HttpClient4 遇到的问题2：httpclient的两个重要的参数maxPerRoute及MaxTotal

httpclient封装了java中进行http网络请求的底层实现，是一个被广泛使用的组件。

httpclient是支持池化机制的，这两个参数就是表示池化设置的。

服务之间发送http请求常用的有Apache的Fluent以及spring的restTemplate等

Apache的Fluent，以及spring的restTemplate都是对httpclient进行了封装

 

以Apache的Fluent为例说明

其Executor类

 
/**
 * An Executor for fluent requests
 * <p/>
 * A {@link PoolingHttpClientConnectionManager} with maximum 100 connections per route and
 * a total maximum of 200 connections is used internally.
 */

最大100 connections per route 以及 最大200个 connection

 
CONNMGR = new PoolingHttpClientConnectionManager(sfr);
 
CONNMGR.setDefaultMaxPerRoute(100);
 
CONNMGR.setMaxTotal(200);
 
CLIENT = HttpClientBuilder.create().setConnectionManager(CONNMGR).build();

其代码中是写死的

这两个参数的含义是什么呢？
下面用测试代码说明一下

测试端

 
public class HttpFluentUtil {
 
    private Logger logger = LoggerFactory.getLogger(HttpFluentUtil.class);
 
    private final static int MaxPerRoute = 2;
 
    private final static int MaxTotal = 4;
 
 
 
    final static PoolingHttpClientConnectionManager CONNMGR;
 
    final static HttpClient CLIENT;
 
    final static Executor EXECUTOR;
 
 
 
    static {
 
        LayeredConnectionSocketFactory ssl = null;
 
        try {
 
            ssl = SSLConnectionSocketFactory.getSystemSocketFactory();
 
        } catch (final SSLInitializationException ex) {
 
            final SSLContext sslcontext;
 
            try {
 
                sslcontext = SSLContext.getInstance(SSLConnectionSocketFactory.TLS);
 
                sslcontext.init(null, null, null);
 
                ssl = new SSLConnectionSocketFactory(sslcontext);
 
            } catch (final SecurityException ignore) {
 
            } catch (final KeyManagementException ignore) {
 
            } catch (final NoSuchAlgorithmException ignore) {
 
            }
 
        }
 
 
 
        final Registry<ConnectionSocketFactory> sfr = RegistryBuilder.<ConnectionSocketFactory>create()
 
                .register("http", PlainConnectionSocketFactory.getSocketFactory())
 
                .register("https", ssl != null ? ssl : SSLConnectionSocketFactory.getSocketFactory()).build();
 
 
 
        CONNMGR = new PoolingHttpClientConnectionManager(sfr);
 
        CONNMGR.setDefaultMaxPerRoute(MaxPerRoute);
 
        CONNMGR.setMaxTotal(MaxTotal);
 
        CLIENT = HttpClientBuilder.create().setConnectionManager(CONNMGR).build();
 
        EXECUTOR = Executor.newInstance(CLIENT);
 
    }
 
 
 
    public static String Get(String uri, int connectTimeout, int socketTimeout) throws IOException {
 
        return EXECUTOR.execute(Request.Get(uri).connectTimeout(connectTimeout).socketTimeout(socketTimeout))
 
                .returnContent().asString();
 
    }
 
 
 
    public static String Post(String uri, StringEntity stringEntity, int connectTimeout, int socketTimeout)
 
            throws IOException {
 
        return EXECUTOR.execute(Request.Post(uri).socketTimeout(socketTimeout)
 
                .addHeader("Content-Type", "application/json").body(stringEntity)).returnContent().asString();
 
    }
 
 
 
 
 
    public static void main(String[] args) {
 
        HttpUtil httpUtil = new HttpUtil();
 
        String url = "http://localhost:9064/app/test"; // 服务端sleep 5秒再返回
 
        for (int i = 0; i < 5; i++) { // MaxPerRoute若设置为2，则5线程分3组返回（2、2、1），共15秒
 
            new Thread(new Runnable() {
 
 
 
                @Override
 
                public void run() {
 
                    try {
 
                        String result = HttpFluentUtil.Get(url, 2000, 2000);
 
                        System.out.println(result);
 
                    } catch (IOException e) {
 
                        e.printStackTrace();
 
                    }
 
 
 
                }
 
            }).start();
 
        }
 
 
 
    }
 
 
 
}

服务器端

很简单的springmvc

 
@GetMapping(value="test")
 
public String test() throws InterruptedException {
 
    Thread.sleep(1000);
 
    return "1";
 
}

测试1：测试端MaxPerRoute=5 MaxTotal=4

服务器端结果

可以看到先接收4个请求，处理完成后，再接收下一次剩余的1个请求。即其一次最多接收MaxTotal次请求

测试2：测试端MaxPerRoute=2 MaxTotal=5

服务器端结果

可以看到接收2个请求，2个请求，1个请求，即说明maxPerRoute意思是某一个服务每次能并行接收的请求数量

什么场景下要设置？
知道了两个参数的含义，那么在什么情况下要对这两个参数进行设置呢？

比如说下面的场景

服务1要通过Fluent调用服务2的接口。服务1发送了400个请求，但由于Fluent默认只支持maxPerRoute=100，MaxTotal=200，比如接口执行时间为500ms，由于maxPerRoute=100，所以要分为100,100,100,100分四批来执行，全部执行完成需要2000ms。而如果maxPerRoute设置为400，全部执行完需要500ms。在这种情况下（提供并发能力时）就要对这两个参数进行设置了。

设置的方法
Apache Fluent可以使用上面测试的HttpFluentUtil工具类来执行请求

RestTemplate类似使用下面的方式

 
@Bean
 
public HttpClient httpClient() {
 
   Registry<ConnectionSocketFactory> registry = RegistryBuilder.<ConnectionSocketFactory>create()
 
            .register("http", PlainConnectionSocketFactory.getSocketFactory())
 
            .register("https", SSLConnectionSocketFactory.getSocketFactory())
 
            .build();
 
    PoolingHttpClientConnectionManager connectionManager = new PoolingHttpClientConnectionManager(registry);
 
    connectionManager.setMaxTotal(restTemplateProperties.getMaxTotal());
 
    connectionManager.setDefaultMaxPerRoute(restTemplateProperties.getDefaultMaxPerRoute());
 
    connectionManager.setValidateAfterInactivity(restTemplateProperties.getValidateAfterInactivity());
 
    RequestConfig requestConfig = RequestConfig.custom()
 
            .setSocketTimeout(restTemplateProperties.getSocketTimeout())
 
            .setConnectTimeout(restTemplateProperties.getConnectTimeout())
 
            .setConnectionRequestTimeout(restTemplateProperties.getConnectionRequestTimeout())
 
            .build();
 
    return HttpClientBuilder.create()
 
            .setDefaultRequestConfig(requestConfig)
 
            .setConnectionManager(connectionManager)
 
            .build();
 
   
 
}

@Bean
 
public ClientHttpRequestFactory httpRequestFactory() {
 
    return new HttpComponentsClientHttpRequestFactory(httpClient());
 
}

@Bean
 
   public RestTemplate restTemplate() {
 
       return new RestTemplate(httpRequestFactory());
 
   }

其中RestTemplateProperties通过配置文件来配置

max-total

default-max-per-route

connect-timeout  获取连接超时

connection-request-timeout  请求超时

socket-timeout  读超时