This is an issue with I am trying to fix, maybe a bug in JRE please vote to reopen this question. It's very selfish approach to mark something off topic if you are not competent enough to understand some problem.
Java版本
java版本“1.7.0_45” Java(TM) SE 运行时环境(版本 1.7.0_45-b18) Java HotSpot(TM) 64 位服务器 VM(内部版本 24.45-b08，混合模式)
遵循两个版本的代码都会产生相同的崩溃。
客户端计算机使用在 Windows Server 2003 64 位上运行的 Server Maching(使用 Oracle 10g 数据库和 Hibernate 3.0)调用部署在 Tomcat 7 中的 Servlet。
当我尝试将负载放在 Servlet 上进行负载测试时，我编写了以下两种方法，它们仅创建一个线程并调用同步方法，但没有任何日志“Eclipse 中的进程终止”的堆栈跟踪。
我不知道崩溃的原因，如果我创建 500 个线程，那么它会在 487 处崩溃，如果我创建 1000 个线程，那么它会在 874 处崩溃。我不知道真正的原因是什么。我配置了StuckThreadDetectionValve检测长时间运行的线程。
https://stackoverflow.com/a/7484800/185022
我将在这里分享所有更新。如果您知道可能出了什么问题，请至少给我一个提示。
** 两个代码都会产生相同的崩溃 **
Code 1:
public class TokenTest {
static int count = 0;
public static void main(String[] args) {
try {
System.out.println("Start " + new Date());
final ApplicationServiceClient app = WSClient.getClient("http://192.168.1.15:81/az-server", ApplicationServiceClient.class);
final int max = 50;
ExecutorService executor = Executors.newFixedThreadPool(max);
for (int i = 1; i <= max; i++) {
executor.execute(new Runnable() {
public void run() {
try {
System.out.print(++count +",");
app.verifyToken("7056451004350030976"); //Network Synchronous Operation, calls a Servlet that is deployed in another Server Machine, Using Tomcat 7.0
} catch (Exception ex) {
ex.printStackTrace();
}
}
});
}
executor.shutdown();
while (!executor.isShutdown()) {
System.out.println("END " + new Date());
}
} catch (Exception ex) {
ex.printStackTrace();
}
}
}
Code 2:
public class TokenTest {
static int count = 0;
public static void main(String[] args) {
try {
System.out.println(new Date());
final ApplicationServiceClient app = WSClient.getClient("http://192.168.1.15:81/az-server", ApplicationServiceClient.class);
final int max=250;
for (int i = 1; i <= max; i++) {
Thread t = new Thread() {
public void run() {
app.verifyToken("7056451004350030976");//Network Synchronous Operation
System.out.print(count++);
System.out.print(", ");
if (count >= max)
System.out.println("\n" + new Date());
}
};
t.start();
}
} catch (Exception ex) {
ex.printStackTrace();
}
}
}
注意: 该服务器托管在另一台 Windows Server 2003 64 位计算机上。
已更新 它只是终止我在 Eclipse 中的 Java 进程，如果我给出像 5000 这样的值，那么它会停留在 2800 上，什么也不会发生，也不会终止我的进程。这是控制台快照。
更新*
感谢 S.Yavari。他引领了正确的方向。配置后Stuck_Thread_Detection_Valve在 Tomcat 上并中断卡住线程的线程，我能够毫无问题地在服务器上生成 500K 客户端的负载，并且在开始新请求之前还提供 100 毫秒的 sleep 。
下面是代码:
public class StageOne {
final ApplicationServiceClient appClient = WSClient.getClient(Main.url, ApplicationServiceClient.class);
final SecureWebServiceClient secureClient = WSClient.getClient(Main.url, SecureWebServiceClient.class);
final static List<Long> times = new Vector<Long>();
String token = "";
public String authenticateClient() {
token = appClient.authenticateClient("az_y", "az", "0");
// System.out.println("Registration Response: " + token);
return token;
}
private void getDefaultDepartmentName() {
String deptname = appClient.getDefaultDepartmentName();
// System.out.println("Department: " + deptname);
}
private void getLocation() {
List<String> location = appClient.listLocation();
// System.out.println("Location: " + location);
}
private void registerClient_() {
secureClient.setToken(token);
String respon = secureClient.registerClient("Buddy AZ", "443", "njuy", "1");
// System.out.println(respon);
}
static int count = 0;
public static void main(String[] args) {
int pool = 500000;
final ExecutorService execotors = Executors.newFixedThreadPool(pool);
for (int i = 0; i < pool; i++) {
try {
Thread.sleep(100);
} catch (InterruptedException e) {
e.printStackTrace();
}
execotors.execute(new Runnable() {
@Override
public void run() {
long startTime = System.nanoTime();
StageOne one = new StageOne();
one.authenticateClient();
one.getDefaultDepartmentName();
one.getLocation();
one.registerClient_();
long EndTime = System.nanoTime();
// System.out.println("Time in Seconds ("+ (++count) +")" + TimeUnit.NANOSECONDS.toSeconds(EndTime - startTime));
times.add(TimeUnit.NANOSECONDS.toSeconds(EndTime - startTime));
}
});
}
execotors.shutdown();
while (!execotors.isShutdown()) {
try {
Thread.sleep(10);
} catch (Exception e) {
// TODO: handle exception
}
}
Collections.sort(times);
long sum = 0;
for (int i = 0; i < times.size(); i++) {
sum += times.get(i);
}
long result = sum / Math.max(1, times.size());
System.out.println("Min Response Time: " + times.get(0));
System.out.println("Average Response Time: " + result);
System.out.println("Max Response Time: " + times.get(times.size() - 1));
}
}
最佳答案
我认为这只是一个线程问题。使用此代码:
public class Test2 {
public static final Object LOCK = new Object();
static int count = 0;
public static void main(String[] args) {
try {
System.out.println(new Date());
final ApplicationServiceClient app = WSClient.getClient("http://192.168.1.15:81/az-server", ApplicationServiceClient.class);
final int max=250;
for (int i = 1; i <= max; i++) {
Thread t = new Thread() {
public void run() {
synchronized (LOCK) {
app.verifyToken("7056451004350030976");
}
System.out.print(count++);
System.out.print(", ");
if (count >= max)
System.out.println("\n" + new Date());
}
};
t.start();
}
} catch (Exception ex) {
ex.printStackTrace();
}
}
}
这个方法(verifyToken)必须是线程安全的，否则在线程内使用这个方法会遇到很多问题。
关于java - JVM 在创建线程并在其中执行网络操作时突然关闭，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/20728451/