前面我们搭建了具有服务降级功能的Hystrix客户端,现在我们来详细了解下Hystrix的一些功能。

Hystrix的意思是豪猪,大家都知道,就是长满刺的猪。。。实际上,它表明了该框架的主要功能:自我保护功能。Hystrix具有服务降级,熔断,线程池隔离,信号量隔离,缓存等功能,基本上能覆盖到微服务中调用依赖服务会遇到的问题。下面我们介绍下,如何理解和使用这些功能。

1、最常用的的服务降级功能

  当执行调用服务方法时,若调用方法出现问题,如:请求超时,抛出异常,线程池拒绝,熔断这些情况下,为该方法定义降级方法,以便在出现问题时执行,实现备用返回。之前我们已经实现了服务降级功能,主要就是通过@HystrixCommand(fallbackMethod = "defaultMethod")注释到需要在出现问题时降级的方法。fallbackMethod指定降级后执行的方法。方法定义在该类中,public,private,protected都可以。在注释的方法出问题后,如超时未返回(execution.isolation.thread.timeoutinMilliseconds来配置),就会执行备用方法,返回备用方法的返回值。当然,降级的方法也可以定义再下一级的降级方法,实现和上面一样。

  上面说到方法抛出异常也会触发服务降级,但是如果我们自定义了异常,并需要将异常抛出给上层做操作,不希望Hystrix捕捉到自定义异常执行服务降级时,可以使用@HystrixCommand(ignoreExceptions = {MyException.class})来定义忽略捕捉的异常。多个异常用逗号隔开。也可以将抛出的异常通过入参传到降级的方法,来实现不同类型异常的不同处理,需要将降级方法定义如下。

@HystrixCommand(fallbackMethod = "back")public String getHello(String id)    {return template.getForObject("http://helloclient/hello", String.class);    }public String back(String id , Throwable e)    {if (e instanceof NullPointerException)        {return "client 2 has some error! NullPointerException";        }else{return "client 2 has some error! Exception";        }    }

 2、熔断器

  熔断器,和物理概念中的断路器类似,断路器在高压高温的过载情况下,会自动断开,实现对电路的保护。熔断器也是一样,下面我们看下主要的接口类:HystrixCircuitBreaker.java,它定义了以下几个方法,并有两个内部实现类HystrixCircuitBreakerImpl,NoOpCircuitBreaker,断路器主要用到HystrixCircuitBreakerImpl。NoOpCircuitBreaker这个类表明不做任何操作,默认熔断器不打开,表明不起用熔断功能。以下的实现方法,都是指HystrixCircuitBreakerImpl的实现。熔断器有三个状态,OPEN,HALF_OPEN,CLOSED,如果要自定义参数配置,下面代码注释中可以找到。

/** * Every {@link HystrixCommand} requests asks this if it is allowed to proceed or not.  It is idempotent and does     * not modify any internal state, and takes into account the half-open logic which allows some requests through     * after the circuit has been opened     *      * @return boolean whether a request should be permitted     */boolean allowRequest();/** * Whether the circuit is currently open (tripped).     *      * @return boolean state of circuit breaker     */boolean isOpen();/** * Invoked on successful executions from {@link HystrixCommand} as part of feedback mechanism when in a half-open state.     */void markSuccess();/** * Invoked on unsuccessful executions from {@link HystrixCommand} as part of feedback mechanism when in a half-open state.     */void markNonSuccess();/** * Invoked at start of command execution to attempt an execution.  This is non-idempotent - it may modify internal     * state.     */boolean attemptExecution();

(1) isOpen()方法用于判断熔断器是否打开。实现方法如下:

 @Overridepublic boolean isOpen() {           //判断熔断器是否被强制打开,如果强制打开,返回true,表示熔断器已打开。circuitBreaker.forceOpen这个配置指定if (properties.circuitBreakerForceOpen().get()) {return true;            }          //判断熔断器是否被强制关闭。circuitBreaker.forceClosedif (properties.circuitBreakerForceClosed().get()) {return false;            }           //判断上一次断路器打开的时间是否大于零,访问成功,该值为-1,访问失败,该值为访问失败时的系统时间。根据是否大于零,判断熔断器是否打开。return circuitOpened.get() >= 0;        }

 

 

(2) attemptExecution(),该方法会在熔断器开启的时候,有访问时,熔断器第一个执行的方法。如果返回false,则直接执行fallback降级方法。

@Overridepublic boolean attemptExecution() {           //判断熔断器是否被强制打开,如果强制打开,返回false后,直接执行fallbackif (properties.circuitBreakerForceOpen().get()) {return false;            }          //判断熔断器是否被强制关闭if (properties.circuitBreakerForceClosed().get()) {return true;            }          //如果circuitOpened为-1,返回true,正常执行if (circuitOpened.get() == -1) {return true;            } else {              //如果circuitOpened不为-1,则表示断路器打开了,此时,服务会从circuitOpened起,休眠5秒(circuitBreaker.sleepWindowInMilliseconds配置,              //默认5000),直接返回false,执行fallback。若休眠时间超过5秒,并且当前熔断状态为打开状态,则会将熔断状态置为半开状态。如它的注释,只有第一个              //请求满足第一次为打开,之后的请求都为半开状态,返回false。if (isAfterSleepWindow()) {if (status.compareAndSet(Status.OPEN, Status.HALF_OPEN)) {//only the first request after sleep window should executereturn true;                    } else {return false;                    }                } else {return false;                }            }        }

 

(3)markSuccess(),在执行完attemptExecution()返回true正常执行成功后(未fallback),才会执行该方法,标注成功,若之前断路器为关闭状态,则不做处理,若为半开状态,则重置熔断器。

 @Overridepublic void markSuccess() {           //如果当前状态为半开,则将state设置成closed,关闭熔断器。如果之前由于断路器打开时,之后的请求,Hystrix会放开一个请求去尝试是否服务正常,并将断路器置为半开,           //如果正常,则将断路器关闭,并重置断路器。if (status.compareAndSet(Status.HALF_OPEN, Status.CLOSED)) {//This thread wins the race to close the circuit - it resets the stream to start it over from 0                metrics.resetStream();                Subscription previousSubscription = activeSubscription.get();if (previousSubscription != null) {                    previousSubscription.unsubscribe();                }                Subscription newSubscription = subscribeToStream();                activeSubscription.set(newSubscription);                circuitOpened.set(-1L);            }        }

(4) markNonSuccess(),用来在正常请求下,请求失败后调用。

 @Overridepublic void markNonSuccess() {           //如果当前为半开状态,且请求失败,则重新打开断路器,将最近一次访问失败的时间置为当前时间。if (status.compareAndSet(Status.HALF_OPEN, Status.OPEN)) {//This thread wins the race to re-open the circuit - it resets the start time for the sleep window                circuitOpened.set(System.currentTimeMillis());            }        }

(5) 熔断器的打开。上面的方法都不会去打开熔断器,熔断器打开是由另一个方法去判断的。这个观察者的方法应该是周期执行的。

private Subscription subscribeToStream() {/* * This stream will recalculate the OPEN/CLOSED status on every onNext from the health stream             */return metrics.getHealthCountsStream()                    .observe()                    .subscribe(new Subscriber<HealthCounts>() {                        @Overridepublic void onCompleted() {                        }                        @Overridepublic void onError(Throwable e) {                        }                        @Overridepublic void onNext(HealthCounts hc) {// check if we are past the statisticalWindowVolumeThreshold                           //检查时间窗内的请求总数小于配置文件中的数量(采用的是buckets,感兴趣的自己研究下)。默认时间窗为10S(metrics.rollingStats.timeInMilliseconds,metrics.rollingStats.numBuckets),默认请求总数为20(circuitBreaker.requestVolumeThreshold)。if (hc.getTotalRequests() < properties.circuitBreakerRequestVolumeThreshold().get()) {// we are not past the minimum volume threshold for the stat window,// so no change to circuit status.// if it was CLOSED, it stays CLOSED// if it was half-open, we need to wait for a successful command execution// if it was open, we need to wait for sleep window to elapse} else {//时间窗内,统计的错误(失败)请求比例是否小于配置比例,默认配置是50%,通过circuitBreaker.errorThresholdPercentage=50指定。if (hc.getErrorPercentage() < properties.circuitBreakerErrorThresholdPercentage().get()) {//we are not past the minimum error threshold for the stat window,// so no change to circuit status.// if it was CLOSED, it stays CLOSED// if it was half-open, we need to wait for a successful command execution// if it was open, we need to wait for sleep window to elapse} else {// our failure rate is too high, we need to set the state to OPEN                                   //如果时间窗内请求数大于定义数,且失败比例大于定义比例,并且当前熔断器关闭的情况下,将熔断器置为打开,并将circuitOpened置为当前时间。if (status.compareAndSet(Status.CLOSED, Status.OPEN)) {                                        circuitOpened.set(System.currentTimeMillis());                                    }                                }                            }                        }                    });        }

(6) 过程:先文字敲吧,没画图工具。

  正常情况:请求——>subscribeToStream未打开熔断器——>attemptExecution——>markSuccess

  异常情况:请求——>subscribeToStream打开熔断器——>attemptExecution最后一个return返回false——>markNonSuccess,这个时候断路器打开状态,且在休眠时间窗内。

       请求——>subscribeToStream未处理——>attemptExecution在超过休眠时间窗后,放开一个请求,并把熔断器设置成半开——>请求成功,执行markSuccess,将熔断器从半开置为关闭,并重置熔断器;请求失败,则将半开状态置为打开状态,失败时间起点重置成当前时间,再次循环。

3、缓存

  之前我以为是每次相同请求,会使用缓存直接返回。其实理解错了,Hystrix的缓存是在当次请求的缓存,当次请求中,多次使用同一方法时,会使用缓存。其他请求不能用到。而且还需初始化HystrixRequestContext,不然直接使用会报错,我们采用定义filter来初始化。不多说了,贴代码大家看下,代码中注释很清楚,启动注册中心和服务实例后(环境搭建见之前章节),就可以测试。

(1)pom.xml,application.yml配置,大家参见之前的章节。

(2)启动类,注意注解上@ServletComponentScan。

package com.example.demo;import org.springframework.boot.SpringApplication;import org.springframework.boot.autoconfigure.SpringBootApplication;import org.springframework.boot.web.servlet.ServletComponentScan;import org.springframework.cloud.client.circuitbreaker.EnableCircuitBreaker;import org.springframework.cloud.client.loadbalancer.LoadBalanced;import org.springframework.cloud.netflix.eureka.EnableEurekaClient;import org.springframework.context.annotation.Bean;import org.springframework.web.client.RestTemplate;@EnableCircuitBreaker@SpringBootApplication@EnableEurekaClient@ServletComponentScanpublic class ConsumerApplication {    @Bean    @LoadBalanced    RestTemplate template()    {return new RestTemplate();    }public static void main(String[] args) {        SpringApplication.run(ConsumerApplication.class, args);    }}

(3)Filter类,用于初始化HystrixRequestContext。

package com.example.demo;import com.netflix.hystrix.strategy.concurrency.HystrixRequestContext;import javax.servlet.*;import javax.servlet.annotation.WebFilter;import java.io.IOException;@WebFilter(filterName = "HystrixRequestContextServletFilter",urlPatterns = "/*",asyncSupported = true)public class HystrixRequestContextServletFilter implements Filter {    @Overridepublic void init(FilterConfig filterConfig) throws ServletException {    }    @Overridepublic void doFilter(ServletRequest request, ServletResponse response, FilterChain chain) throws IOException, ServletException {        HystrixRequestContext context = HystrixRequestContext.initializeContext();try{            chain.doFilter(request,response);        }finally {            context.shutdown();        }    }    @Overridepublic void destroy() {    }}

(4)controller类。

package com.example.demo;import org.springframework.beans.factory.annotation.Autowired;import org.springframework.web.bind.annotation.RequestMapping;import org.springframework.web.bind.annotation.RestController;@RestControllerpublic class ConsumerContorller {    @Autowired    HystrixServer server;//注意,在这个controller中调用具有缓存功能的方法才会具备缓存效果。@RequestMapping("/hello")public String sayHello()    {        System.out.println("请求了一次hello2");        server.getHello2("1","ibethfy");        System.out.println("请求了二次hello2,不会打印hello2 initinized");        server.getHello2("1","ibethfy");        System.out.println("请求了三次hello2,清空缓存,会打印hello2 initinized");        server.updateHello2("1","ibethfy");        server.getHello2("1","ibethfy");        System.out.println("请求了四次hello2,入参不同,会打印hello2 initinized");        server.getHello2("1","ibethfy1");return server.getHello2("1","ibethfy1");    }}

(5)server类。

package com.example.demo;import com.netflix.hystrix.contrib.javanica.annotation.HystrixCommand;import com.netflix.hystrix.contrib.javanica.annotation.HystrixProperty;import com.netflix.hystrix.contrib.javanica.cache.annotation.CacheKey;import com.netflix.hystrix.contrib.javanica.cache.annotation.CacheRemove;import com.netflix.hystrix.contrib.javanica.cache.annotation.CacheResult;import com.netflix.hystrix.strategy.concurrency.HystrixRequestContext;import org.springframework.beans.factory.annotation.Autowired;import org.springframework.stereotype.Service;import org.springframework.web.client.RestTemplate;@Servicepublic class HystrixServer {    @Autowired    RestTemplate template;//通过指定生成缓存key的方法生成key,commandKey指定一个HystrixCommand的key,表示注解@HystrixCommand的方法的key。groupKey表示一个类型分组的key。threadPoolKey指定线程池的key。//fallbackMethod指定降级方法,commandProperties指定该HystrixCommand方法的参数,是个数组类型,里面的值是@HystrixProperty,多个用逗号隔开。@CacheResult(cacheKeyMethod = "generateCacheKey")    @HystrixCommand(commandKey = "getHello1",groupKey = "getHello",threadPoolKey = "getHelloThreadPool",fallbackMethod = "back",commandProperties = {            @HystrixProperty(name="execution.isolation.thread.timeoutinMilliseconds", value = "5000")    })public String getHello1()    {        System.out.println("hello1 initinized");return template.getForObject("http://helloclient/hello", String.class);    }private String generateCacheKey()    {return "myHelloKey";    }//若不指定cache的key,默认使用方法的所有参数作为key    @CacheResult    @HystrixCommand(commandKey = "getHello2",groupKey = "getHello",threadPoolKey = "getHelloThreadPool")public String getHello2(String id,String name)    {        System.out.println("hello2 initinized");return template.getForObject("http://helloclient/hello", String.class);    }//使用@CacheRemove在数据更新时,移除对应key的缓存,需要指定commandKey,@HystrixCommand里面的参数可以指定亦可以不用@CacheRemove(commandKey = "getHello2")    @HystrixCommand(commandKey = "getHello2",groupKey = "getHello",threadPoolKey = "getHelloThreadPool")public void updateHello2(String id,String name)    {        System.out.println("hello2 id = "+ id + ", name = "+ name + " removed");    }//使用@CacheKey指定参数作为key    @CacheResult    @HystrixCommand(commandKey = "getHello3",groupKey = "getHello",threadPoolKey = "getHelloThreadPool")public String getHello3(@CacheKey("id") String id, String name)    {        System.out.println("请求了一次hello3");return "hello3 " + id + name;    }public String back(Throwable e)    {if (e instanceof NullPointerException)        {return "client 2 has some error! NullPointerException";        }else{return "client 2 has some error! Exception";        }    }}


4、线程隔离和信号量隔离。

  Hystrix为了避免多个不同服务间的调用影响,使用了线程隔离模式,它为每个依赖服务单独创建自己的线程池,就算某个服务延迟或问题阻塞,其余的服务也能正常执行。总之,使得我们的服务更加健壮。当然,创建这么多线程池,可能会对性能造成影响,但Hystrix测试后,独立线程池带来的好处,远大于性能损耗的坏处。所以,大家可以放心使用。

  ExecutionIsolationStrategy枚举中定义了两个THREAD, SEMAPHORE,一个是线程池,一个是信号量,Hystix默认使用线程池。通过execution.isolation.strategy可以切换。

  Hystrix默认情况下,会让配置了同组名的groupKey的command使用同一线程池,但也支持使用threadPoolKey配置线程池key。

  对于那些本来延迟就比较小的请求(例如访问本地缓存成功率很高的请求)来说,线程池带来的开销是非常高的,这时,可以考虑采用非阻塞信号量(不支持超时),来实现依赖服务的隔离,使用信号量的开销很小。但绝大多数情况下,Netflix 更偏向于使用线程池来隔离依赖服务,因为其带来的额外开销可以接受,并且能支持包括超时在内的所有功能。

 

好了,Hystrix的主要功能基本介绍完了,码字不容易呀,,,,

  

©著作权归作者所有:来自51CTO博客作者独孤一笑的原创作品,如需转载,请注明出处,否则将追究法律责任

更多相关文章

  1. JavaScript中的信标beacon是什么?
  2. 浅谈Web安全常见问题
  3. crmeb H5本机访问正常,其他机器访问出现请求失败
  4. JavaScript:演示Ajax的get和post请求,练习选顶卡和换肤案例
  5. nodejs搭建本地服务器
  6. OSI七层模型和TCP/IP模型
  7. 【JS基础入门】JavaScript原生Ajax请求与响应处理方式及JSON数据
  8. Tomcat日常维护笔记 - 运维小结
  9. JS跳转引发浏览器NS_BINDING_ABORTED

随机推荐

  1. Android onActivityResult()不执行的几个
  2. Android的视频播放之VideoView与SurfaceV
  3. Android入门之helloworld
  4. Android Zygote
  5. android framework 图解分享
  6. Android 布局 & 一些控件
  7. Android(安卓)Material Design动画 View
  8. Android-AbsoluteLayout(绝对布局)
  9. Android 之shape 的用法介绍
  10. Android(安卓)数据存储ContentProvider(类