如何使用Netty在单独的线程池中执行业务逻辑处理程序

问题描述

我有一个需要执行一些业务逻辑的处理程序，我希望在一个单独的线程池中执行该处理程序，以免阻止io事件循环.我已按照 http://netty中的指定，将DefaultEventExecutorGroup添加到管道中. io/4.0/api/io/netty/channel/ChannelPipeline.html javadoc和 http://netty.io/wiki/new-and-noteworthy-in-4.0.html#no-more-executionhandler-- -其核心 Wiki:

I have a handler that needs to execute some business logic and I want that to be executed in a separate thread pool to not block the io event loop. I have added DefaultEventExecutorGroup into the pipeline as specified in http://netty.io/4.0/api/io/netty/channel/ChannelPipeline.html javadoc and http://netty.io/wiki/new-and-noteworthy-in-4.0.html#no-more-executionhandler---its-in-the-core wiki:

ch.pipeline().addLast(new DefaultEventExecutorGroup(10), new ServerHandler());

出于测试目的，我的ServerHandler只是将当前线程置于休眠状态5秒钟:

Just for testing purposes my ServerHandler just puts the current thread to sleep for 5 seconds:

protected void channelRead0(ChannelHandlerContext ctx, Command cmd) throws Exception {
    System.out.println("Starting.");

    try {
        Thread.currentThread().sleep(5000);
    } catch (InterruptedException e) {
        // TODO Auto-generated catch block
        e.printStackTrace();
    }

    System.out.println("Finished.");        
}

但是显然，业务逻辑仍然是同步执行的:

But apparently the business logic is still executed synchronously:

Starting.
Finished.
Starting.
Finished.
Starting.
Finished.

我想念什么?

推荐答案

如果您的目标不是阻止IO事件循环-您做对了.但是由于特定于网络，您的处理程序将始终附加在EventExecutorGroup的同一线程上，因此可以预期您在上述情况下的行为.

In case your goal is not to block IO event loop - you did it right. But due to netty specific, your handler will be always attached to the same thread of EventExecutorGroup and thus behavior you described above is expected.

如果要在到达时立即并行执行阻止操作，则需要使用另一种方式-单独的ThreadPoolExecutor.像这样:

In case you want to execute blocking operation in parallel as soon as it arrives you need to use the another way - separate ThreadPoolExecutor. Like this:

ch.pipeline().addLast(new ServerHandler(blockingThreadPool));

其中blockingThreadPool是常规ThreadPoolExecutor.

例如:

ExecutorService blockingThreadPool = Executors.newFixedThreadPool(10);

现在，在您的逻辑处理程序中，您可以像这样将阻止任务提交给该执行器:

Now, within your logic handler you can submit blocking tasks to this executor like this:

protected void channelRead0(ChannelHandlerContext ctx, Command cmd) throws Exception {

    blockingIOProcessor.execute(new Runnable() {
        @Override
        public void run() {
            System.out.println("Starting.");

            try {
                Thread.currentThread().sleep(5000);
            } catch (InterruptedException e) {
                // TODO Auto-generated catch block
                e.printStackTrace();
            }

            System.out.println("Finished.");
        }
    });

}

如果需要，您还可以将上下文传递给此可运行对象，以便在处理完成时返回响应.

You can also pass context to this runnable in order to return the response back when processing is finished if needed.

这篇关于如何使用Netty在单独的线程池中执行业务逻辑处理程序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！