What is the difference between orchestration and choreography in sagas?

Expected Interview Answer

Step-by-Step Explanation

1. Step 1

   Identify the saga steps

   List the local transactions across services that make up the distributed workflow.

2. Step 2

   Decide who owns the flow

   Choose orchestration if one component should own and sequence the workflow, or choreography if services should react to events independently.

3. Step 3

   For orchestration, build a coordinator

   Create an orchestrator that sends commands to each service, tracks saga state, and triggers compensations on failure.

4. Step 4

   For choreography, define events

   Have each service publish domain events and subscribe to the events that should trigger its own next local transaction.

5. Step 5

   Handle failures per style

   Orchestrator invokes compensations directly; in choreography, failure events cascade to trigger each service's compensating action.

6. Step 6

   Add observability

   Correlate steps with a saga id so the flow can be traced regardless of style, especially important for choreography.

Code Example

// Central orchestrator issues commands in order
async function orderOrchestrator(order) {
  await commands.reserveInventory(order)
  await commands.chargePayment(order)
  await commands.dispatchShipment(order)
}

// Inventory service reacts to an OrderCreated event
bus.on('OrderCreated', async (order) => {
  await inventory.reserve(order)
  bus.emit('InventoryReserved', order)
})

// Payment service reacts to InventoryReserved
bus.on('InventoryReserved', async (order) => {
  await payment.charge(order)
  bus.emit('PaymentCharged', order)
})

Follow-up Questions

• When does choreography become hard to maintain?
• How do you monitor a choreographed saga end to end?
• Can you mix orchestration and choreography in one system?
• How does each style handle a compensating transaction?
• What role does a message broker play in choreography?