How do you build microservices with Spring Boot: config, discovery, resilience?

Question

IT Interview Practice · Accepted Answer

Spring Boot plus Spring Cloud gives you the building blocks for microservices. The three problems every distributed system must solve are **centralized configuration**, **service discovery**, and **resilience** against partial failure. ## The moving parts ```text ┌── Config Server (git-backed properties) │ Client ──▶ API Gateway ──▶ Service A ──▶ Service B │ │ │ └──▶ Discovery (Eureka / K8s DNS) ◀── register + look up ``` - **Config Server** serves each service's properties from a central git repo, so you change config without redeploying. Services pull it on startup (and refresh via `/actuator/refresh`). - **Service discovery** (Eureka, Consul, or Kubernetes DNS) lets Service A find Service B by *name* instead of a hard-coded host, so instances can scale and move freely. - **API Gateway** (Spring Cloud Gateway) is the single entry point for routing, auth, and rate limiting. ## Resilience — the part juniors skip In a distributed system every network call *will* fail eventually. Use Resilience4j to stop one slow service cascading into an outage: ```java @CircuitBreaker(name = "inventory", fallbackMethod = "cachedStock") @Retry(name = "inventory") // retry transient blips @TimeLimiter(name = "inventory") // cap the wait public CompletableFuture stock(String sku) { return inventoryClient.get(sku); // remote call } // Called when the circuit is open or the call fails — degrade gracefully. public CompletableFuture cachedStock(String sku, Throwable ex) { return CompletableFuture.completedFuture(Stock.unknown(sku)); } ``` The **circuit breaker** trips open after repeated failures, so calls fail fast instead of piling up threads waiting on a dead dependency. ## Why it matters Senior interviews probe distributed-systems maturity. Naming discovery and config is table stakes; the differentiator is talking about **failure**: circuit breakers, timeouts, idempotency, and observability (distributed tracing with Micrometer/Zipkin). Strong candidates also push back — noting microservices add operational cost and a monolith is often the right call until scale demands otherwise.

The moving parts

Resilience — the part juniors skip

Why it matters

Related questions