DefaultEventExecutorGroup线程不断增长，服务器对请求的响应速度很慢

问题描述

我正在使用Netty-4.1.0-Beta5，我正在使用DefaultEventExecutorGroup(具有2个线程)作为长时间运行的处理程序(解除阻塞I/O线程)，在服务器运行了几个小时之后，我看到了大约2155个组每个都有1个线程

I am using Netty-4.1.0-Beta5, i am using DefaultEventExecutorGroup(with 2 threads) for long running handler (to unblock I/O thread), after running server for couple of hours, i am seeing around 2155 groups and each with 1 thread

 defaultEventExecutorGroup-10-1 tid=134 [WAITING]
sun.misc.Unsafe.park(boolean, long) Unsafe.java
java.util.concurrent.locks.LockSupport.parkNanos(Object, long) LockSupport.java:215
java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.awaitNanos(long) AbstractQueuedSynchronizer.java:2078
java.util.concurrent.LinkedBlockingQueue.poll(long, TimeUnit) LinkedBlockingQueue.java:467
io.netty.util.concurrent.SingleThreadEventExecutor.takeTask() SingleThreadEventExecutor.java:194
io.netty.util.concurrent.DefaultEventExecutor.run() DefaultEventExecutor.java:54
io.netty.util.concurrent.SingleThreadEventExecutor$4.run() SingleThreadEventExecutor.java:703
io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run() DefaultThreadFactory.java:137
java.lang.Thread.run() Thread.java:745

..

defaultEventExecutorGroup-1900-1 tid=2114 [WAITING]

大多数HTTP请求都需要20秒钟以上才能响应，这是我的代码:

and most of HTTP requests taking more than 20 seconds to respond, here is my code:

class NettyServer extends Logging {
    val allChannels = new DefaultChannelGroup("enter", GlobalEventExecutor.INSTANCE)
    val webSocketConnections = new WebSocketConnections("websockets")
    val sslManager: Option[SslManager] = Some(new SslManager(this))

    def start(): Unit = {
        allChannels.clear()
        val bootstrap = new ServerBootstrap()
        bootstrap.group(new NioEventLoopGroup(), new NioEventLoopGroup())
        bootstrap.channel(classOf[NioServerSocketChannel])
        bootstrap.childOption[java.lang.Boolean](ChannelOption.TCP_NODELAY, true)
        bootstrap.childHandler(new PipelineFactory(this))
        val bindFutures = "0.0.0.0".split(",").map(address => {
        address.trim match
            case "0.0.0.0" =>
                bootstrap.bind(9999)
            case _ if (!address.isEmpty) =>
                bootstrap.bind(address, 9999)
        })

        allChannels.addAll(bindFutures.map(_.channel).toList)
        val latch = new CountDownLatch(bindFutures.length)
        val bindFutureListener = new ChannelFutureListener {
            def operationComplete(future: ChannelFuture) = {
                latch.countDown
            }
        }
        bindFutures.foreach(_.addListener(bindFutureListener))
        latch.await
    }

    def stop(): Unit = {
        val future = allChannels.close()
        future.awaitUninterruptibly()
        allChannels.clear()
    }
}

class PipelineFactory(server: NettyServer) extends ChannelInitializer[SocketChannel] with Logging {
  override def initChannel(ch: SocketChannel) {
    val pipeline = ch.pipeline
    pipeline.addLast("ssl", new SslHandler(server.sslManager.get.createSSLEngine()))
    pipeline.addLast("httpRequestDecoder", new HttpRequestDecoder(4096, 8192, 8192, false))
    pipeline.addLast("httpResponseEncoder", new HttpResponseEncoder())
    pipeline.addLast("chunkAggregator", new HttpObjectAggregator(65536))
    pipeline.addLast("idleStateHandler", new IdleStateHandler(0, 0, 0))
    val corsConfig = CorsConfig.withAnyOrigin()
    pipeline.addLast("corsHandler", new CorsHandler(corsConfig))
    pipeline.addLast(new DefaultEventExecutorGroup(2), "handler", new MyHandler())
  }
}

任何想法，我的代码有什么问题?如果我使用curl在紧密循环中发送10000个HTTP请求，那么我可以重新创建此问题.感谢您的帮助！

Any idea, what is wrong with my code? if i send 10000 HTTP requests in tight loop using curl then i could recreate this issue.Thanks for the help!

推荐答案

您只需创建一次DefaultEventExecutorGroup并共享该实例.目前，您为每个频道创建了

You need to create the DefaultEventExecutorGroup only one time and share the instance. At the moment you create it for each Channel

这篇关于DefaultEventExecutorGroup线程不断增长，服务器对请求的响应速度很慢的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！