Netty服务器不关闭/释放套接字

Question

我在netty server应用程序中面临资源问题。

[io.netty.channel.DefaultChannelPipeline] An exceptionCaught() event was fired, and it reached at the tail of the pipeline. It usually means the last handler in the pipeline did not handle the exception.: java.io.IOException: Too many open files
    at sun.nio.ch.ServerSocketChannelImpl.accept0(Native Method) [rt.jar:1.7.0_60]
    at sun.nio.ch.ServerSocketChannelImpl.accept(ServerSocketChannelImpl.java:241) [rt.jar:1.7.0_60]
    at io.netty.channel.socket.nio.NioServerSocketChannel.doReadMessages(NioServerSocketChannel.java:135) [netty-all-4.0.25.Final.jar:4.0.25.Final]
    at io.netty.channel.nio.AbstractNioMessageChannel$NioMessageUnsafe.read(AbstractNioMessageChannel.java:69) [netty-all-4.0.25.Final.jar:4.0.25.Final]
    at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:511) [netty-all-4.0.25.Final.jar:4.0.25.Final]
    at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:468) [netty-all-4.0.25.Final.jar:4.0.25.Final]
    at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:382) [netty-all-4.0.25.Final.jar:4.0.25.Final]
    at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:354) [netty-all-4.0.25.Final.jar:4.0.25.Final]
    at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:116) [netty-all-4.0.25.Final.jar:4.0.25.Final]
    at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:137) [netty-all-4.0.25.Final.jar:4.0.25.Final]
    at java.lang.Thread.run(Thread.java:745) [rt.jar:1.7.0_60]

lsof -p 5604 | grep socket | wc -l

无法看到与Netstat的任何打开或挂起的连接...

我使用ReadTimeoutHandler关闭未使用的连接，并使用以下exceptionHandler代码：

@Override
public void exceptionCaught(ChannelHandlerContext ctx, Throwable cause) throws Exception {
  if (cause instanceof ReadTimeoutException) {
    logger.debug("Read timeout - close connection");
  } else {
    logger.info(cause.getMessage());
  }
  ctx.close();
}

服务器引导程序如下所示：

ServerBootstrap b = new ServerBootstrap();
b.group(bossGroup, workerGroup).channel(NioServerSocketChannel.class).childHandler(new ChannelInitializer<SocketChannel>() {
@Override
public void initChannel(SocketChannel ch) throws Exception {
    ch.pipeline().addLast(new ReadTimeoutHandler(60));
    ch.pipeline().addLast(new LoggingHandler(mySpec.getPortLookupKey().toLowerCase()));
    ch.pipeline().addLast(new RawMessageEncoder());
    ch.pipeline().addLast(new RawMessageDecoder());
    ch.pipeline().addLast(new RequestServerHandler(ctx.getWorkManager(), factory));
}
}).option(ChannelOption.SO_BACKLOG, 128).childOption(ChannelOption.SO_KEEPALIVE, true);

ChannelFuture channelFuture = b.bind(port).sync();

更新2：根据请求，我将日志处理程序移到ReadTimeouthandler前面，以下是日志。客户端通常断开连接的情况：

09:41:39,755 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 => /127.0.0.1:4300] REGISTERED
09:41:39,756 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 => /127.0.0.1:4300] ACTIVE
09:41:39,810 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 => /127.0.0.1:4300] RECEIVED(1024B)
09:41:39,813 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 => /127.0.0.1:4300] RECEIVED(1024B)
09:41:39,814 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 => /127.0.0.1:4300] RECEIVED(150B)
09:41:40,854 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 => /127.0.0.1:4300] WRITE(1385B)
09:41:40,855 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 => /127.0.0.1:4300] FLUSH
09:41:40,861 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 :> /127.0.0.1:4300] INACTIVE
09:41:40,864 [3-1] [id: 0xca6601a2, /127.0.0.1:64258 :> /127.0.0.1:4300] UNREGISTERED

客户端不断开连接的情况：

10:04:24,104 [3-1] [id: 0x48076684, /127.0.0.1:50525 => /127.0.0.1:4300] REGISTERED
10:04:24,107 [3-1] [id: 0x48076684, /127.0.0.1:50525 => /127.0.0.1:4300] ACTIVE
10:04:24,594 [3-1] [id: 0x48076684, /127.0.0.1:50525 => /127.0.0.1:4300] RECEIVED(1024B)
10:04:24,597 [3-1] [id: 0x48076684, /127.0.0.1:50525 => /127.0.0.1:4300] RECEIVED(1024B)
10:04:24,598 [3-1] [id: 0x48076684, /127.0.0.1:50525 => /127.0.0.1:4300] RECEIVED(150B)
10:04:25,638 [3-1] [id: 0x48076684, /127.0.0.1:50525 => /127.0.0.1:4300] WRITE(1383B)
10:04:25,639 [3-1] [id: 0x48076684, /127.0.0.1:50525 => /127.0.0.1:4300] FLUSH
10:05:25,389 [3-1] [id: 0x48076684, /127.0.0.1:50525 => /127.0.0.1:4300] CLOSE()
10:05:25,390 [3-1] [id: 0x48076684, /127.0.0.1:50525 :> /127.0.0.1:4300] CLOSE()
10:05:25,390 [3-1] [id: 0x48076684, /127.0.0.1:50525 :> /127.0.0.1:4300] INACTIVE
10:05:25,394 [3-1] [id: 0x48076684, /127.0.0.1:50525 :> /127.0.0.1:4300] UNREGISTERED

所以收盘前有60秒的差距（正如ReadTimeouthHandler所预期的那样）

经过更多的分析，我有了这样的印象，即使与客户端正常断开连接，打开的文件数量也会增加！而且，在这种情况下没有CLOSE（）...

Answer 1

也许这与这个棘手的问题有关https://github.com/netty/netty/issues/1731

这是预期的行为，是无法改变的。JVM正在发出信号，表示它无法接受该通道--因此无法启动连接，也无法发送响应。客户端将看到连接失败。如果您有一个负载平衡器，它应该对备用主机重试，或者代表您的应用程序返回503。