Spring Cloud Hystrix源码篇（十二）

一、Hystrix处理流程

Hystrix流程图如下：

Hystrix整个工作流如下：

1. 构造一个 HystrixCommand或HystrixObservableCommand对象，用于封装请求，并在构造方法配置请求被执行需要的参数；
2. 执行命令，Hystrix提供了4种执行命令的方法，后面详述；
3. 判断是否使用缓存响应请求，若启用了缓存，且缓存可用，直接使用缓存响应请求。Hystrix支持请求缓存，但需要用户自定义启动；
4. 判断熔断器是否打开，如果打开，跳到第8步；
5. 判断线程池/队列/信号量是否已满，已满则跳到第8步；
6. 执行HystrixObservableCommand.construct()或HystrixCommand.run()，如果执行失败或者超时，跳到第8步；否则，跳到第9步；
7. 统计熔断器监控指标；
8. 走Fallback备用逻辑
9. 返回请求响应

从流程图上可知道，第5步线程池/队列/信号量已满时，还会执行第7步逻辑，更新熔断器统计信息，而第6步无论成功与否，都会更新熔断器统计信息。

二、Hystrix的核心原理

hystrix在服务降级熔断的过程中有几个步骤他是必须要去完成的

•  可配置化的降级策略（根据不同的服务降级类型配置不同的降级策略方案）:
  • 三种方式：信号量/线程 、超时（默认1s）、熔断（错误率）
  • 在HystrixCommandProperty类中通过相关属性去配置改变他的默认策略（上篇中有说明过）
• 可以识别的降级边界：
  • @HystrixCommand(Spring AOP通过注解标注一个接口的资源，去表示说明这个接口需要通过Hystrix来接管这个请求，如果达到注解内的配置要求就熔断)
  • 自己去继承HystrixCommand 抽象类，等下演示下，这玩意还挺好玩的
• 数据采集：
  • 如何触发熔断（上篇幅也说过10s 内20个请求 ，错误率达到50），这里引出的问题是如何采集数据，如何统计数据.
  • SEMAPHORE，最大并发数量 （它底层其实就是个AQS 统计次数tryAcquire(), acquire()）
• 行为干预： 触发降级/熔断之后，对正常业务产生影响
• 结果干预： 通过fallback()返回数据
• 自动恢复（处于熔断状态下，会每隔5s尝试去恢复）

 2.1、通过HystrixCommand 接管我们定义的请求

上一篇幅我是通过注解的方式来进行服务熔错的，这次不通过注解换一种方式，首先在spring-cloud-user服务中写以下内容

 

 然后启动服务访问浏览器，结果如果我想的一样

 2.2、Hystrix是如何工作的

下面演示个带超时降级的Hystrix注解

 然后用AOP写自己的拦截规则

/**
 *这里面用到的是AOP的知识点，如果不了解可以先自行补下，后面我有空把Spring的AOP原理也写下，这样回头看这个就没这么难了
 */
@Component
@Aspect  //切入
public class GhyHystrixAspect {
    //通过线程池去请求
    ExecutorService executorService= Executors.newFixedThreadPool(10);
    //定义切点针对GhyHystrix进行切入
    @Pointcut(value = "@annotation(GhyHystrix)")
    public void pointCut(){}
    //切入后执行的方法
    @Around(value = "pointCut()&&@annotation(hystrixCommand)")
    public Object doPointCut(ProceedingJoinPoint joinPoint, GhyHystrix hystrixCommand) throws InterruptedException, ExecutionException, TimeoutException, NoSuchMethodException, InvocationTargetException, IllegalAccessException {
        //定义超时降级
        int timeout=hystrixCommand.timeout();
        //前置的判断逻辑
        Future future=executorService.submit(()->{
            try {
                return joinPoint.proceed(); //执行目标方法
            } catch (Throwable throwable) {
                throwable.printStackTrace();
            }
            return null;
        });
        Object result;
        try {
            //得到开始和结束时间判断是否超时，如果超时就降级
            result=future.get(timeout, TimeUnit.MILLISECONDS);
        } catch (InterruptedException | ExecutionException | TimeoutException e) {
            e.printStackTrace();
            //超时了就取消请求
            future.cancel(true);
            // 先判断是否为空如果空就把异常抛出去
            if(StringUtils.isBlank(hystrixCommand.fallback())){
                throw e;
            }
            //调用fallback
            result=invokeFallback(joinPoint,hystrixCommand.fallback());
        }
        return result;
    }
   //反射调用
    private Object invokeFallback(ProceedingJoinPoint joinPoint,String fallback) throws NoSuchMethodException, InvocationTargetException, IllegalAccessException {
        MethodSignature signature=(MethodSignature)joinPoint.getSignature();
        //拿到方法的信息
        Method method=signature.getMethod();
        //得到参数类型
        Class<?>[] parameterTypes=method.getParameterTypes();
        //以上是获取被代理的方法的参数和Method
        //得到fallback方法
        try {
            Method fallbackMethod=joinPoint.getTarget().getClass().getMethod(fallback,parameterTypes);
            fallbackMethod.setAccessible(true);
            //完成反射调用
            return fallbackMethod.invoke(joinPoint.getTarget(),joinPoint.getArgs());
        } catch (Exception e) {
            e.printStackTrace();
            throw e;
        }
    }
}

然后再写个调用逻辑，用自己定义的注解

 浏览器访问，返回的不是我们刚刚定义的降级内容，其实这也挺好想的，我用的是之前的项目，之前在spring-cloud-api工程中定义了熔断规则，改一下就好

 将这此内容改下就好，还有配置文件隐藏下，这里就不搞了

 三、Hystrix的熔断的原理以及请求代理的原理

当请求过来时，如果请求失败，先判断请求次数是否达到了最小请求次数，再判断错误率是否达到了阈值，如果没达到就继续请求，这个错误率的统计时间默认是10S；如果达到了阈值就要打开断路器，打开断 路器后有5秒的时间是熔断状态，5秒后，如果有请求过来，就会试着把请求发送到远程服务，如果成功，断路器就关闭；如果失败断路器继续开启；这个流程就引出第一个概念，那就是滑动窗口

3.1、滑动窗口

在 hystrix 里，大量使用了 RxJava 这个响应式函数编程框架，滑动窗口的实现也是使用了 RxJava 框架。它其实就是一个 流量控制技术；竟然提到了滑动窗口，那就必须要提两上东西，一个是计数器，另一个就是滑动窗口;为了更通俗的理解计数器和滑动窗口关系，就以一个例子说明；假如有一个场景：要做一个请求限制，限制要求一分钟内最多只能有60个请求通过，这时最通用的做方就是用个计数器，计数一分钟内请求的次数，在这一分钟内每来一个请求计数器就加1；一分钟过后进入下一个一分钟时计数器就把计数归零重新计数；所以说如果要限流判断就只用判断这一分钟内的计数量就可以了，但这种做法在每个1分钟的临界值时是有问题的，问题是啥呢，假如说在0到58S时都没有请求，但是突然在第59S时一下子来了60个请求，在60S时再来60个请求，这个时候发生的情况是在相邻两秒内一下子来了120个请求，此时因为59S在第一个时间段；60S在第二个时间段，所以没有满足触发熔断条件，这就导至了相邻两秒间的请求量过了阈值，系统很可能炸了，为此引出了另一个玩意，那就是滑动窗口；滑动窗口把一分钟分成6个窗口，每个窗口是10S，红色框代表可以滑动的滑动窗口，黑色的窗口代表10S的统计数值，第一个10S统计 完成后红色滑动窗口会向前滑动一格，改成滑动窗口后他统计的就是红色滑动窗口内的访问量总和了

hystrix是通过滑动窗口统计的，他一共有10个窗口，每个窗口代表1S，所以他统计的是他10S内的数据

 

上图的每个小矩形代表一个桶，可以看到，每个桶都记录着1秒内的四个指标数据：成功量、失败量、超时量和拒绝量，这里的拒绝量指的就是上面流程图中【信号量/线程池资源检查】中被拒绝的流量。10个桶合起来是一个完整的滑动窗口，所以计算一个滑动窗口的总数据需要将10个桶的数据加起来。

 四、Hystrix熔断的源码分析

Hystrix熔断的@HystrixCommand注解，是通过HystrixCommandAspect这个切面来处理的。其中关注@Around注解声明的方法，它针对于请求合并，以及降级的注解进行代理。这里重点针对HystrixCommand这个注解进行详细分析。

• getMethodFromTarget 获取目标方法信息
• MetaHolder metaHolder = metaHolderFactory.create(joinPoint); 获取元数据，比如调用方法，HystrixProperty注解数据、方法参数等
• HystrixCommandFactory.getInstance().create 获取调用者，它持有一个命令对象，并且可以在合适的时候通过这个命令对象完成具体的业务逻辑
• execute，执行命令

@Around("hystrixCommandAnnotationPointcut() ||
hystrixCollapserAnnotationPointcut()")
public Object methodsAnnotatedWithHystrixCommand(final ProceedingJoinPoint
joinPoint) throws Throwable {
  Method method = getMethodFromTarget(joinPoint);
  //省略代码...
  MetaHolderFactory metaHolderFactory =
META_HOLDER_FACTORY_MAP.get(HystrixPointcutType.of(method));
  MetaHolder metaHolder = metaHolderFactory.create(joinPoint);
  //如果是异步，则创建GenericObservableCommand， 否则，则创建GenericCommand
  HystrixInvokable invokable =
HystrixCommandFactory.getInstance().create(metaHolder);
 
  ExecutionType executionType = metaHolder.isCollapserAnnotationPresent() ?
    metaHolder.getCollapserExecutionType() : metaHolder.getExecutionType();
  Object result;
  try {
    if (!metaHolder.isObservable()) { //是否是响应式的（由于我们这些都是同步的会走
这个逻辑）
     //默认是走这里面，用命令执行器去执行
      result = CommandExecutor.execute(invokable, executionType,
metaHolder);
   } else {
      result = executeObservable(invokable, executionType, metaHolder);
   }
 } catch (HystrixBadRequestException e) {
    throw e.getCause();
 } catch (HystrixRuntimeException e) {
    throw hystrixRuntimeExceptionToThrowable(metaHolder, e);
}
  return result;
}

点击进入 CommandExecutor类的execute方法，这个方法主要用来执行命令，从代码中可以看出这里有三个执行类型，分别是同步、异步、以及响应式。其中，响应式又分为Cold Observable（observable.toObservable()） 和 HotObservable（observable.observe()）

默认的executionType=SYNCHRONOUS ，同步请求。

• execute():同步执行，返回一个单一的对象结果，发生错误时抛出异常。
• queue():异步执行，返回一个 Future 对象，包含着执行结束后返回的单一结果。
• observe():这个方法返回一个 Observable 对象，它代表操作的多个结果，但是已经被订阅者消费掉了。
• toObservable():这个方法返回一个 Observable 对象，它代表操作的多个结果，需要咱们自己手动订阅并消费掉。

需要注意的是，Hystrix用到了RxJava这个框架，它是一个响应式编程框架，在Android里面用得比较多 

    public static Object execute(HystrixInvokable invokable, ExecutionType executionType, MetaHolder metaHolder) throws RuntimeException {
        Validate.notNull(invokable);
        Validate.notNull(metaHolder);
         
        switch (executionType) {

            case SYNCHRONOUS: {
                return castToExecutable(invokable, executionType).execute();
            }
            case ASYNCHRONOUS: {
                HystrixExecutable executable = castToExecutable(invokable, executionType);
                if (metaHolder.hasFallbackMethodCommand()
                        && ExecutionType.ASYNCHRONOUS == metaHolder.getFallbackExecutionType()) {
                    return new FutureDecorator(executable.queue());
                }
                return executable.queue();
            }
            case OBSERVABLE: {
                HystrixObservable observable = castToObservable(invokable);
                return ObservableExecutionMode.EAGER == metaHolder.getObservableExecutionMode() ? observable.observe() : observable.toObservable();
            }
            default:
                throw new RuntimeException("unsupported execution type: " + executionType);
        }
    }

因为是走默认的，所以进入HystrixCommand类的execute()方法；这个方法中，首先调用queue()，这个方法会返回一个future对象。

  public R execute() {
        try {
            return queue().get();
        } catch (Exception e) {
            throw Exceptions.sneakyThrow(decomposeException(e));
        }
    }

queue这个方法中，返回了一个Future对象，这个future对象的实现是f，f是以匿名内部类，它是Java.util.concurrent中定一个的一个异步带返回值对象。当调用queue().get()方法时，最终是委派给了delegate.get 方法。

public Future<R> queue() {
  /*
    * The Future returned by Observable.toBlocking().toFuture() does not
implement the
    * interruption of the execution thread when the "mayInterrupt" flag of
Future.cancel(boolean) is set to true;
    * thus, to comply with the contract of Future, we must wrap around it.
    */
  final Future<R> delegate = toObservable().toBlocking().toFuture();
  final Future<R> f = new Future<R>() {
    @Override
    public boolean cancel(boolean mayInterruptIfRunning) {
      if (delegate.isCancelled()) {
        return false;
     }
      if
(HystrixCommand.this.getProperties().executionIsolationThreadInterruptOnFutureCa
ncel().get()) {
        /*
          * The only valid transition here is false -> true. If there
are two futures, say f1 and f2, created by this command
          * (which is super-weird, but has never been prohibited),
and calls to f1.cancel(true) and to f2.cancel(false) are
          * issued by different threads, it's unclear about what
value would be used by the time mayInterruptOnCancel is checked.
          * The most consistent way to deal with this scenario is to
say that if *any* cancellation is invoked with interruption,
          * than that interruption request cannot be taken back.
          */
        interruptOnFutureCancel.compareAndSet(false,
mayInterruptIfRunning);
     }
      final boolean res = delegate.cancel(interruptOnFutureCancel.get());
      if (!isExecutionComplete() && interruptOnFutureCancel.get()) {
        final Thread t = executionThread.get();
        if (t != null && !t.equals(Thread.currentThread())) {
t.interrupt();
       }
     }
      return res;
   }
    @Override
    public boolean isCancelled() {
      return delegate.isCancelled();
   }
    @Override
    public boolean isDone() {
      return delegate.isDone();
   }
    //最终会调用此方法
    @Override
    public R get() throws InterruptedException, ExecutionException {
      return delegate.get();
   }
    @Override
    public R get(long timeout, TimeUnit unit) throws InterruptedException,
ExecutionException, TimeoutException {
      return delegate.get(timeout, unit);
   }
 };
  /* special handling of error states that throw immediately */
  if (f.isDone()) {
    try {
      f.get();
      return f;
   } catch (Exception e) {
      Throwable t = decomposeException(e);
      if (t instanceof HystrixBadRequestException) {
        return f;
     } else if (t instanceof HystrixRuntimeException) {
        HystrixRuntimeException hre = (HystrixRuntimeException) t;
        switch (hre.getFailureType()) {
          case COMMAND_EXCEPTION:
          case TIMEOUT:
            // we don't throw these types from queue() only from
queue().get() as they are execution errors
            return f;
          default:
            // these are errors we throw from queue() as they as
rejection type errors
            throw hre;
       }
     } else {
        throw Exceptions.sneakyThrow(t);
     }
   }
 }
return f;
}

因为最终是委派给了delegate.get 方法执行，而delegate在开头final Future<R> delegate = toObservable().toBlocking().toFuture();中，所以进入toObservable()方法中，在RxJava中，分为几种角色

• Observable（被观察者），它的主要作用是产生事件
• Observer（观察者），它的作用是接收事件并作出相应
• Subscribe（订阅），它用来连接被观察者和观察者
• Event（事件），被观察者、观察者、沟通的载体

在queue中，调用toObservable()方法创建一个被观察者。通过Observable定义一个被观察者，这个被观察者会被toObservable().toBlocking().toFuture() ，实际上就是返回可获得 run() 抽象方法执行结果的Future 。 run() 方法由子类实现，执行正常的业务逻辑。在下面这段代码中，当存在subscriber时，便会调用Func0#call() 方法，而这个subscriber是在 toBlocking() 中被订阅的。

• 调用 isRequestCachingEnabled(); 判断请求结果缓存功能是否开启，如果开启并且命中了缓存，则会以Observable形式返回一个缓存结果
• 创建执行命令的Observable： hystrixObservable
• 当缓存处于开启状态并且没有命中缓存时，则创建一个“订阅了执行命令的Observable”：HystrixCommandResponseFromCache
  • 创建存储到缓存的Observable： HystrixCachedObservable
  • 将toCache添加到缓存中，返回获取缓存的Observable：fromCache
  • 如果添加失败: fromCache!=null, 则调用 toCache.unsubscribe() 方法，取消HystrixCachedObservable 的订阅
  • 如果添加成功，则调用 toCache.toObservable(); 获得缓存Observable
• 当缓存特性没有开启时，则返回执行命令的Observable。

   public Observable<R> toObservable() {
        final AbstractCommand<R> _cmd = this;

        //doOnCompleted handler already did all of the SUCCESS work
        //doOnError handler already did all of the FAILURE/TIMEOUT/REJECTION/BAD_REQUEST work
        final Action0 terminateCommandCleanup = new Action0() {

            @Override
            public void call() {
                if (_cmd.commandState.compareAndSet(CommandState.OBSERVABLE_CHAIN_CREATED, CommandState.TERMINAL)) {
                    handleCommandEnd(false); //user code never ran
                } else if (_cmd.commandState.compareAndSet(CommandState.USER_CODE_EXECUTED, CommandState.TERMINAL)) {
                    handleCommandEnd(true); //user code did run
                }
            }
        };

        //mark the command as CANCELLED and store the latency (in addition to standard cleanup)
        final Action0 unsubscribeCommandCleanup = new Action0() {
            @Override
            public void call() {
                if (_cmd.commandState.compareAndSet(CommandState.OBSERVABLE_CHAIN_CREATED, CommandState.UNSUBSCRIBED)) {
                    if (!_cmd.executionResult.containsTerminalEvent()) {
                        _cmd.eventNotifier.markEvent(HystrixEventType.CANCELLED, _cmd.commandKey);
                        try {
                            executionHook.onUnsubscribe(_cmd);
                        } catch (Throwable hookEx) {
                            logger.warn("Error calling HystrixCommandExecutionHook.onUnsubscribe", hookEx);
                        }
                        _cmd.executionResultAtTimeOfCancellation = _cmd.executionResult
                                .addEvent((int) (System.currentTimeMillis() - _cmd.commandStartTimestamp), HystrixEventType.CANCELLED);
                    }
                    handleCommandEnd(false); //user code never ran
                } else if (_cmd.commandState.compareAndSet(CommandState.USER_CODE_EXECUTED, CommandState.UNSUBSCRIBED)) {
                    if (!_cmd.executionResult.containsTerminalEvent()) {
                        _cmd.eventNotifier.markEvent(HystrixEventType.CANCELLED, _cmd.commandKey);
                        try {
                            executionHook.onUnsubscribe(_cmd);
                        } catch (Throwable hookEx) {
                            logger.warn("Error calling HystrixCommandExecutionHook.onUnsubscribe", hookEx);
                        }
                        _cmd.executionResultAtTimeOfCancellation = _cmd.executionResult
                                .addEvent((int) (System.currentTimeMillis() - _cmd.commandStartTimestamp), HystrixEventType.CANCELLED);
                    }
                    handleCommandEnd(true); //user code did run
                }
            }
        };

        final Func0<Observable<R>> applyHystrixSemantics = new Func0<Observable<R>>() {
            @Override
            public Observable<R> call() {
                if (commandState.get().equals(CommandState.UNSUBSCRIBED)) {
                    return Observable.never();
                }
                return applyHystrixSemantics(_cmd);
            }
        };

        final Func1<R, R> wrapWithAllOnNextHooks = new Func1<R, R>() {
            @Override
            public R call(R r) {
                R afterFirstApplication = r;

                try {
                    afterFirstApplication = executionHook.onComplete(_cmd, r);
                } catch (Throwable hookEx) {
                    logger.warn("Error calling HystrixCommandExecutionHook.onComplete", hookEx);
                }

                try {
                    return executionHook.onEmit(_cmd, afterFirstApplication);
                } catch (Throwable hookEx) {
                    logger.warn("Error calling HystrixCommandExecutionHook.onEmit", hookEx);
                    return afterFirstApplication;
                }
            }
        };

        final Action0 fireOnCompletedHook = new Action0() {
            @Override
            public void call() {
                try {
                    executionHook.onSuccess(_cmd);
                } catch (Throwable hookEx) {
                    logger.warn("Error calling HystrixCommandExecutionHook.onSuccess", hookEx);
                }
            }
        };
return Observable.defer(new Func0<Observable<R>>() {
  @Override
  public Observable<R> call() {
    /* this is a stateful object so can only be used once */
    /* CAS保证命令只执行一次 */
    if (!commandState.compareAndSet(CommandState.NOT_STARTED,
CommandState.OBSERVABLE_CHAIN_CREATED)) {
      IllegalStateException ex = new IllegalStateException("This instance
can only be executed once. Please instantiate a new instance.");
      //TODO make a new error type for this
      throw new HystrixRuntimeException(FailureType.BAD_REQUEST_EXCEPTION,
_cmd.getClass(), getLogMessagePrefix() + " command executed multiple times -
this is not permitted.", ex, null);
   }
   // 命令开始时间戳
    commandStartTimestamp = System.currentTimeMillis();
    // 打印日志
    if (properties.requestLogEnabled().get()) {
      // log this command execution regardless of what happened
      if (currentRequestLog != null) {
currentRequestLog.addExecutedCommand(_cmd);
     }
   }
   // 缓存开关，缓存KEY（这个是Hystrix中请求缓存功能，hystrix支持将一个请求结果缓存起
来，下一个具有相同key的请求将直接从缓存中取出结果，减少请求开销）
    final boolean requestCacheEnabled = isRequestCachingEnabled();
    final String cacheKey = getCacheKey();
    /* try from cache first */
    if (requestCacheEnabled) {//如果开启了缓存机制，则从缓存中获取结果
      HystrixCommandResponseFromCache<R> fromCache =
(HystrixCommandResponseFromCache<R>) requestCache.get(cacheKey);
      if (fromCache != null) {
        isResponseFromCache = true;
        return handleRequestCacheHitAndEmitValues(fromCache, _cmd);
     }
   }
    // 声明执行命令的Observable
    Observable<R> hystrixObservable =
      Observable.defer(applyHystrixSemantics)
     .map(wrapWithAllOnNextHooks);
    Observable<R> afterCache;
    //保存请求结果到缓存中
    if (requestCacheEnabled && cacheKey != null) {
      // wrap it for caching
      HystrixCachedObservable<R> toCache =
HystrixCachedObservable.from(hystrixObservable, _cmd);
      HystrixCommandResponseFromCache<R> fromCache =
(HystrixCommandResponseFromCache<R>) requestCache.putIfAbsent(cacheKey,
toCache);
      if (fromCache != null) {
        // another thread beat us so we'll use the cached value instead
        toCache.unsubscribe();
        isResponseFromCache = true;
        return handleRequestCacheHitAndEmitValues(fromCache, _cmd);
     } else {
        // we just created an ObservableCommand so we cast and return it
        afterCache = toCache.toObservable();
     }
   } else {
      afterCache = hystrixObservable;
   }
    return afterCache
     .doOnTerminate(terminateCommandCleanup)   // perform cleanup once
(either on normal terminal state (this line), or unsubscribe (next line))
     .doOnUnsubscribe(unsubscribeCommandCleanup) // perform cleanup once
     .doOnCompleted(fireOnCompletedHook);
 }
});

执行命令的Observable的定义如下，通过defer定义了一个 applyHystrixSemantics 的事件。

final Func0<Observable<R>> applyHystrixSemantics = new Func0<Observable<R>>() {
  @Override
 public Observable<R> call() {
    // 当commandState处于UNSUBSCRIBED时，不执行命令
    if (commandState.get().equals(CommandState.UNSUBSCRIBED)) {
      return Observable.never();
   }
    //返回执行命令的Observable
    return applyHystrixSemantics(_cmd);
 }
};
Observable<R> hystrixObservable =
  Observable.defer(applyHystrixSemantics)
 .map(wrapWithAllOnNextHooks);

applyHystrixSemantics方法；假设缓存特性未开启或者未命中缓存，那么代码将执行 applyHystrixSemantics 。

• 传入的_cmd是一个GenericCommand，最终执行这个command中的run方法，本质就是完成对queryOrder方法的代理
• circuitBreaker.allowRequest() 如果为true，表示当前不处于熔断状态，正常执行，否则，调用 handleShortCircuitViaFallback 实现服务降级，如果我们配置了fallback方法，则会获得我们配置的fallback执行
• 如果当前hystrix处于未熔断状态，则

• • getExecutionSemaphore 判断当前策略是否为信号量（TryableSemaphoreNoOp/TryableSemaphoreActual），如果是，则调用 tryAcquire 来获取信号量。如果当前信号量满了，则调用 handleSemaphoreRejectionViaFallback 方法。
  • 调用 executeCommandAndObserve 获取命令执行Observable。

   private Observable<R> applyHystrixSemantics(final AbstractCommand<R> _cmd) {
        // mark that we're starting execution on the ExecutionHook
        // if this hook throws an exception, then a fast-fail occurs with no fallback.  No state is left inconsistent
        executionHook.onStart(_cmd);

        /* determine if we're allowed to execute */
        if (circuitBreaker.allowRequest()) {
            final TryableSemaphore executionSemaphore = getExecutionSemaphore();
            final AtomicBoolean semaphoreHasBeenReleased = new AtomicBoolean(false);
            final Action0 singleSemaphoreRelease = new Action0() {
                @Override
                public void call() {
                    if (semaphoreHasBeenReleased.compareAndSet(false, true)) {
                        executionSemaphore.release();
                    }
                }
            };

            final Action1<Throwable> markExceptionThrown = new Action1<Throwable>() {
                @Override
                public void call(Throwable t) {
                    eventNotifier.markEvent(HystrixEventType.EXCEPTION_THROWN, commandKey);
                }
            };

            if (executionSemaphore.tryAcquire()) {
                try {
                    /* used to track userThreadExecutionTime */
                    executionResult = executionResult.setInvocationStartTime(System.currentTimeMillis());
                   //跟进
                    return executeCommandAndObserve(_cmd)
                            .doOnError(markExceptionThrown)
                            .doOnTerminate(singleSemaphoreRelease)
                            .doOnUnsubscribe(singleSemaphoreRelease);
                } catch (RuntimeException e) {
                    return Observable.error(e);
                }
            } else {
                return handleSemaphoreRejectionViaFallback();
            }
        } else {
            return handleShortCircuitViaFallback();
        }
    }

executeCommandAndObserve

• 定义不同的回调，doOnNext、doOnCompleted、onErrorResumeNext、doOnEach。
• 调用executeCommandWithSpecifiedIsolation获得执行命令的Observable
• 若执行命令超时特性开启，调用 Observable.lift 方法实现执行命令超时功能。

private Observable<R> executeCommandAndObserve(final AbstractCommand<R> _cmd) {
  final HystrixRequestContext currentRequestContext =
HystrixRequestContext.getContextForCurrentThread();
  //Action和Func都是定义的一个动作，Action是无返回值，Func是有返回值
  // doOnNext中的回调。即命令执行之前执行的操作
  final Action1<R> markEmits //...
// doOnCompleted中的回调。命令执行完毕后执行的操作
  final Action0 markOnCompleted = //...
// onErrorResumeNext中的回调。命令执行失败后的回退逻辑
  final Func1<Throwable, Observable<R>> handleFallback = //...
// doOnEach中的回调。`Observable`每发射一个数据都会执行这个回调，设置请求上下文
  final Action1<Notification<? super R>> setRequestContext =//...
  Observable<R> execution;
  if (properties.executionTimeoutEnabled().get()) {
    execution = executeCommandWithSpecifiedIsolation(_cmd)
      .lift(new HystrixObservableTimeoutOperator<R>(_cmd));
  } else {
    execution = executeCommandWithSpecifiedIsolation(_cmd);
  }
  return execution.doOnNext(markEmits)
.doOnCompleted(markOnCompleted)
    .onErrorResumeNext(handleFallback)
    .doOnEach(setRequestContext);
}

executeCommandWithSpecifiedIsolation方法，这个方法首先是根据当前不同的资源隔离策略执行不同的逻辑，THREAD、SEMAPHORE，这里就不展开实现细节，我们直接看执行的方法 getUserExecutionObservable 。

private Observable<R> executeCommandWithSpecifiedIsolation(final
AbstractCommand<R> _cmd) {
  if (properties.executionIsolationStrategy().get() ==
ExecutionIsolationStrategy.THREAD) {
    // mark that we are executing in a thread (even if we end up being
rejected we still were a THREAD execution and not SEMAPHORE)
    return Observable.defer(new Func0<Observable<R>>() {
      @Override
      public Observable<R> call() {
        executionResult = executionResult.setExecutionOccurred();
        if
(!commandState.compareAndSet(CommandState.OBSERVABLE_CHAIN_CREATED,
CommandState.USER_CODE_EXECUTED)) {
          return Observable.error(new IllegalStateException("execution
attempted while in state : " + commandState.get().name()));
       }
        metrics.markCommandStart(commandKey, threadPoolKey,
ExecutionIsolationStrategy.THREAD);
        if (isCommandTimedOut.get() == TimedOutStatus.TIMED_OUT) {
          // the command timed out in the wrapping thread so we will
return immediately
          // and not increment any of the counters below or other such
logic
          return Observable.error(new RuntimeException("timed out
before executing run()"));
       }
        if (threadState.compareAndSet(ThreadState.NOT_USING_THREAD,
ThreadState.STARTED)) {
          //we have not been unsubscribed, so should proceed
          HystrixCounters.incrementGlobalConcurrentThreads();
          threadPool.markThreadExecution();
          // store the command that is being run
          endCurrentThreadExecutingCommand =
Hystrix.startCurrentThreadExecutingCommand(getCommandKey());
          executionResult = executionResult.setExecutedInThread();
          /**
            * If any of these hooks throw an exception, then it
appears as if the actual execution threw an error
            */
          try {
            executionHook.onThreadStart(_cmd);
            executionHook.onRunStart(_cmd);
            executionHook.onExecutionStart(_cmd);
            return getUserExecutionObservable(_cmd);
} catch (Throwable ex) {
            return Observable.error(ex);
         }
       } else {
          //command has already been unsubscribed, so return
immediately
          return Observable.error(new RuntimeException("unsubscribed
before executing run()"));
       }
     }
   }).doOnTerminate(new Action0() {
      @Override
      public void call() {
        if (threadState.compareAndSet(ThreadState.STARTED,
ThreadState.TERMINAL)) {
          handleThreadEnd(_cmd);
       }
        if (threadState.compareAndSet(ThreadState.NOT_USING_THREAD,
ThreadState.TERMINAL)) {
          //if it was never started and received terminal, then no
need to clean up (I don't think this is possible)
       }
        //if it was unsubscribed, then other cleanup handled it
     }
   }).doOnUnsubscribe(new Action0() {
      @Override
      public void call() {
        if (threadState.compareAndSet(ThreadState.STARTED,
ThreadState.UNSUBSCRIBED)) {
          handleThreadEnd(_cmd);
       }
        if (threadState.compareAndSet(ThreadState.NOT_USING_THREAD,
ThreadState.UNSUBSCRIBED)) {
          //if it was never started and was cancelled, then no need to
clean up
       }
        //if it was terminal, then other cleanup handled it
     }
   }).subscribeOn(threadPool.getScheduler(new Func0<Boolean>() {
      @Override
      public Boolean call() {
        return
properties.executionIsolationThreadInterruptOnTimeout().get() &&
_cmd.isCommandTimedOut.get() == TimedOutStatus.TIMED_OUT;
     }
   }));
 } else {
    return Observable.defer(new Func0<Observable<R>>() {
      @Override
      public Observable<R> call() {
        executionResult = executionResult.setExecutionOccurred();
        if
(!commandState.compareAndSet(CommandState.OBSERVABLE_CHAIN_CREATED,
CommandState.USER_CODE_EXECUTED)) {
          return Observable.error(new IllegalStateException("execution
attempted while in state : " + commandState.get().name()));
       }
metrics.markCommandStart(commandKey, threadPoolKey,
ExecutionIsolationStrategy.SEMAPHORE);
        // semaphore isolated
        // store the command that is being run
        endCurrentThreadExecutingCommand =
Hystrix.startCurrentThreadExecutingCommand(getCommandKey());
        try {
          executionHook.onRunStart(_cmd);
          executionHook.onExecutionStart(_cmd);
          return getUserExecutionObservable(_cmd);  //the
getUserExecutionObservable method already wraps sync exceptions, so this
shouldn't throw
       } catch (Throwable ex) {
          //If the above hooks throw, then use that as the result of
the run method
          return Observable.error(ex);
       }
     }
   });
 }
}

调用 getExecutionObservable 方法创建 命令执行Observable 。 getExecutionObservable 方法是个抽象方法， HystrixCommand 实现了该方法。

private Observable<R> getUserExecutionObservable(final AbstractCommand<R> _cmd)
{
  Observable<R> userObservable;
  try {
    userObservable = getExecutionObservable();
 } catch (Throwable ex) {
    // the run() method is a user provided implementation so can throw
instead of using Observable.onError
    // so we catch it here and turn it into Observable.error
    userObservable = Observable.error(ex);
 }
  return userObservable
   .lift(new ExecutionHookApplication(_cmd))
   .lift(new DeprecatedOnRunHookApplication(_cmd));
}

调用 HystrixCommand.getExecutionObservable 方法创建命令执行Observable这里最终调用的是run方法，通过Observable.just， just是RxJava中的一个操作符，它可以接受一个或者多个参数来创建一个Observable对象。而这个run（）方法是一个抽象方法，在HystrixCommand中并没有实现，而是在子类中实现，而此时传递的cmd=GenricCommand正好实现了HystrixCommand，重写了run方法。

@Override
final protected Observable<R> getExecutionObservable() {
  return Observable.defer(new Func0<Observable<R>>() {
    @Override
    public Observable<R> call() {
      try {
        return Observable.just(run());
     } catch (Throwable ex) {
        return Observable.error(ex);
     }
   }
 }).doOnSubscribe(new Action0() {
    @Override
    public void call() {
      // Save thread on which we get subscribed so that we can interrupt
it later if needed
      executionThread.set(Thread.currentThread());
   }
 });
}

enericCommand类中的run方法，这里的实现和前面自定义的 HystrixCommandService 实现是一样的，同样是集成HystrixCommand，重写run方法。这里也是如此。

• 首先调用 getCommandAction() 方法获取 CommandAction ，我们的示例中获取到的是MethodExecutionAction 。
• 然后调用 MethodExecutionAction.execute 方法，传入 ExecutionType 参数，我们的示例中传入的是 ExecutionType.SYNCHRONOUS 。

@Override
protected Object run() throws Exception {
  LOGGER.debug("execute command: {}", getCommandKey().name());
  return process(new Action() {
    @Override
    Object execute() {
      return getCommandAction().execute(getExecutionType());
   }
 });
}