如果网络被修改，Java 应用程序将挂起
我正在开发一个有多个视图的 Eclipse 应用程序。从这些视图中，我必须通过网络访问一些数据。在以下情况下会引发错误。
第 1 步：启动应用程序，一切正常。
第 2 步：禁用网络
第 3 步：启用网络
第 4 步：尝试从其中一个视图通过线路访问数据。该应用程序显然挂起。
第 5 步：现在，如果您尝试从其他视图访问数据，一切都会正常。
最初我认为这是由于 JVM 的 DNS 缓存而发生的。所以我做了这样的事情。
System.setProperty( "networkaddress.cache.ttl", "0" );
System.setProperty( "networkaddress.cache.negative.ttl" , "0" );
但问题仍然存在。任何帮助表示赞赏。
I am working on an eclipse app that has got multiple views. From these views I have to access some data over the wire. An error is thrown in following scenario.
Step 1: Start app, everything works fine.
Step 2: Disable network
Step 3: Enable Network
Step 4: Try to access data over wire from one of the views. The app apparently hangs.
Step 5: Now if you try to access data from some other view, everything works.
Initially I thought that it is happening due to DNS caching by JVM. So I did something like this.
System.setProperty( "networkaddress.cache.ttl", "0" );
System.setProperty( "networkaddress.cache.negative.ttl" , "0" );
But the problem is still there. Any help is appreciated.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（3）
可能在步骤2和步骤3之间。抛出什么类型的异常？
it's probably between step 2 and 3. What type of exception is thrown?
当应用程序挂起时查看当前堆栈的另一种选择是使用 jps/jstack（与从 1.5 开始的 Sun JDK 捆绑在一起）或 Visualvm（在 1.6 SDK 中提供，并与 https://visualvm.dev.java.net/)
jps 列出计算机上当前的所有 java 进程。然后，您可以使用每个进程 ID 调用 jstack 来获取当前堆栈。
也许如果您可以发布这样的堆栈跟踪，我们可以提供帮助。
Another option to see the current stack when your app is hanging would be to use jps/jstack (bundled with Sun JDKs beginning with 1.5) or visualvm (available in 1.6 SDKs and separately from https://visualvm.dev.java.net/)
jps lists all current java processes on a machine. you can then call jstack with each of these process ids to get the current stack.
Maybe if you could post such a stack trace we could help.
完全疯狂的猜测：如果您使用 TCP，您的应用程序可能不会注意到其套接字已断开连接。如果它在读取中被阻止，它可能不会注意到，直到它尝试向套接字写入内容。这取决于套接字关闭的“干净”程度——如果套接字没有从另一端收到 RST 数据包，它就不会知道另一端已经消失。
您可以尝试从应用程序中强制执行线程转储。通常，JVM 将在终端上运行时响应 Ctrl-\ 或响应带有每个线程的堆栈跟踪的
kill -QUIT信号。这可以让您看到有问题的线程被阻塞的位置。如果您可以发布一些可疑代码，那么尝试做出诊断就会容易得多。或者您能解释一下为什么您认为 DNS 缓存会成为一个问题吗？
Completely Wild Guess: if you're using TCP your app may not be noticing that its socket has been disconnected. If it's blocked in a read it may not notice until it tries to write something to the socket. It depends on how "clean" the socket shutdown is--if the socket doesn't get a RST packet from the other side it won't know that the other end has gone away.
You could try coercing a thread dump out of your app. Typically the JVM will respond to either Ctrl-\ when running at a terminal or to a
kill -QUITsignal with a stack trace of every thread. That'd let you see where the thread in question is blocked.If you can post some suspect code it'd be much easier to try to make a diagnosis. Or can you explain why you think DNS caching would be an issue?