Play Framework/Netty不释放套接字

Question

已经问过类似的问题，但是删掉了，既然以为我修好了，但是我错了。

我在生产中为我的一个web项目使用Play framework。有时播放不呈现主页面或不返回一些静态内容文件。

最初应用程序运行良好，它必须工作5-7天，我可以看到这个问题。很难重现，它发生了15次中的1次，我不得不删除缓存数据和重新加载页面。（按下FF中的CRTL-F5)。问题可以在不同机器和操作系统的大多数浏览器中重现。最初，我认为主机提供商有一些问题。但我已经改变了它，问题没有消失。

我怀疑，Play Framework使用的netty存在一些问题。Netty 3.5.7最后的罐子是由游戏使用的。

cd /proc/28761/fd
ls -l | wc -l
337

在几天内，打开的文件描述符从140个增加到350个。注意，网站在开始和以后的平均负载是相同的。

我可以看到很多由进程打开的套接字，这些套接字后来没有释放。

lrwx------ 1 root root 64 Nov 11 10:34 300 -> socket:[1079566]
lrwx------ 1 root root 64 Nov 11 10:34 301 -> socket:[1079568]
lrwx------ 1 root root 64 Nov 11 10:34 302 -> socket:[1149958]
lrwx------ 1 root root 64 Nov 11 10:34 303 -> socket:[1160807]
lrwx------ 1 root root 64 Nov 11 10:34 304 -> socket:[1160605]
lrwx------ 1 root root 64 Nov 11 10:34 305 -> socket:[1157435]
lrwx------ 1 root root 64 Nov 11 10:34 306 -> socket:[1160607]
lrwx------ 1 root root 64 Nov 11 10:34 307 -> socket:[1160609]
lrwx------ 1 root root 64 Nov 11 10:34 308 -> socket:[1155542]
lrwx------ 1 root root 64 Nov 11 10:34 309 -> socket:[1120231]

更新

应用程序启动时（运行几小时）打开的TCP连接数为63。

Total: 150 (kernel 181)
TCP:   63 (estab 38, closed 5, orphaned 0, synrecv 0, timewait 3/0), ports 44

Transport Total     IP        IPv6
*         181       -         -
RAW       0         0         0
UDP       7         4         3
TCP       58        9         49
INET      65        13        52
FRAG      0         0         0

运行2天后，打开的TCP连接数为490。

[root@82711-2 fd]# ss -s
Total: 459 (kernel 490)
TCP:   378 (estab 271, closed 23, orphaned 0, synrecv 0, timewait 9/0), ports 37

Transport Total     IP        IPv6
*         490       -         -
RAW       0         0         0
UDP       7         4         3
TCP       355       12        343
INET      362       16        346
FRAG      0         0         0

Name: New I/O  worker #21
State: BLOCKED on org.jboss.netty.handler.stream.ChunkedWriteHandler@15e057 owned by: play-thread-2
Total blocked: 44  Total waited: 9

Stack trace: 
org.jboss.netty.handler.stream.ChunkedWriteHandler.flush(ChunkedWriteHandler.java:188)
org.jboss.netty.handler.stream.ChunkedWriteHandler.handleUpstream(ChunkedWriteHandler.java:140)
org.jboss.netty.channel.DefaultChannelPipeline.sendUpstream(DefaultChannelPipeline.java:564)
org.jboss.netty.channel.DefaultChannelPipeline$DefaultChannelHandlerContext.sendUpstream(DefaultChannelPipeline.java:792)
org.jboss.netty.channel.SimpleChannelUpstreamHandler.channelClosed(SimpleChannelUpstreamHandler.java:212)
org.jboss.netty.channel.SimpleChannelUpstreamHandler.handleUpstream(SimpleChannelUpstreamHandler.java:93)
org.jboss.netty.channel.DefaultChannelPipeline.sendUpstream(DefaultChannelPipeline.java:564)
org.jboss.netty.channel.DefaultChannelPipeline$DefaultChannelHandlerContext.sendUpstream(DefaultChannelPipeline.java:792)
org.jboss.netty.handler.codec.replay.ReplayingDecoder.cleanup(ReplayingDecoder.java:636)
org.jboss.netty.handler.codec.replay.ReplayingDecoder.channelClosed(ReplayingDecoder.java:533)
org.jboss.netty.channel.SimpleChannelUpstreamHandler.handleUpstream(SimpleChannelUpstreamHandler.java:93)
org.jboss.netty.channel.DefaultChannelPipeline.sendUpstream(DefaultChannelPipeline.java:564)
org.jboss.netty.channel.DefaultChannelPipeline.sendUpstream(DefaultChannelPipeline.java:559)
org.jboss.netty.channel.Channels.fireChannelClosed(Channels.java:476)
org.jboss.netty.channel.socket.nio.AbstractNioWorker.close(AbstractNioWorker.java:631)
org.jboss.netty.channel.socket.nio.NioServerSocketPipelineSink.handleAcceptedSocket(NioServerSocketPipelineSink.java:109)
org.jboss.netty.channel.socket.nio.NioServerSocketPipelineSink.eventSunk(NioServerSocketPipelineSink.java:66)
org.jboss.netty.channel.DefaultChannelPipeline$DefaultChannelHandlerContext.sendDownstream(DefaultChannelPipeline.java:780)
org.jboss.netty.handler.codec.oneone.OneToOneEncoder.handleDownstream(OneToOneEncoder.java:55)
org.jboss.netty.channel.DefaultChannelPipeline.sendDownstream(DefaultChannelPipeline.java:591)
org.jboss.netty.channel.DefaultChannelPipeline$DefaultChannelHandlerContext.sendDownstream(DefaultChannelPipeline.java:785)
org.jboss.netty.handler.stream.ChunkedWriteHandler.handleDownstream(ChunkedWriteHandler.java:111)
org.jboss.netty.channel.DefaultChannelPipeline.sendDownstream(DefaultChannelPipeline.java:591)
org.jboss.netty.channel.DefaultChannelPipeline.sendDownstream(DefaultChannelPipeline.java:582)
org.jboss.netty.channel.Channels.close(Channels.java:821)
org.jboss.netty.channel.AbstractChannel.close(AbstractChannel.java:194)
org.jboss.netty.channel.ChannelFutureListener$1.operationComplete(ChannelFutureListener.java:41)
org.jboss.netty.channel.DefaultChannelFuture.notifyListener(DefaultChannelFuture.java:399)
org.jboss.netty.channel.DefaultChannelFuture.notifyListeners(DefaultChannelFuture.java:385)
org.jboss.netty.channel.DefaultChannelFuture.setSuccess(DefaultChannelFuture.java:334)
org.jboss.netty.channel.socket.nio.AbstractNioWorker.write0(AbstractNioWorker.java:493)
   - locked java.lang.Object@3b7e28
org.jboss.netty.channel.socket.nio.AbstractNioWorker.writeFromTaskLoop(AbstractNioWorker.java:431)
org.jboss.netty.channel.socket.nio.AbstractNioChannel$WriteTask.run(AbstractNioChannel.java:364)
org.jboss.netty.channel.socket.nio.AbstractNioWorker.processWriteTaskQueue(AbstractNioWorker.java:349)
org.jboss.netty.channel.socket.nio.AbstractNioWorker.run(AbstractNioWorker.java:245)
org.jboss.netty.channel.socket.nio.NioWorker.run(NioWorker.java:38)
org.jboss.netty.util.ThreadRenamingRunnable.run(ThreadRenamingRunnable.java:102)
org.jboss.netty.util.internal.DeadLockProofWorker$1.run(DeadLockProofWorker.java:42)
java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:886)
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:908)
java.lang.Thread.run(Thread.java:662)

Name: play-thread-2
State: BLOCKED on java.lang.Object@3b7e28 owned by: New I/O  worker #21
Total blocked: 23  Total waited: 34 778

Stack trace: 
org.jboss.netty.channel.socket.nio.AbstractNioWorker.cleanUpWriteBuffer(AbstractNioWorker.java:654)
org.jboss.netty.channel.socket.nio.AbstractNioWorker.writeFromUserCode(AbstractNioWorker.java:408)
org.jboss.netty.channel.socket.nio.NioServerSocketPipelineSink.handleAcceptedSocket(NioServerSocketPipelineSink.java:127)
org.jboss.netty.channel.socket.nio.NioServerSocketPipelineSink.eventSunk(NioServerSocketPipelineSink.java:66)
org.jboss.netty.channel.DefaultChannelPipeline$DefaultChannelHandlerContext.sendDownstream(DefaultChannelPipeline.java:780)
org.jboss.netty.handler.codec.oneone.OneToOneEncoder.handleDownstream(OneToOneEncoder.java:63)
org.jboss.netty.channel.DefaultChannelPipeline.sendDownstream(DefaultChannelPipeline.java:591)
org.jboss.netty.channel.DefaultChannelPipeline$DefaultChannelHandlerContext.sendDownstream(DefaultChannelPipeline.java:785)
org.jboss.netty.channel.Channels.write(Channels.java:733)
org.jboss.netty.handler.stream.ChunkedWriteHandler.flush(ChunkedWriteHandler.java:262)
   - locked org.jboss.netty.handler.stream.ChunkedWriteHandler@15e057
org.jboss.netty.handler.stream.ChunkedWriteHandler.handleDownstream(ChunkedWriteHandler.java:121)
org.jboss.netty.channel.DefaultChannelPipeline.sendDownstream(DefaultChannelPipeline.java:591)
org.jboss.netty.channel.DefaultChannelPipeline.sendDownstream(DefaultChannelPipeline.java:582)
org.jboss.netty.channel.Channels.write(Channels.java:712)
org.jboss.netty.channel.Channels.write(Channels.java:679)
org.jboss.netty.channel.AbstractChannel.write(AbstractChannel.java:245)
play.server.PlayHandler.serveStatic(PlayHandler.java:886)
play.server.PlayHandler$NettyInvocation.init(PlayHandler.java:182)
play.Invoker$Invocation.run(Invoker.java:276)
play.server.PlayHandler$NettyInvocation.run(PlayHandler.java:229)
java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:441)
java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:303)
java.util.concurrent.FutureTask.run(FutureTask.java:138)
java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$301(ScheduledThreadPoolExecutor.java:98)
java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:206)
java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:886)
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:908)
java.lang.Thread.run(Thread.java:662)

我在Tomcat7的相同环境中部署了相同的Play应用程序。24小时过去了，问题已经解决，打开的TCP连接数保持不变。打开的文件描述符数不超过70。这是相同的生产主机、相同的数据库和相同的应用程序用户。

Answer 1

实际上，我不是在play中遇到了类似的bug，而是在JVM（play运行的JVM）中遇到了这样一个bug，即指向filehandles的封闭通道直到关闭JVM才被强制释放。唉，我不记得我是如何找到bug报告的，或者我会链接到它，但它是JVM中已知的bug。我最终不得不努力解决它。我能建议的最好的办法是重写代码，尽可能使用相同的通道/文件句柄。