microservices_patterns

Decompose by business capability or DDD subdomains, mapping bounded contexts to services like OrderService and PaymentService. Use the Strangler Fig pattern to gradually extract functionality while a proxy routes traffic between old and new systems.

Implement a Saga as an orchestrated sequence of steps, each with an action and a compensating action. On failure, execute compensations in reverse order, such as refunding a payment after inventory reservation fails, to maintain eventual consistency.

Use synchronous REST or gRPC for request-response queries needing immediate results, with retries and timeouts. Prefer asynchronous events via Kafka for decoupled workflows, since event-driven communication reduces tight coupling and improves resilience.

A circuit breaker tracks failures and opens after a threshold, rejecting requests immediately instead of waiting for timeouts. After a recovery timeout it enters half-open state to test the dependency, closing again once enough calls succeed.

Avoid microservices for early-stage products or small teams, since premature distribution adds network complexity, distributed data management, and operational overhead. A well-structured monolith is often better until clear service boundaries and scale requirements emerge.