AsyncRestTemplate ListenableFuture<T>回调+超时

Question

我配置了一个像这样的AsyncRestTemplate，这里只是一个例子，说明我使用的是一个HttpComponentsAsyncClientHttpRequestFactory，其中connectTimeout和readTimeout都是用值初始化的--使用Spring4.0.8发行版：

    <bean id="myAsynchRequestFactory" class="org.springframework.http.client.HttpComponentsAsyncClientHttpRequestFactory">
        <property name="httpAsyncClient" ref="myCloseableHttpAsynchClient"></property>
        <property name="connectTimeout" value="444"></property>
        <property name="readTimeout" value="555"></property>
    </bean>

    <bean id="myAsynchRestTemplate" class="org.springframework.web.client.AsyncRestTemplate">
        <constructor-arg>
            <ref bean="myAsynchRequestFactory"/>
        </constructor-arg>
    </bean>

我还配置了一个RequestConfig，因为spring代码建议不再推荐前一个方法，因此我按如下方式添加了它：

<bean id="myRequestConfigBuilder" class="org.apache.http.client.config.RequestConfig" factory-method="custom">
    <property name="connectionRequestTimeout" value="555"></property>
    <property name="connectTimeout" value="400"></property>
    <property name="socketTimeout" value="555"></property>
</bean>

<bean id="myRequestConfig" factory-bean="myRequestConfigBuilder" factory-method="build">
</bean>

<bean id="myHttpAsyncClientBuilder" class="org.apache.http.impl.nio.client.HttpAsyncClientBuilder">
    <property name="connectionManager" ref="myHttpAsyncConnectionManager"></property>
    <property name="defaultRequestConfig" ref="myRequestConfig"></property>
</bean>

<bean id="myHttpAsyncClientBuilder" class="org.apache.http.impl.nio.client.HttpAsyncClientBuilder">
    <property name="connectionManager" ref="myHttpAsyncConnectionManager"></property>
    <property name="defaultRequestConfig" ref="myRequestConfig"></property>
</bean>
<bean id="myCloseableHttpAsynchClient" factory-bean="myHttpAsyncClientBuilder" factory-method="build">
</bean>

现在，我使用AsyncRestTemplate和它的addCallback()方法(如图所示)：

  response = myAsynchRestTemplate.getForEntity( ... );
  response.addCallback(new org.springframework.util.concurrent.ListenableFutureCallback<ResponseEntity<?>>() {

                    @Override
                    public void onSuccess(ResponseEntity<?> result) {
                        System.out.println("SUCCESS");
                    }

                    @Override
                    public void onFailure(Throwable ex) {
                        System.out.println("FAILURE")
                    }
                }

我希望onSuccess(.)当服务超过555 is才能回复时，将不会调用。但这是不可能的。即使我的服务假装延迟了5000 my，无论如何也会调用onSuccess方法。

我在网上搜索了一些可能的解决方案，以便在回调中添加某种超时，但没有找到合适的解决方案。我试着在自4.0以来的任何版本中阅读org.Spring Frawork.util.Concurr.*代码，但是还没有找到任何可以阻止注册回调执行的内容。

我已经看到了以下Stackoverflow问题：

How to cancel AsyncRestTemplate HTTP request if they are taking too much time?这个建议超时应该能工作

这一个是使用Google的例子，但我使用的是org.springframework.util.concurrent.ListenableFuture

有谁能解决这个问题吗？

更新

我找到了这不起作用的根本原因..。

HttpComponentsAsyncClientHttpRequestFactory不接受我提供的RequestConfig。相反，它总是空的，并且被重新创建为RequestConfig.DEFAULT。

来自HttpComponentesAsyncClientHttpRequestFactory：

@Override
public AsyncClientHttpRequest createAsyncRequest(URI uri, HttpMethod httpMethod) throws IOException {
    HttpAsyncClient asyncClient = getHttpAsyncClient();
    startAsyncClient();
    HttpUriRequest httpRequest = createHttpUriRequest(httpMethod, uri);
    postProcessHttpRequest(httpRequest);



    /** HERE: It tries to create a HttpContext **/
    HttpContext context = createHttpContext(httpMethod, uri);




    if (context == null) {
        context = HttpClientContext.create();
    }
    // Request configuration not set in the context
    if (context.getAttribute(HttpClientContext.REQUEST_CONFIG) == null) {
        // Use request configuration given by the user, when available
        RequestConfig config = null;
        if (httpRequest instanceof Configurable) {
            config = ((Configurable) httpRequest).getConfig();
        }
        if (config == null) {
            config = RequestConfig.DEFAULT;
        }
        context.setAttribute(HttpClientContext.REQUEST_CONFIG, config);
    }
    return new HttpComponentsAsyncClientHttpRequest(asyncClient, httpRequest, context);
}

但是方法createHttpContext(httpMethod，uri)总是返回null：

/**
 * Template methods that creates a {@link HttpContext} for the given HTTP method and URI.
 * <p>The default implementation returns {@code null}.
 * @param httpMethod the HTTP method
 * @param uri the URI
 * @return the http context
 */
protected HttpContext createHttpContext(HttpMethod httpMethod, URI uri) {
    return null;
}

这样，就可以重新创建RequestConfig.DEFAULT并获得附加的HttpContext。

现在，HttpComponentsClientHttpRequestFactory. 将这个与非同步的版本进行比较。这一次使用超时重新创建RequestConfig：

@Override
public ClientHttpRequest createRequest(URI uri, HttpMethod httpMethod) throws IOException {
    CloseableHttpClient client = (CloseableHttpClient) getHttpClient();
    Assert.state(client != null, "Synchronous execution requires an HttpClient to be set");
    HttpUriRequest httpRequest = createHttpUriRequest(httpMethod, uri);
    postProcessHttpRequest(httpRequest);
    HttpContext context = createHttpContext(httpMethod, uri);
    if (context == null) {
        context = HttpClientContext.create();
    }
    // Request configuration not set in the context
    if (context.getAttribute(HttpClientContext.REQUEST_CONFIG) == null) {
        // Use request configuration given by the user, when available
        RequestConfig config = null;
        if (httpRequest instanceof Configurable) {
            config = ((Configurable) httpRequest).getConfig();
        }
        if (config == null) {




            /** LOOK HERE - THE SYNC WORLD HAS THIS WORKAROUND */
            if (this.socketTimeout > 0 || this.connectTimeout > 0) {
                config = RequestConfig.custom()
                        .setConnectTimeout(this.connectTimeout)
                        .setSocketTimeout(this.socketTimeout)
                        .build();
            }



            else {
                config = RequestConfig.DEFAULT;
            }
        }
        context.setAttribute(HttpClientContext.REQUEST_CONFIG, config);
    }
    if (this.bufferRequestBody) {
        return new HttpComponentsClientHttpRequest(client, httpRequest, context);
    }
    else {
        return new HttpComponentsStreamingClientHttpRequest(client, httpRequest, context);
    }
}

当我使用SimpleClientHttpRequestFactory代替时，我配置的超时将被接受，而不是onSuccess(.)，onFailure(.)方法是用SocketTimeoutException调用的。

这完全令人费解--也许任何人都能给出一个提示，为什么HttpComponentsAsyncClientHttpRequestFactory是按原样实现的？

是否有正确使用RequestConfig对象的示例？

Answer 1

这个问题将在

https://jira.spring.io/browse/SPR-12540

在Spring4.3中有一个解决方案(当前的Spring是4.2)，此时的解决方案是子类HttpComponentsAsyncClientHttpRequestFactory和@重写getHttpContext方法。