NettyUtils

最新推荐文章于 2023-03-10 11:18:17 发布

宝哥大数据

最新推荐文章于 2023-03-10 11:18:17 发布

阅读量456

点赞数

分类专栏： # spark

本文链接：https://blog.csdn.net/wuxintdrh/article/details/89322916

版权

spark 专栏收录该内容

145 篇文章 14 订阅

订阅专栏

NettyUtils 用于基于NIO或EPOLL创建Netty概念

Utilities for creating various Netty constructs based on whether we’re using EPOLL or NIO.

1.1、getClientChannelClass

根据IOMode,获取对应的SocketChannel

    /**
     * Returns the correct (client) SocketChannel class based on IOMode.
     */
    public static Class<? extends Channel> getClientChannelClass(IOMode mode) {
        switch (mode) {
            case NIO:
                return NioSocketChannel.class;
            case EPOLL:
                return EpollSocketChannel.class;
            default:
                throw new IllegalArgumentException("Unknown io mode: " + mode);
        }
    }

1.2、createEventLoop

根据IOMode，创建Netty的EventLoopGroup。

    /**
     * Creates a Netty EventLoopGroup based on the IOMode.
     */
    public static EventLoopGroup createEventLoop(IOMode mode, int numThreads, String threadPrefix) {
        ThreadFactory threadFactory = createThreadFactory(threadPrefix);

        switch (mode) {
            case NIO:
                return new NioEventLoopGroup(numThreads, threadFactory);
            case EPOLL:
                return new EpollEventLoopGroup(numThreads, threadFactory);
            default:
                throw new IllegalArgumentException("Unknown io mode: " + mode);
        }
    }

1.3、createPooledByteBufAllocator

创建一个汇集ByteBuf但对本地线程缓存禁用的分配器。为什么要对本地线程缓存禁用？因为ByteBuf都是由事件循环线程分配，所以线程本地缓存对于TransportClient是禁用的。但是ByteBuf的释放却是由Executor线程，而不是事件循环线程来完成。本地线程缓存经常会延迟ByteBuf的回收，导致巨大的内存消耗。

    /**
     * Create a pooled ByteBuf allocator but disables the thread-local cache. Thread-local caches
     * are disabled for TransportClients because the ByteBufs are allocated by the event loop thread,
     * but released by the executor thread rather than the event loop thread. Those thread-local
     * caches actually delay the recycling of buffers, leading to larger memory usage.
     */
    public static PooledByteBufAllocator createPooledByteBufAllocator(
            boolean allowDirectBufs,
            boolean allowCache,
            int numCores) {
        if (numCores == 0) {
            numCores = Runtime.getRuntime().availableProcessors();
        }
        return new PooledByteBufAllocator(
                allowDirectBufs && PlatformDependent.directBufferPreferred(),
                Math.min(getPrivateStaticField("DEFAULT_NUM_HEAP_ARENA"), numCores),
                Math.min(getPrivateStaticField("DEFAULT_NUM_DIRECT_ARENA"), allowDirectBufs ? numCores : 0),
                getPrivateStaticField("DEFAULT_PAGE_SIZE"),
                getPrivateStaticField("DEFAULT_MAX_ORDER"),
                allowCache ? getPrivateStaticField("DEFAULT_TINY_CACHE_SIZE") : 0,
                allowCache ? getPrivateStaticField("DEFAULT_SMALL_CACHE_SIZE") : 0,
                allowCache ? getPrivateStaticField("DEFAULT_NORMAL_CACHE_SIZE") : 0
        );
    }

1.4、createFrameDecoder

对从管道中读取的ByteBuf按照数据帧进行解析；

    /**
     * Creates a LengthFieldBasedFrameDecoder where the first 8 bytes are the length of the frame.
     * This is used before all decoders.
     */
    public static TransportFrameDecoder createFrameDecoder() {
        return new TransportFrameDecoder();
    }