Add retries and a circuit breaker with Resilience4j

Approach

Use the Resilience4j Spring Boot starter: @Retry for transient failures, @CircuitBreaker to fail fast when the dependency is unhealthy, with a fallback method. Configure thresholds in YAML and expose state via Actuator/metrics.

Solution

@Service
class RecommendationClient {
    private final RestClient client;
    RecommendationClient(RestClient recommendationsRestClient) { this.client = recommendationsRestClient; }

    @CircuitBreaker(name = "recommendations", fallbackMethod = "fallback")
    @Retry(name = "recommendations")
    public List<ProductSummary> forUser(String userId) {
        return client.get().uri("/users/{id}/recommendations", userId).retrieve()
                .body(new ParameterizedTypeReference<>() {});
    }

    private List<ProductSummary> fallback(String userId, Throwable t) {
        return popularProducts.cached();      // degrade gracefully
    }
}

resilience4j:
  circuitbreaker:
    instances:
      recommendations:
        sliding-window-size: 50
        failure-rate-threshold: 50
        slow-call-duration-threshold: 2s
        slow-call-rate-threshold: 50
        wait-duration-in-open-state: 30s
  retry:
    instances:
      recommendations:
        max-attempts: 3
        wait-duration: 200ms
        enable-exponential-backoff: true
        retry-exceptions: [org.springframework.web.client.ResourceAccessException]

Pitfalls

• Retrying inside an open circuit (order of aspects matters; check circuitBreakerAspectOrder/retryAspectOrder).
• Fallbacks that hide outages without metrics/alerts.
• No HTTP timeouts — the breaker can't detect slowness without them.

Variations

• Bulkheads to cap concurrent calls per dependency.
• Spring Cloud Circuit Breaker abstraction.

Further reading: Resilience4j: Spring Boot