Java Netty ReadTimeoutHandler和CLOSE\u等待

我正在使用带有readTimeoutHandler的Netty服务器（4.0.33）上工作，但由于某些原因，当存在大量流量时，许多套接字会以关闭等待状态堆积起来。我已将readTimeoutHandler超时设置为5秒，但在收到请求时将其删除。以下是管道初始值设定项：

@Override
protected void initChannel(Channel channel) throws Exception {

    ChannelPipeline p = channel.pipeline();
    p.addLast("decoder", new HttpRequestDecoder(4096, 8192, 64 * 1024));
    p.addLast("encoder", new HttpResponseEncoder());
    p.addLast("chunkedwritehandler", new ChunkedWriteHandler());
    p.addLast("readtimeout", new ReadTimeoutHandler(5, TimeUnit.SECONDS) 
    p.addLast("writetimeout", new WriteTimeoutHandler(0, TimeUnit.SECONDS));
    p.addLast("handler", new MyHandler());
}

以及我的简化处理程序方法：

@Override
public void channelRead0(ChannelHandlerContext ctx, HttpObject msg) throws Exception {

    // If it's an HTTP request
    if(msg instanceof HttpRequest)
    {
        this.request = (HttpRequest) msg;
        this.prepareHttpDecoder(ctx);
    }

    // If it's HTTP content
    else if(msg instanceof HttpContent)
    {
        HttpContent chunk = (HttpContent) msg;

        try {
            // Offer the chunk to the HTTP decoder
            httpDecoder.offer(chunk);
        }
        catch(HttpPostRequestDecoder.ErrorDataDecoderException e) {

            // Log the error

            // Write the response with a force close
            ctx.writeAndFlush("DecoderError").addListener(ChannelFutureListener.CLOSE);

            // Add a 15 second readTimeout
            ctx.pipeline().replace("readtimeout", "readtimeout", new ReadTimeoutHandler(15, TimeUnit.SECONDS));
            return;
        }

        // Handle the final chunk
        if(chunk instanceof LastHttpContent)
        {
            // We've received the whole request, cancel the readTimeout
            ctx.pipeline().replace("readtimeout", "readtimeout", new ReadTimeoutHandler(0, TimeUnit.SECONDS));

            // Process the request

            resetHttpDecoder();
        }
    }

@Override
public void channelUnregistered(ChannelHandlerContext ctx) throws Exception {

    // Log things, clean session

    // Close the context
    if(ctx != null)
    {
        ctx.close();
    }
}

@Override
public void exceptionCaught(ChannelHandlerContext ctx, Throwable e) throws Exception {

    // Log things, clean the session

    // Close the context
    if(ctx != null)
    {
        ctx.close();
    }
}

对于仍然处于CLOSE_WAIT状态的套接字，我得到以下Netty日志：

2016-02-18 01:46:34,994 DEBUG [nioEventLoopGroup-3-2] [id: 0xbe795f05, /10.221.2.135:44859 => /10.221.1.128:631] REGISTERED
2016-02-18 01:46:34,994 DEBUG [nioEventLoopGroup-3-2] [id: 0xbe795f05, /10.221.2.135:44859 => /10.221.1.128:631] ACTIVE
2016-02-18 01:46:39,995 DEBUG [nioEventLoopGroup-3-2] [id: 0xbe795f05, /10.221.2.135:44859 => /10.221.1.128:631] CLOSE()
2016-02-18 01:46:39,995 DEBUG [nioEventLoopGroup-3-2] [id: 0xbe795f05, /10.221.2.135:44859 :> /10.221.1.128:631] CLOSE()
2016-02-18 01:46:39,995 DEBUG [nioEventLoopGroup-3-2] [id: 0xbe795f05, /10.221.2.135:44859 :> /10.221.1.128:631] INACTIVE
2016-02-18 01:46:39,995 DEBUG [nioEventLoopGroup-3-2] [id: 0xbe795f05, /10.221.2.135:44859 :> /10.221.1.128:631] UNREGISTERED
2016-02-18 01:46:39,995 DEBUG [nioEventLoopGroup-3-2] [id: 0xbe795f05, /10.221.2.135:44859 :> /10.221.1.128:631] CLOSE()

---

看起来，当Netty没有收到任何数据时，readTimeout将在5秒后按预期发生。但是，套接字仍处于关闭状态，请等待。有人知道我可能做错了什么吗？事先谢谢

尝试将

ReadTimeoutHandler

放在管道的最开始处，看看是否有区别。尝试将

ReadTimeoutHandler

放在管道的最开始处，看看是否有区别。