熔断器

当某个机器接口失败率达到某一条件（比如失败率>50%），就可以采用熔断器将服务器从集群机器中移除，或者采用自定义的业务逻辑降级方案，比如从一个接口转到另一个接口。熔断器只是一个工具，重要的是业务逻辑。

Resilience4j是什么？

Resilience4j 是受Netflix的Hysrix项目启发，专门为Java 8 和函数式编程设计的轻量级容错框架。Resilicenes4j 仅使用了一个第三方开源库Vavr，Vavr不依赖其他库。相比较而言，Netflix Hysrix对Archaius存在编译依赖，Archaius有许多外部依赖，比如Guava和Apache Commons Configuration。Resilience4j 按模块发布，可以有选择的使用其中某些功能而无需引入全部的Resilience4j 组件。

resilience4j是一款受hystrix启发的容错组件，提供了如下几款核心组件：

• resilience4j-circuitbreaker: Circuit breaking
• resilience4j-ratelimiter: Rate limiting
• resilience4j-bulkhead: Bulkheading
• resilience4j-retry: Automatic retrying (sync and async)
• resilience4j-cache: Response caching

Resilience4j的使用

1. CircuitBreaker（熔断器）

CircuitBreaker主要是实现针对接口异常的断路统计以及断路处理。

@Test
    public void testCircuitBreaker(){
        // Create a CircuitBreaker (use default configuration)
        CircuitBreakerConfig circuitBreakerConfig = CircuitBreakerConfig
                .custom()
                .enableAutomaticTransitionFromOpenToHalfOpen()
                .build();
        CircuitBreaker circuitBreaker = CircuitBreaker
                .of("backendName",circuitBreakerConfig);
        String result = circuitBreaker.executeSupplier(() -> backendService.doSomethingWithArgs("world"));
        System.out.println(result);
    }

2. Timelimiter

主要是实现超时的控制。

 @Test
    public void testTimelimiter(){
        ExecutorService executorService = Executors.newSingleThreadExecutor();
        TimeLimiterConfig config = TimeLimiterConfig.custom()
                .timeoutDuration(Duration.ofMillis(600))
                .cancelRunningFuture(true)
                .build();
        TimeLimiter timeLimiter = TimeLimiter.of(config);

        Supplier<Future<String>> futureSupplier = () -> {
            return executorService.submit(backendService::doSomethingThrowException);
        };
        Callable<String> restrictedCall = TimeLimiter.decorateFutureSupplier(timeLimiter,futureSupplier);
        Try.of(restrictedCall::call)
                .onFailure(throwable -> System.out.println("We might have timed out or the circuit breaker has opened."));
    }

3. Bulkhead

Bulkhead目前来看是用来控制并行(parallel)调用的次数。

/**
     * A Bulkhead can be used to limit the amount of parallel executions
     */
    @Test
    public void testBulkhead(){
        Bulkhead bulkhead = Bulkhead.of("test", BulkheadConfig.custom()
                .maxConcurrentCalls(1)
                .build());
        Supplier<String> decoratedSupplier = Bulkhead.decorateSupplier(bulkhead, backendService::doSomethingSlowly);
        IntStream.rangeClosed(1,2)
                .parallel()
                .forEach(i -> {
                    String result = Try.ofSupplier(decoratedSupplier)
                            .recover(throwable -> "Hello from Recovery").get();
                    System.out.println(result);
                });
    }

4. RateLimiter

用来做流控。

    public void testRateLimiter(){
        // Create a custom RateLimiter configuration
        RateLimiterConfig config = RateLimiterConfig.custom()
                .timeoutDuration(Duration.ofMillis(100))
                .limitRefreshPeriod(Duration.ofSeconds(1))
                .limitForPeriod(1)
                .build();
        // Create a RateLimiter
        RateLimiter rateLimiter = RateLimiter.of("backendName", config);

        // Decorate your call to BackendService.doSomething()
        Supplier<String> restrictedSupplier = RateLimiter
                .decorateSupplier(rateLimiter, backendService::doSomething);

        IntStream.rangeClosed(1,5)
                .parallel()
                .forEach(i -> {
                    Try<String> aTry = Try.ofSupplier(restrictedSupplier);
                    System.out.println(aTry.isSuccess());
                });
    }

5. Fallback

fallback基本上是高可用操作的标配。

@Test
    public void testFallback(){
        // Execute the decorated supplier and recover from any exception
        String result = Try.ofSupplier(() -> backendService.doSomethingThrowException())
                .recover(throwable -> "Hello from Recovery").get();
        System.out.println(result);
    }

    @Test
    public void testCircuitBreakerAndFallback(){
        CircuitBreaker circuitBreaker = CircuitBreaker.ofDefaults("backendName");
        Supplier<String> decoratedSupplier = CircuitBreaker
                .decorateSupplier(circuitBreaker, backendService::doSomethingThrowException);
        String result = Try.ofSupplier(decoratedSupplier)
                .recover(throwable -> "Hello from Recovery").get();
        System.out.println(result);
    }

6. Retry

retry用来控制重试。

@Test
    public void testRetry(){
        CircuitBreaker circuitBreaker = CircuitBreaker.ofDefaults("backendName");
        // Create a Retry with at most 3 retries and a fixed time interval between retries of 500ms
        Retry retry = Retry.ofDefaults("backendName");

        // Decorate your call to BackendService.doSomething() with a CircuitBreaker
        Supplier<String> decoratedSupplier = CircuitBreaker
                .decorateSupplier(circuitBreaker, backendService::doSomething);

        // Decorate your call with automatic retry
        decoratedSupplier = Retry
                .decorateSupplier(retry, decoratedSupplier);

        // Execute the decorated supplier and recover from any exception
        String result = Try.ofSupplier(decoratedSupplier)
                .recover(throwable -> "Hello from Recovery").get();
        System.out.println(result);
    }

Sentinel 与 Hystrix、resilience4j 的对比