ribbon 详解

ribbon 详解

1. 顶层架构

2. 简单的示例：使用ResourceTemplate方式

@Test
public void testGroup(){
    HttpResourceGroup httpResourceGroup = Ribbon.createHttpResourceGroup("test",
            ClientOptions.create().withMaxAutoRetries(3).
                    withConfigurationBasedServerList("localhost:8081,localhost:8080"));
    HttpRequestTemplate<ByteBuf> recommendationsByUserIdTemplate = httpResourceGroup.newTemplateBuilder("recommendationsByUserId", ByteBuf.class)
            .withMethod("GET")
            .withUriTemplate("/aa/index")
            .withHeader("X-Auth-Token", "abc")
            //.withFallbackProvider(new RecommendationServiceFallbackHandler())
            //.withResponseValidator(new RecommendationServiceResponseValidator())
            .build();

    RibbonRequest<ByteBuf> request = recommendationsByUserIdTemplate.requestBuilder()
            .withRequestProperty("userId", "test")
            .build();
    ByteBuf buf = request.execute();
    byte[] bytes = new byte[buf.capacity()];
    buf.readBytes(bytes);
    System.out.println(new String(bytes));
}

3.实现：

3.1 关键对象

public interface RibbonRequest<T> {

    /**
     * Blocking API that returns a single (or last element if there is a sequence of objects from the execution) element
     */
    public T execute();
    
    /**
     * Non blocking API that returns a {@link Future}, where its {@link Future#get()} method is blocking and returns a 
     * single (or last element if there is a sequence of objects from the execution) element
     */
    public Future<T> queue();

    /**
     * Non blocking API that returns an {@link Observable} while the execution is started asynchronously.
     * Subscribing to the returned {@link Observable} is guaranteed to get the complete sequence from 
     * the beginning, which might be replayed by the framework. Use this API for "fire and forget".
     */
    public Observable<T> observe();
    
    /**
     * Non blocking API that returns an Observable. The execution is not started until the returned Observable is subscribed to.
     */
    public Observable<T> toObservable();

    /**
     * Create a decorated {@link RequestWithMetaData} where you can call its similar blocking or non blocking 
     * APIs to get {@link RibbonResponse}, which in turn contains returned object(s) and 
     * some meta data from Hystrix execution.
     */
    public RequestWithMetaData<T> withMetadata();
}

明显，最终生成的关键对象RibbonRequest,使用的时观察者模式，底层实现肯定会使用RxJava或者Hystrix。

3.2 关键实现：HystrixObservableCommandChain

HystrixObservableCommandChain<T> createHystrixCommandChain() {
    List<HystrixObservableCommand<T>> commands = new ArrayList<HystrixObservableCommand<T>>(2);
    if (cacheProvider != null) {
        commands.add(new CacheObservableCommand<T>(cacheProvider.getCacheProvider(), cacheProvider.getKey(), cacheHystrixCacheKey,
                requestProperties, template.cacheHystrixProperties()));
    }
    commands.add(new HttpResourceObservableCommand<T>(client, httpRequest, hystrixCacheKey, requestProperties, template.fallbackHandler(),
            template.responseValidator(), template.getClassType(), template.hystrixProperties()));

    return new HystrixObservableCommandChain<T>(commands);
}

在HystrixObservableCommandChain执行toObservable的时候，会依次便利集合中所有的HystrixObservableCommand,知道其toObservable不为null。
所以，如果cacheProvider不为null会调用CacheObservableCommand的toObservable,如果cacheProvider为null，则直接调用HttpResourceObservableCommand的toObservable.

3.3 http请求：

Observable<HttpClientResponse<ByteBuf>> httpResponseObservable = httpClient.submit(httpRequest);

Observable<T> o = 
  (server == null ? selectServer() : Observable.just(server))
      .concatMap(new Func1<Server, Observable<T>>() {
        @Override
        // Called for each server being selected
        public Observable<T> call(Server server) {
          context.setServer(server);
          final ServerStats stats = loadBalancerContext.getServerStats(server);
          
          // Called for each attempt and retry
          Observable<T> o = Observable
          .just(server)
          .concatMap(new Func1<Server, Observable<T>>() {
            ...
            });

            if (maxRetrysSame > 0) 
            o = o.retry(retryPolicy(maxRetrysSame, true));
            return o;
          }
        }
if (maxRetrysSame > 0) 
  o = o.retry(retryPolicy(maxRetrysSame, true));
      return o;              
                                

会调用统一服务器n次，后调用下一个服务器m次（n,m为设定的最大调用次数）。

4.简单示例：使用注解方式

public interface AAIndex {
    @Http(  method = Http.HttpMethod.GET,
            uri = "http://localhost:8080/aa/index")
    RibbonRequest<ByteBuf> index();
}

@Test
public void testRibbonAnnotation(){
    AAIndex aaIndex = Ribbon.from(AAIndex.class);
    ByteBuf buf = aaIndex.index().execute();
    byte[] bytes = new byte[buf.capacity()];
    buf.readBytes(bytes);
    System.out.println(new String(bytes));
}

5.实现：

5.1 动态代理：

return (T) Proxy.newProxyInstance(
        Thread.currentThread().getContextClassLoader(),
        new Class[]{clientInterface, ProxyLifeCycle.class},
        new RibbonDynamicProxy<T>(clientInterface, resourceGroupFactory, configFactory, transportFactory, processors)
);

使用JDK自带的代理类实现，代理了client接口，和ProxyLifeCycle,其实现具体类为RibbonDynamicProxy。

5.2 支持的注解：

static void registerAnnotationProcessors(AnnotationProcessorsProvider processors) {
    processors.register(new HttpAnnotationProcessor());
    processors.register(new HystrixAnnotationProcessor());
    processors.register(new CacheProviderAnnotationProcessor());
    processors.register(new ClientPropertiesProcessor());
}

最终的实现跟非注解方式的实现是一致的。

@SuppressWarnings("unchecked")
public <O> RibbonRequest<O> executeFromTemplate(Object[] args) {
    HttpRequestBuilder<?> requestBuilder = httpRequestTemplateBuilder.build().requestBuilder();
    withParameters(requestBuilder, args);
    withContent(requestBuilder, args);
    return (RibbonRequest<O>) requestBuilder.build();
}