How can data consistency be ensured when performing transactions across different microservices?

Ensuring data consistency during transactions across different microservices is a complex task, as traditional distributed transactions (such as two-phase commit) often do not fit due to performance and fault tolerance considerations.

Main approaches:

• Sagas: breaking the transaction into a series of local transactions with compensating actions in case of errors. Each service performs its part and publishes an event, which the next service reacts to.

• Idempotency and repeatability: operations should be safe to repeat to avoid inconsistencies.

• Use of events and queues: message exchange between services to coordinate state.

An example in C# (simplified) — implementation of a saga using events:

public class OrderSaga
{
    public async Task HandleOrderCreated(OrderCreatedEvent evt)
    {
        await ReserveInventory(evt.OrderId, evt.Items);
        await ProcessPayment(evt.OrderId, evt.PaymentDetails);
        // etc.
    }

    private async Task ReserveInventory(Guid orderId, List<Item> items) { /* ... */ }
    private async Task ProcessPayment(Guid orderId, PaymentDetails payment) { /* ... */ }
}

Thus, each operation is a separate transaction, and consistency is achieved through saga logic and compensating actions.