows:

• Centralized Logic: All saga logic resides in one place, making it easier to understand and manage.
• Easier Changes: Modifying saga steps or adding new ones is simpler as changes are contained within the orchestrator.
• Clearer Debugging: It's easier to trace the flow of a transaction and pinpoint exactly where a failure occurred.

A saga orchestrator has crucial responsibilities to ensure the transaction completes successfully or is properly compensated:

• Initiates Saga: Starts the transaction by sending the first command.
• Tracks State: Maintains the current progress and state of the saga.
• Coordinates Steps: Sends commands to participant services based on the current state and responses.
• Handles Compensation: Triggers rollback actions if any step fails.

Let's consider an 'Order Placement' saga involving multiple services:

1. Order Service: Creates the order.
2. Payment Service: Processes the payment.
3. Inventory Service: Updates stock.
4. Shipping Service: Arranges delivery.

The orchestrator will guide the order through these steps.

The orchestrator must keep track of where the saga is in its lifecycle. This is its state. It typically stores:

• sagaId: A unique identifier for the current transaction.
• currentStep: Which step is currently active or has been completed.
• status: e.g., IN_PROGRESS, COMPLETED, FAILED, COMPENSATING.

This state must be persisted (saved) so the orchestrator can recover if it crashes.

The orchestrator's design centers around a state machine-like flow:

1. Orchestrator sends a command to a participant service (e.g., 'ProcessPayment').
2. Participant service performs its action and sends an event back (e.g., 'PaymentProcessed' or 'PaymentFailed').
3. Orchestrator receives the event, updates its state, and decides the next command to send, or initiates compensation.

Here's a simplified Java example showing how an orchestrator might manage its state and react to responses. This code simulates the flow:

For reliable communication, orchestrators typically interact with participant services via a message broker (like Apache Kafka or RabbitMQ).

• Orchestrator publishes commands to queues/topics for specific services.
• Participant services publish events (success or failure) back to topics that the orchestrator subscribes to.

This asynchronous messaging ensures loose coupling and resilience.

If a participant service reports a failure, the orchestrator must initiate compensation. This means reversing any successfully completed steps.

For example, if payment succeeded but inventory update failed, the orchestrator would send a 'RefundPayment' command to the Payment Service.

The orchestrator's persisted state is vital here, as it knows exactly which steps need to be undone.

Which of the following are key responsibilities of a Saga Orchestrator?

In this lesson, we learned about designing a saga orchestrator. It's a dedicated service that acts as a central coordinator for distributed transactions.

• It initiates steps, tracks state, and coordinates participant services.
• Key to its design are persistent state management and robust communication via message brokers.
• Orchestrators simplify debugging and managing complex business flows, especially with their built-in compensation logic.

Next, we'll explore how to use state machines to build even more robust orchestrators!