Spring Boot微服务熔断降级实战：Sentinel与Resilience4j配置对比指南

微服务熔断降级为什么是刚需

微服务架构中，服务间调用链路越长，单点故障的雪崩效应越严重。一个支付服务的数据库连接池耗尽，导致支付接口超时，上游订单服务线程池被支付超时请求占满，再上游商品服务因为订单服务无响应而堆积请求——三层调用下来，一个小故障就能拖垮整条链路。

熔断器的核心机制是：当下游服务的错误率或慢调用比例超过阈值时，自动切断请求（Open状态），直接走降级逻辑，给下游恢复的时间。经过一个冷却窗口后进入Half-Open状态，放少量请求探测，成功则恢复（Closed），失败则继续熔断。

Spring Cloud生态中有两个主流熔断组件：Sentinel（阿里巴巴开源）和Resilience4j。两者的设计哲学差异决定了选型方向。

Sentinel配置实战

Sentinel的核心概念是资源（Resource）和规则（Rule）。资源是被保护的业务逻辑，规则定义了触发流控、熔断、降级的条件。

引入依赖与基础配置

<!-- pom.xml -->
<dependency>
    <groupId>com.alibaba.cloud</groupId>
    <artifactId>spring-cloud-starter-alibaba-sentinel</artifactId>
    <version>2023.0.1.0</version>
</dependency>

<dependency>
    <groupId>com.alibaba.csp</groupId>
    <artifactId>sentinel-datasource-nacos</artifactId>
</dependency>

# application.yml
spring:
  cloud:
    sentinel:
      transport:
        dashboard: localhost:8080
        port: 8719
      datasource:
        flow:
          nacos:
            server-addr: localhost:8848
            data-id: sentinel-flow-rules
            rule-type: flow
        degrade:
          nacos:
            server-addr: localhost:8848
            data-id: sentinel-degrade-rules
            rule-type: degrade

熔断规则配置

Sentinel支持三种熔断策略：慢调用比例、异常比例、异常数。

@Configuration
public class SentinelRuleConfig {

    @PostConstruct
    public void initRules() {
        // 熔断规则：慢调用比例策略
        DegradeRule slowCallRule = new DegradeRule("paymentService")
            .setGrade(CircuitBreakerStrategy.SLOW_REQUEST_RATIO.getType())
            .setCount(500)
            .setSlowRatioThreshold(0.6)
            .setTimeWindow(30)
            .setMinRequestAmount(10)
            .setStatIntervalMs(10000);

        // 熔断规则：异常比例策略
        DegradeRule errorRatioRule = new DegradeRule("paymentService")
            .setGrade(CircuitBreakerStrategy.ERROR_RATIO.getType())
            .setCount(0.5)
            .setTimeWindow(30)
            .setMinRequestAmount(5)
            .setStatIntervalMs(10000);

        List<DegradeRule> rules = List.of(slowCallRule, errorRatioRule);
        DegradeRuleManager.loadRules(rules);
    }
}

Feign集成Sentinel降级

# 开启Feign的Sentinel支持
feign:
  sentinel:
    enabled: true

// 降级回调类
@Component
public class PaymentServiceFallback implements PaymentService {
    @Override
    public PaymentResult pay(PaymentRequest request) {
        return PaymentResult.fail("服务暂不可用，请稍后重试");
    }

    @Override
    public PaymentResult query(String orderId) {
        return PaymentResult.cached("缓存数据");
    }
}

// Feign客户端声明
@FeignClient(
    name = "payment-service",
    fallbackFactory = PaymentServiceFallbackFactory.class
)
public interface PaymentService {
    @PostMapping("/pay")
    PaymentResult pay(PaymentRequest request);

    @GetMapping("/query/{orderId}")
    PaymentResult query(@PathVariable String orderId);
}

Resilience4j配置实战

Resilience4j是轻量级方案，不依赖外部Dashboard，所有配置通过代码或YAML完成。它的优势在于与Spring Boot 3和Virtual Threads的兼容性更好。

引入依赖

<dependency>
    <groupId>io.github.resilience4j</groupId>
    <artifactId>resilience4j-spring-boot3</artifactId>
    <version>2.2.0</version>
</dependency>
<dependency>
    <groupId>io.github.resilience4j</groupId>
    <artifactId>resilience4j-circuitbreaker</artifactId>
</dependency>

# application.yml - Resilience4j熔断配置
resilience4j:
  circuitbreaker:
    instances:
      paymentService:
        registerHealthIndicator: true
        slidingWindowType: COUNT_BASED
        slidingWindowSize: 10
        minimumNumberOfCalls: 5
        failureRateThreshold: 50
        slowCallDurationThreshold: 500ms
        slowCallRateThreshold: 60
        waitDurationInOpenState: 30s
        permittedNumberOfCallsInHalfOpenState: 3
  timelimiter:
    instances:
      paymentService:
        timeoutDuration: 3s

编程式使用

@Service
public class OrderService {

    private final CircuitBreaker circuitBreaker;

    public OrderService(CircuitBreakerRegistry registry) {
        this.circuitBreaker = registry.circuitBreaker("paymentService");
    }

    public PaymentResult processPayment(PaymentRequest request) {
        return CircuitBreaker.decorateSupplier(circuitBreaker,
            () -> paymentClient.pay(request)
        ).get();
    }

    @CircuitBreaker(name = "paymentService", fallbackMethod = "payFallback")
    public PaymentResult pay(PaymentRequest request) {
        return paymentClient.pay(request);
    }

    private PaymentResult payFallback(PaymentRequest request, Exception e) {
        log.warn("支付服务熔断降级: {}", e.getMessage());
        return PaymentResult.fail("服务暂不可用");
    }
}

Sentinel vs Resilience4j选型对比

选型建议：如果团队已有Nacos作为配置中心且需要Dashboard实时调整规则，选Sentinel；如果项目追求轻量、无外部依赖、更倾向代码即配置的风格，选Resilience4j。两者在熔断核心功能上差异不大，差异主要在运维体验和生态集成上。

熔断降级的监控告警

无论选择哪个组件，都必须建立熔断事件的告警机制。熔断触发意味着服务已经出现异常，需要人工介入。

// Sentinel熔断事件监听
EventObserverRegistry.getInstance().addStateChangeObserver(
    "alert-observer",
    (preState, newState, rule) -> {
        if (newState == CircuitBreaker.State.OPEN) {
            alertService.send(
                "熔断触发: " + rule.getResource()
                + " 规则: " + rule.toString()
            );
        }
    }
);

// Resilience4j熔断事件监听
circuitBreaker.getEventPublisher()
    .onStateTransition(event -> {
        if (event.getStateTransition() == StateTransition.CLOSED_TO_OPEN) {
            alertService.send("熔断触发: " + circuitBreaker.getName());
        }
    });

完善的熔断降级方案不是装个组件就结束的，还需要配套的降级策略设计、监控告警、规则调优。降级策略要根据业务影响分级——核心链路降级走缓存，非核心功能降级直接返回默认值或关闭。规则参数需要在压测和灰度中持续调整，初始阈值建议设置宽松一些（如异常比例80%），观察一周后再收紧。