在多线程环境下使用HttpClient组件对某个HTTP服务发起请求,运行一段时间以后发现客户端主机CPU利用率呈现出降低趋势,而不是一个稳定的状态。
并且,从程序日志中判断有线程处于hang住的状态,应该是被阻塞了。
html
一开始找不到缘由,怀疑是多线程并发致使的死锁问题,可是经过代码审查并未定位到任何可能的多线程并发问题。
甚至开始怀疑是不是由于内存资源不够引发JVM频繁GC到致使业务线程被暂停,可是从GC的日志输出结果看,GC是正常的。
因而,进入一种丈二和尚摸不着头脑头脑的状态,再次Review代码,发现并未设置请求超时时间,因而设置超时控制,发现问题依然存在,完全懵逼了。
最后,dump线程堆栈和内存堆栈,再对堆栈数据进行分析。从分析结果看,确认是由于Socket链接在读取数据时被阻塞引发线程hang住。搜索“httpclient 超时”关键字,找到各式各样设置HttpClient超时控制的方式,均尝试过可是并未生效。
实际上到后来才知道,HttpCient的超时控制在不一样的版本中设置请求超时参数的方式均各不相同,这才致使了我使用了网上看到的方式设置以后并未生效。固然,根本缘由仍是由于对HttpClient这个组件不熟悉致使的.java
1.HttpClient版本git
<dependency> <groupId>org.apache.httpcomponents</groupId> <artifactId>httpclient</artifactId> <version>4.5.2</version> </dependency>
2.Java代码github
public class HttpClientTest { private AtomicInteger counter = new AtomicInteger(0); private String url = "http://www.baidu.com/"; public static void main(String[] args) { new HttpClientTest().test(); } // 执行测试 private void test() { int number = 100000; // 总请求数 int concurrent = 50; // 每次并发请求数 CountDownLatch countDownLatch = new CountDownLatch(number); // 计数器 ExecutorService threadPool = Executors.newFixedThreadPool(concurrent); // 线程池 int concurrentPer = concurrent; boolean over = false; while(!over) { number = number - concurrent; if(number <= 0) { concurrentPer = number + concurrent; over = true; } // 线程池批量提交 for(int i = 0; i < concurrentPer; i++) { threadPool.execute(new Runnable() { @Override public void run() { try { request(url); Thread.sleep(100); } catch (IOException | InterruptedException e) { e.printStackTrace(); } finally { countDownLatch.countDown(); } } }); } } try { countDownLatch.await(); threadPool.shutdown(); } catch (InterruptedException e) { e.printStackTrace(); } } // 访问指定地址 private void request(String url) throws IOException { HttpGet httpGet = new HttpGet(url); commnicate(httpGet); } // 负责底层通讯处理 private void commnicate(HttpRequestBase request) throws IOException { ResponseHandler<String> responseHandler = new ResponseHandler<String>() { @Override public String handleResponse(HttpResponse response) throws ClientProtocolException, IOException { return EntityUtils.toString(response.getEntity()); } }; HttpClient client = HttpClients.createDefault(); String body = client.execute(request, responseHandler); // 线程可能会在这里被阻塞 System.out.println(String.format("body size: %s, counter: %s", body.length(), counter.incrementAndGet())); } }
运行上述代码一段时间后很容易能够重现出问题,以下为运行控制台信息:
shell
而且线程所有hang住,进程没法正常结束.apache
查看端口状态存在大量请求处于创建链接状态(ESTABLISHED):性能优化
# netstat -anpt tcp 0 0 172.17.7.81:56408 180.118.128.179:56033 ESTABLISHED 3766/java tcp 0 0 172.17.7.81:57644 115.202.238.177:27016 ESTABLISHED 3847/java tcp 0 0 172.17.7.81:36616 117.57.21.0:25719 ESTABLISHED 3766/java tcp 0 0 172.17.7.81:59944 112.245.197.118:57220 ESTABLISHED 3847/java tcp 0 0 172.17.7.81:48722 218.5.215.10:40835 ESTABLISHED 4007/java tcp 0 0 172.17.7.81:52734 115.194.17.14:45210 ESTABLISHED 4007/java tcp 0 0 172.17.7.81:60586 59.32.37.129:16637 ESTABLISHED 3686/java tcp 0 0 172.17.7.81:36776 222.89.86.109:21667 ESTABLISHED 3766/java tcp 0 0 172.17.7.81:51690 60.161.249.162:59039 ESTABLISHED 3927/java tcp 0 0 172.17.7.81:42226 58.218.200.59:80 TIME_WAIT - tcp 0 0 172.17.7.81:56566 117.70.47.194:40879 ESTABLISHED 3686/java tcp 0 0 172.17.7.81:43266 182.120.202.204:45893 ESTABLISHED 3766/java tcp 0 0 172.17.7.81:55630 60.169.223.16:21280 ESTABLISHED 3927/java tcp 0 0 172.17.7.81:54922 60.168.81.26:25464 ESTABLISHED 3927/java tcp 0 0 172.17.7.81:53352 112.252.97.83:53584 ESTABLISHED 3847/java tcp 0 0 172.17.7.81:52684 113.121.242.43:14447 ESTABLISHED 3927/java tcp 0 0 172.17.7.81:54750 113.121.241.168:45173 ESTABLISHED 3686/java tcp 0 0 172.17.7.81:41510 113.105.202.106:47288 ESTABLISHED 4007/java tcp 0 0 172.17.7.81:38804 121.232.148.62:57938 ESTABLISHED 3847/java tcp 0 0 172.17.7.81:41468 113.105.202.106:47288 ESTABLISHED 3927/java tcp 0 0 172.17.7.81:45444 123.163.81.185:22012 ESTABLISHED 3766/java tcp 0 0 172.17.7.81:54810 113.121.241.168:45173 ESTABLISHED 4007/java tcp 0 0 172.17.7.81:51542 175.153.23.147:20766 ESTABLISHED 3927/java tcp 0 0 172.17.7.81:45644 218.5.215.10:40835 ESTABLISHED 4007/java tcp 0 0 172.17.7.81:35730 116.53.197.198:30042 ESTABLISHED 3766/java tcp 0 0 172.17.7.81:54738 113.121.241.168:45173 ESTABLISHED 3686/java tcp 0 0 172.17.7.81:60600 59.32.37.129:16637 ESTABLISHED 3686/java tcp 0 0 172.17.7.81:54862 113.121.241.168:45173 ESTABLISHED 4007/java tcp 0 0 172.17.7.81:40980 115.225.153.215:17292 ESTABLISHED 3686/java tcp 0 0 172.17.7.81:54166 123.149.162.129:18269 ESTABLISHED 3766/java tcp 0 0 172.17.7.81:60712 120.35.190.184:33054 ESTABLISHED 3766/java tcp 0 0 172.17.7.81:55802 106.42.211.65:59547 ESTABLISHED 3766/java
同时,分析线程堆栈信息(jstack -F -l pid > thread_stack.log
)能够看到以下信息:多线程
"pool-1-thread-45" #55 prio=5 os_prio=0 tid=0x00007f78702df000 nid=0x33d5 runnable [0x00007f7830c1d000] java.lang.Thread.State: RUNNABLE at java.net.SocketInputStream.socketRead0(Native Method) at java.net.SocketInputStream.socketRead(SocketInputStream.java:116) at java.net.SocketInputStream.read(SocketInputStream.java:171) at java.net.SocketInputStream.read(SocketInputStream.java:141) at org.apache.http.impl.io.SessionInputBufferImpl.streamRead(SessionInputBufferImpl.java:139) at org.apache.http.impl.io.SessionInputBufferImpl.fillBuffer(SessionInputBufferImpl.java:155) at org.apache.http.impl.io.SessionInputBufferImpl.readLine(SessionInputBufferImpl.java:284) at org.apache.http.impl.conn.DefaultHttpResponseParser.parseHead(DefaultHttpResponseParser.java:140) at org.apache.http.impl.conn.DefaultHttpResponseParser.parseHead(DefaultHttpResponseParser.java:57) at org.apache.http.impl.io.AbstractMessageParser.parse(AbstractMessageParser.java:261) at org.apache.http.impl.DefaultBHttpClientConnection.receiveResponseHeader(DefaultBHttpClientConnection.java:165) at org.apache.http.impl.conn.CPoolProxy.receiveResponseHeader(CPoolProxy.java:167) at org.apache.http.protocol.HttpRequestExecutor.doReceiveResponse(HttpRequestExecutor.java:272) at org.apache.http.protocol.HttpRequestExecutor.execute(HttpRequestExecutor.java:124) at org.apache.http.impl.execchain.MainClientExec.execute(MainClientExec.java:271) at org.apache.http.impl.execchain.ProtocolExec.execute(ProtocolExec.java:184) at org.apache.http.impl.execchain.RetryExec.execute(RetryExec.java:88) at org.apache.http.impl.execchain.RedirectExec.execute(RedirectExec.java:110) at org.apache.http.impl.client.InternalHttpClient.doExecute(InternalHttpClient.java:184) at org.apache.http.impl.client.CloseableHttpClient.execute(CloseableHttpClient.java:71) at org.apache.http.impl.client.CloseableHttpClient.execute(CloseableHttpClient.java:220) at org.apache.http.impl.client.CloseableHttpClient.execute(CloseableHttpClient.java:164) at org.apache.http.impl.client.CloseableHttpClient.execute(CloseableHttpClient.java:139) at org.chench.extra.HttpClientTest.commnicate(HttpClientTest.java:106) # 线程在这里阻塞 at org.chench.extra.HttpClientTest.request(HttpClientTest.java:93) at org.chench.extra.HttpClientTest.access$100(HttpClientTest.java:31) at org.chench.extra.HttpClientTest$1.run(HttpClientTest.java:62) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) at java.lang.Thread.run(Thread.java:748) Locked ownable synchronizers: - <0x0000000086d50638> (a java.util.concurrent.ThreadPoolExecutor$Worker)
从堆栈日志中能够看到,线程处于RUNNABLE
状态,而且阻塞在at org.chench.extra.HttpClientTest.commnicate(HttpClientTest.java:106)
处.并发
经过线程堆栈日志分析能够定位到线程hang住是由于HttpClient在执行访问时被阻塞了,结合源代码找到阻塞缘由是由于未设置请求超时时间.app
上述问题本质上是由于HttpClient组件并未设置请求超时控制致使的:虽然链接超时,可是读取失败,致使线程一直被阻塞.
那么,应该如何设置HttpClient的超时时间呢?鉴于HttpClient的官方文档没有明确说明,而且不一样版本的HttpClient组件设置超时控制的方式不一致,因此建议直接查看源码.
HttpClient执行访问请求时序图以下:
顺藤摸瓜,在MainClientExec.java
的execute()
方法中看到有2处使用了timeout参数,其含义各不相同:
(1)在获取HttpClientConnection对象时须要读取配置参数中的ConnectionRequestTimeout
值,该参数用于控制获取链接的超时时间.
(2)获取到HttpClientConnection对象以后读取配置参数中的SocketTimeout
值,设置Socket超时时间.
显然,这2个timeout参数都须要从RequestConfig对象中获取.
既然找到了使用timeout参数的地方,下一步须要肯定该参数是如何设置的.沿着HttpClient的请求时序图路径往回查找,在InternalHttpClient.java
类的doExecute()
方法中能够很清晰地看到设置了RequestConfig对象参数.
@Override protected CloseableHttpResponse doExecute( final HttpHost target, final HttpRequest request, final HttpContext context) throws IOException, ClientProtocolException { Args.notNull(request, "HTTP request"); HttpExecutionAware execAware = null; if (request instanceof HttpExecutionAware) { execAware = (HttpExecutionAware) request; } try { final HttpRequestWrapper wrapper = HttpRequestWrapper.wrap(request, target); final HttpClientContext localcontext = HttpClientContext.adapt( context != null ? context : new BasicHttpContext()); // 经过RequestConfig对象配置链接参数 RequestConfig config = null; if (request instanceof Configurable) { // 若是在HttpRequest对象中设置了RequestConfig属性,直接使用 config = ((Configurable) request).getConfig(); } if (config == null) { // 若是在HttpRequest对象中未设置RequestConfig对象属性,则获取HttpParams属性构造RequestConfig对象 final HttpParams params = request.getParams(); if (params instanceof HttpParamsNames) { if (!((HttpParamsNames) params).getNames().isEmpty()) { config = HttpClientParamConfig.getRequestConfig(params); } } else { config = HttpClientParamConfig.getRequestConfig(params); } } if (config != null) { // 使用RequestConfig对象配置链接参数 localcontext.setRequestConfig(config); } setupContext(localcontext); final HttpRoute route = determineRoute(target, wrapper, localcontext); return this.execChain.execute(route, wrapper, localcontext, execAware); } catch (final HttpException httpException) { throw new ClientProtocolException(httpException); } }
(3)HttpClient默认使用的链接池为PoolingHttpClientConnectionManager
,在创建链接时(connect()方法
)会使用其中的SocketConfig
配置参数对Socket进行配置,以下所示:
@Override public void connect( final HttpClientConnection managedConn, final HttpRoute route, final int connectTimeout, final HttpContext context) throws IOException { Args.notNull(managedConn, "Managed Connection"); Args.notNull(route, "HTTP route"); final ManagedHttpClientConnection conn; synchronized (managedConn) { final CPoolEntry entry = CPoolProxy.getPoolEntry(managedConn); conn = entry.getConnection(); } final HttpHost host; if (route.getProxyHost() != null) { host = route.getProxyHost(); } else { host = route.getTargetHost(); } final InetSocketAddress localAddress = route.getLocalSocketAddress(); SocketConfig socketConfig = this.configData.getSocketConfig(host); if (socketConfig == null) { // 使用配置参数SocketConfig socketConfig = this.configData.getDefaultSocketConfig(); } if (socketConfig == null) { socketConfig = SocketConfig.DEFAULT; } this.connectionOperator.connect( conn, host, localAddress, connectTimeout, socketConfig, context); }
@Override public void connect( final ManagedHttpClientConnection conn, final HttpHost host, final InetSocketAddress localAddress, final int connectTimeout, final SocketConfig socketConfig, final HttpContext context) throws IOException { final Lookup<ConnectionSocketFactory> registry = getSocketFactoryRegistry(context); final ConnectionSocketFactory sf = registry.lookup(host.getSchemeName()); if (sf == null) { throw new UnsupportedSchemeException(host.getSchemeName() + " protocol is not supported"); } final InetAddress[] addresses = host.getAddress() != null ? new InetAddress[] { host.getAddress() } : this.dnsResolver.resolve(host.getHostName()); final int port = this.schemePortResolver.resolve(host); for (int i = 0; i < addresses.length; i++) { final InetAddress address = addresses[i]; final boolean last = i == addresses.length - 1; Socket sock = sf.createSocket(context); // 使用socketConfig参数中的超时时间对Socket进行配置 sock.setSoTimeout(socketConfig.getSoTimeout()); sock.setReuseAddress(socketConfig.isSoReuseAddress()); sock.setTcpNoDelay(socketConfig.isTcpNoDelay()); sock.setKeepAlive(socketConfig.isSoKeepAlive()); if (socketConfig.getRcvBufSize() > 0) { sock.setReceiveBufferSize(socketConfig.getRcvBufSize()); } if (socketConfig.getSndBufSize() > 0) { sock.setSendBufferSize(socketConfig.getSndBufSize()); } final int linger = socketConfig.getSoLinger(); if (linger >= 0) { sock.setSoLinger(true, linger); } conn.bind(sock); // ... } }
通过源码解读能够很明确地知道,在HttpClient 4.5.2版本中,设置链接参数有3种方式:
(1)在HttpRequest对象中设置RequestConfig
对象属性
(2)在HttpRequest对象中设置HttpParams
对象属性.
(3)在链接池对象中设置SocketConfig
对象属性
既然找到了根源,下面分别经过这3种方式设置超时参数进行验证.
方式1: 经过RequestConfig对象设置超时参数
int timeOut = 5000; RequestConfig requestConfig = RequestConfig.custom() .setConnectionRequestTimeout(timeOut) // 获取链接超时时间 .setConnectTimeout(timeOut) // 设置HTTP链接超时时间 .setSocketTimeout(timeOut) // 设置Socket超时时间 .build(); request.setConfig(requestConfig);
方式2: 经过HttpParams对象设置超时参数
int timeOut = 5000; HttpParams params = new BasicHttpParams(); params.setParameter(CoreConnectionPNames.SO_TIMEOUT, timeOut); // 设置Socket超时时间 params.setParameter(CoreConnectionPNames.CONNECTION_TIMEOUT, timeOut); // 设置HTTP链接超时时间 request.setParams(params);
方式3: 经过链接池对象设置超时参数
int timeOut = 5000; PoolingHttpClientConnectionManager connManager = new PoolingHttpClientConnectionManager(); // 对链接池设置SocketConfig对象 connManager.setDefaultSocketConfig(SocketConfig.custom().setSoTimeout(timeOut).build()); client = HttpClients.custom().setConnectionManager(connManager).build();
经过上述3种方式分别设置超时参数验证,虽然在运行过程当中会有报错,可是不会致使线程被阻塞,进程能正常运行结束:
解决问题以后客户端CPU使用率恢复正常:
在Java平台使用Http客户端组件,能够有多个选择:
(1)直接使用JDK提供的URL类访问
(2)使用HttpClient组件,有坑,不一样版本设置参数的方式变更较大,最好是阅读一下当前使用版本的源码实现,正确设置好超时时间等参数
(3)若是使用了Spring MVC框架,还可使用Spring MVC提供的RestTemplate组件,底层是使用Netty实现的客户端
遇到的这个坑自己并不属于技术难点,可是面对这个问题的解决思路值得总结:
(1)程序日志,运行日志很是关键,是定位问题时第一时间须要查看的
(2)代码review,逐行逐行地审查,首先排除可能存在的代码逻辑问题,好比:死锁等
(3)经过jstack命令查看线程堆栈信息: jstack -l -F <pid> > stack.log
(4)经过jmap命令查看内存堆栈信息: jmap -dump:live format=b,file=heap.bin <pid>
(5)若是结合搜索引擎和上述排查步骤依然未能解决问题,应该第一时间考虑直接阅读组件的源代码实现,特别是使用了开源组件时这可能才是真正解决问题的最佳路径
【参考】 https://blog.csdn.net/u011191463/article/details/78664896 HttpClient超时设置详解 https://my.oschina.net/jywm/blog/1834702 解决httpclient超时设置不生效的问题 https://www.jianshu.com/p/4b3e172c4f2d HttpClient 4.5.2-(四)链接超时的配置 https://www.jianshu.com/p/6a41c95855e3 HttpClient 4.5.2-(五)链接池的配置 https://www.jianshu.com/p/c852cbcf3d68 HttpClient高并发下性能优化-http链接池 https://blog.csdn.net/u011402596/article/details/44619443 HttpClient 多线程处理 https://field-notes.iteye.com/blog/2383759 多线程消费使用HttpClient引起的坑 http://blog.51cto.com/lihao001/1788490 httpclient4.3 致使线程阻塞 https://study121007.iteye.com/blog/2304274 HttpClient4.5.2 链接管理 https://www.jianshu.com/p/c852cbcf3d68 HttpClient高并发下性能优化-http链接池 https://alafqq.iteye.com/blog/2325041 httpclient 多线程执行(网上版本太多了。。。误人子弟) https://gaozzsoft.iteye.com/blog/2352241 HttpClient 4.5.2版本设置链接超时时间-CloseableHttpClient设置Timeout https://www.cnblogs.com/softidea/p/6964347.html HttpClient 专题 https://blog.csdn.net/Fhaohaizi/article/details/78217903 httpclient4.5如何确保资源释放 https://blog.csdn.net/u010634066/article/details/83120122 一场HttpClient调用未关闭流引起的问题 https://www.cnblogs.com/mumuxinfei/p/5066633.html Apache HttpClient使用之阻塞陷阱 https://issues.apache.org/jira/browse/HTTPCLIENT-1584 CloseableHttpClient - SSL Handshake has no Socket Timeout https://monkeyissexy.github.io/2016/11/11/httpclient_ssl_handshake_socketTimeout_bug/ httpclient ssl handshake socketTimeout bug 分析解决过程 http://geekerwang.com/2017/10/22/%E8%AE%B0%E5%BD%95%E4%B8%80%E6%AC%A1%E5%9D%91%EF%BC%9AsocketRead-hang/ 记录一次坑:socketRead hang https://www.cnblogs.com/jessezeng/p/7448636.html 解决: httpclient ssl 验证致使死锁问题 http://itbang.me/solu/detail/201 JAVA线程卡死问题如何定位?