Netty（四）NIO-优化与源码

Netty优化与源码

1. 优化

1.1 扩展序列化算法

序列化，反序列化主要用于消息正文的转换。
序列化：将java对象转为要传输对象(byte[]或json，最终都是byte[])
反序列化：将正文还原成java对象。

//java自带的序列化
// 反序列化
byte[] body = new byte[bodyLength];
byteByf.readBytes(body);
ObjectInputStream in = new ObjectInputStream(new ByteArrayInputStream(body));
Message message = (Message) in.readObject();
message.setSequenceId(sequenceId);
// 序列化
ByteArrayOutputStream out = new ByteArrayOutputStream();
new ObjectOutputStream(out).writeObject(message);
byte[] bytes = out.toByteArray();

为了支持更多序列化算法，抽象一个 Serializer 接口，提供两个实现，将实现加入了枚举类 Serializer.Algorithm 中：

enum SerializerAlgorithm implements Serializer {// Java 实现Java {@Overridepublic <T> T deserialize(Class<T> clazz, byte[] bytes) {try {ObjectInputStream in = new ObjectInputStream(new ByteArrayInputStream(bytes));Object object = in.readObject();return (T) object;} catch (IOException | ClassNotFoundException e) {throw new RuntimeException("SerializerAlgorithm.Java 反序列化错误", e);}}@Overridepublic <T> byte[] serialize(T object) {try {ByteArrayOutputStream out = new ByteArrayOutputStream();new ObjectOutputStream(out).writeObject(object);return out.toByteArray();} catch (IOException e) {throw new RuntimeException("SerializerAlgorithm.Java 序列化错误", e);}}}, // Json 实现(引入了 Gson 依赖)Json {@Overridepublic <T> T deserialize(Class<T> clazz, byte[] bytes) {return new Gson().fromJson(new String(bytes, StandardCharsets.UTF_8), clazz);}@Overridepublic <T> byte[] serialize(T object) {return new Gson().toJson(object).getBytes(StandardCharsets.UTF_8);}};// 需要从协议的字节中得到是哪种序列化算法public static SerializerAlgorithm getByInt(int type) {SerializerAlgorithm[] array = SerializerAlgorithm.values();if (type < 0 || type > array.length - 1) {throw new IllegalArgumentException("超过 SerializerAlgorithm 范围");}return array[type];}
}

增加配置类和配置文件：

public abstract class Config {static Properties properties;static {try (InputStream in = Config.class.getResourceAsStream("/application.properties")) {properties = new Properties();properties.load(in);} catch (IOException e) {throw new ExceptionInInitializerError(e);}}public static int getServerPort() {String value = properties.getProperty("server.port");if(value == null) {return 8080;} else {return Integer.parseInt(value);}}public static Serializer.Algorithm getSerializerAlgorithm() {String value = properties.getProperty("serializer.algorithm");if(value == null) {return Serializer.Algorithm.Java;} else {return Serializer.Algorithm.valueOf(value);}}
}

配置文件

serializer.algorithm=Json

修改编解码器

/*** 必须和 LengthFieldBasedFrameDecoder 一起使用，确保接到的 ByteBuf 消息是完整的*/
public class MessageCodecSharable extends MessageToMessageCodec<ByteBuf, Message> {@Overridepublic void encode(ChannelHandlerContext ctx, Message msg, List<Object> outList) throws Exception {// 3. 1 字节的序列化方式 jdk 0 , json 1out.writeByte(Config.getSerializerAlgorithm().ordinal());byte[] bytes = Config.getSerializerAlgorithm().serialize(msg);}@Overrideprotected void decode(ChannelHandlerContext ctx, ByteBuf in, List<Object> out) throws Exception {byte serializerAlgorithm = in.readByte(); // 0 或 1// 找到反序列化算法Serializer.Algorithm algorithm = Serializer.Algorithm.values()[serializerAlgorithm];// 确定具体消息类型Class<? extends Message> messageClass = Message.getMessageClass(messageType);Message message = algorithm.deserialize(messageClass, bytes);out.add(message);}
}

1.2 参数调优

CONNECT_TIMEOUT_MILLIS

• 属于SocketChannel参数，用在客户端建立连接时，如超时则抛出timeout异常
• SO_TIMEOUT主要用在阻塞IO，阻塞IO中accept，read等都是无限等待的

Bootstrap bootstrap = new Bootstrap().group(group).option(ChannelOption.CONNECT_TIMEOUT_MILLIS, 5000).channel(NioServerSocketChannel.class).handler(new LoggingHandler());

附源码部分 io.netty.channel.nio.AbstractNioChannel.AbstractNioUnsafe#connect

@Override
public final void connect(final SocketAddress remoteAddress, final SocketAddress localAddress, final ChannelPromise promise) {// ...// Schedule connect timeout.int connectTimeoutMillis = config().getConnectTimeoutMillis();if (connectTimeoutMillis > 0) {connectTimeoutFuture = eventLoop().schedule(new Runnable() {@Overridepublic void run() {                ChannelPromise connectPromise = AbstractNioChannel.this.connectPromise;ConnectTimeoutException cause =new ConnectTimeoutException("connection timed out: " + remoteAddress); // 断点2if (connectPromise != null && connectPromise.tryFailure(cause)) {close(voidPromise());}}}, connectTimeoutMillis, TimeUnit.MILLISECONDS);}// ...
}

SO_BACKLOG

• 属于ServerSocketChannel参数
  
  sync queue - 半连接队列
  • 大小通过 /proc/sys/net/ipv4/tcp_max_syn_backlog 指定，在 syncookies 启用的情况下，逻辑上没有最大值限制，这个设置便被忽略
    accept queue - 全连接队列
  • 其大小通过 /proc/sys/net/core/somaxconn 指定，在使用 listen 函数时，内核会根据传入的 backlog 参数与系统参数，取二者的较小值
  • 如果 accpet queue 队列满了，server 将发送一个拒绝连接的错误信息到 client
    netty中可通过option(ChannelOption.SO_BACKLOG,值)来设置大小

public class DefaultServerSocketChannelConfig extends DefaultChannelConfigimplements ServerSocketChannelConfig {private volatile int backlog = NetUtil.SOMAXCONN;// ...默认大小
}

ulimit -n

• 限制一个进程打开最大文件描述符的数目，属于操作系统参数

TCP_NODELAY

• nagle算法的延迟，一般设为true不延迟，数据赞属于 SocketChannal 参数

SO_SNDBUF & SO_RECVBUF

滑动接口的参数，现在的操作系统会根据实际情况自动调整。

• SO_SNDBUF 属于 SocketChannal 参数
• SO_RCVBUF 既可用于 SocketChannal 参数，也可以用于 ServerSocketChannal 参数（建议设置到 ServerSocketChannal 上）

ALLOCATOR

ByteBuf分配器，属于 SocketChannal 参数，用来分配 ByteBuf， ctx.alloc()。源码详解P128

RCVBUF_ALLOCATOR

• 属于 SocketChannal 参数，控制 netty 接收缓冲区大小。源码详解：P129
• 负责入站数据的分配，决定入站缓冲区的大小（并可动态调整），统一采用 direct 直接内存，具体池化还是非池化由 allocator 决定

1.3 RPC 框架

通过反射获取配置

public class ServicesFactory {static Properties properties;static Map<Class<?>, Object> map = new ConcurrentHashMap<>();static {try {InputStream in = Config.class.getResourceAsStream("/application.properties");properties = new Properties();properties.load(in);Set<String> names = properties.stringPropertyNames();for (String name : names) {if (name.endsWith("Services")) {Class<?> interfaceClass = Class.forName(name);Class<?> instanceClass = Class.forName(properties.getProperty(name));map.put(interfaceClass, instanceClass.newInstance());}}} catch (IOException | ClassNotFoundException | InstantiationException | IllegalAccessException e) {throw new ExceptionInInitializerError(e);}}public static <T> T getService(Class<T> interfaceClass) {return (T) map.get(interfaceClass);}
}

RPC消息处理器

@ChannelHandler.Sharable
public class RpcRequestMessageHandler extends SimpleChannelInboundHandler<RpcRequestMessage> {@Overrideprotected void channelRead0(ChannelHandlerContext ctx, RpcRequestMessage message) {RpcResponseMessage response = new RpcResponseMessage();try {HelloService service = (HelloService) ServicesFactory.getService(Class.forName(message.getInterfaceName()));Method method = service.getClass().getMethod(message.getMethodName(), message.getParameterTypes());Object invoke = method.invoke(service, message.getParameterValue());response.setReturnValue(invoke);} catch (Exception e) {e.printStackTrace();response.setExceptionValue(e);}ctx.writeAndFlush(response);}//本地调试public static void main(String[] args) throws ClassNotFoundException, NoSuchMethodException, InvocationTargetException, IllegalAccessException {RpcRequestMessage message = new RpcRequestMessage(1,"com.aric.server.service.HelloService","sayHello",String.class,new Class[]{String.class},new Object[]{"aric"});HelloService service = (HelloService) ServicesFactory.getService(Class.forName(message.getInterfaceName()));Method method = service.getClass().getMethod(message.getMethodName(), message.getParameterTypes());Object invoke = method.invoke(service, message.getParameterValue());System.out.println(invoke);}
}

客户端优化，抽取使用代理对象发送消息

/*** 使用代理对象替换，主线程发送* NioEventLoop线程接收结果，需要线程间通信，使用promise对象接收结果* @author* @created by xuyu on 2023/9/23-23:10*/
@Slf4j
public class RpcClientManager {public static void main(String[] args) {//后期创建代理类优化发送结构getChannel().writeAndFlush(new RpcRequestMessage(1,"com.aric.server.service.HelloService","sayHello",String.class,new Class[]{String.class},new Object[]{"test"}));//使用代理发送HelloService service = getProxyService(HelloService.class);service.sayHello("test");}//创建代理类public static <T> T getProxyService(Class<T> serviceClass) {ClassLoader loader = serviceClass.getClassLoader();  //当前类加载器Class[] interfaces = new Class[]{serviceClass};//代理类要实现的接口//jdk自带的代理Object o = Proxy.newProxyInstance(loader, interfaces, (proxy, method, arg) -> {//proxy代理对象，method：代理方法，arg：代理参数//1.将方法调用转换为消息对象RpcRequestMessage message = new RpcRequestMessage(SequenceIdGenerator.nextId(),serviceClass.getName(),method.getName(),method.getReturnType(),method.getParameterTypes(),arg);//2.将消息对象发送出去getChannel().writeAndFlush(message);//3.TODO：待优化异步等待返回结果return null;});return (T)o;}private static Channel channel = null;private static final Object LOCK = new Object();//单例构造获取唯一channel对象public static Channel getChannel() {if (channel != null) {return channel;}synchronized (LOCK) {if (channel != null) {return channel;}initChannel();return channel;}}//初始化channel方法private static void initChannel() {NioEventLoopGroup group = new NioEventLoopGroup();LoggingHandler LOGGING_HANDLER = new LoggingHandler(LogLevel.DEBUG);MessageCodecSharable MESSAGE_CODEC = new MessageCodecSharable();RpcResponseMessageHandler RPC_HANDLER = new RpcResponseMessageHandler();Bootstrap bootstrap = new Bootstrap();bootstrap.channel(NioSocketChannel.class);bootstrap.group(group);bootstrap.handler(new ChannelInitializer<SocketChannel>() {@Overrideprotected void initChannel(SocketChannel ch) throws Exception {ch.pipeline().addLast(new ProtocolFrameDecoder());ch.pipeline().addLast(LOGGING_HANDLER);ch.pipeline().addLast(MESSAGE_CODEC);ch.pipeline().addLast(RPC_HANDLER);}});try {channel = bootstrap.connect("localhost", 8080).sync().channel();//改为异步channel.closeFuture().addListener(future -> {group.shutdownGracefully();});} catch (InterruptedException e) {log.debug("client error", e);}}
}

优化：线程间通信：异步获取返回结果

//创建代理类public static <T> T getProxyService(Class<T> serviceClass) {ClassLoader loader = serviceClass.getClassLoader();  //当前类加载器Class[] interfaces = new Class[]{serviceClass};//代理类要实现的接口//jdk自带的代理Object o = Proxy.newProxyInstance(loader, interfaces, (proxy, method, arg) -> {//proxy代理对象，method：代理方法，arg：代理参数//1.将方法调用转换为消息对象int sequenceId = SequenceIdGenerator.nextId();RpcRequestMessage message = new RpcRequestMessage(sequenceId,serviceClass.getName(),method.getName(),method.getReturnType(),method.getParameterTypes(),arg);//2.将消息对象发送出去getChannel().writeAndFlush(message);//3.返回//准备好空的promise对象来接收结果,参数为指定promise对象异步接收结果的线程DefaultPromise<Object> promise = new DefaultPromise<>(getChannel().eventLoop());RpcResponseMessageHandler.PROMISE.put(sequenceId, promise);
//            promise.addListener(future -> {
//                //创建线程处理任务
//            });//原线程等待promise的结果promise.await();if (promise.isSuccess()) {return promise.getNow();} else {throw new RuntimeException(promise.cause());}});return (T) o;}

/*** rpc响应消息处理器*/
@Slf4j
@ChannelHandler.Sharable
public class RpcResponseMessageHandler extends SimpleChannelInboundHandler<RpcResponseMessage> {//序号-promise<结果类型>，多个线程访问，用于异步接收rpc调用的返回结果public static final Map<Integer, Promise<Object>> PROMISE = new ConcurrentHashMap<>();@Overrideprotected void channelRead0(ChannelHandlerContext ctx, RpcResponseMessage msg) throws Exception {//拿到空的promisePromise<Object> promise = PROMISE.remove(msg.getSequenceId());  //返回并移除if (promise != null) {Object returnValue = msg.getReturnValue();Exception exceptionValue = msg.getExceptionValue();if (exceptionValue != null) {promise.setFailure(exceptionValue);} else {promise.setSuccess(returnValue);}}System.out.println(msg);}

代码：https://gitee.com/xuyu294636185/netty-demo.git

2. 源码

2.1 netty启动剖析

        //1. netty中使用EventLoopGroup(Nio boss线程),来封装线程和selectorSelector selector = Selector.open();//创建NioServerSocketChannel，同时初始化它关联的handler，以及为原生ssc存储configNioServerSocketChannel attachment = new NioServerSocketChannel();ServerSocketChannel ssc = ServerSocketChannel.open();ssc.configureBlocking(false);//2.启动nio boss线程执行//建立selector和channel的注册,sscKey是事件的句柄，是将来事件发生后，通过它可以知道事件和哪个channel的事件SelectionKey sscKey = ssc.register(selector, 0, attachment);ssc.bind(new InetSocketAddress(8080));//表示sscKey只关注accept事件sscKey.interestOps(SelectionKey.OP_ACCEPT);

启动流程

EventLoop

EventLoop重要组成：selector，线程，任务队列
EventLoop既会处理io事件，也会处理普通任务和定时任务

1. selector何时创建？
   在构造方法创建时通过SelectorProvider.openSelector();
2. eventloop为什么会有两个selector成员？
   为了在遍历selectedKey时提高性能。
   一个是原始的unwrappedselector（底层是hashset实现），一个是包装后的selector（底层是数组实现）
3. eventLoop的nio线程在何时启动？
   在首次调用exectue方法时executor中将当前线程赋给nio线程，并通过state状态控制位只会启动一次
4. 提交普通任务会不会结束select阻塞？
   会
   int selectedKeys = selector.select(timeoutMillis);
   protected void wakeup(boolean inEventLoop) {
   if(!inEventLoop && wakeUp.compareAndSet(false,true)) {
   selector.wakeup();
   }
   }
5. wakeup方法理解
   inEventLoop：用于判断当前wakeup线程是否和nio线程是否相同，不同才能进入。
   wakeUp：原子Boolean变量，如果有多个线程来提交任务，为了避免wakeup被频繁调用。只有一个成功。
6. 每次循环时，什么时候会进入SelectStrategy.SELECT分支？
   public void run(){
   for(;😉 {
   switch(selectStrategy.calculateStrategy(selectNowSupplier, hasTask())) {
   case SelectStrategy.CONTINUE:
   continue;
   case SelectStrategy.BUSY_WAIT:
   case SelectStrategy.SELECT:
   select(wakeUp.getAndSet(false));
   if(wakeUp.get()) {…}
   default:
   }
   }
   }
   public int calculateStrategy(IntSupplier supplier,boolean hasTasks) {
   return hasTasks ? suppplier.get() : SelectStrategy.SELECT;
   }
   没有任务时，才会进入SELECT。
   当有任务时，会调用SelectNow方法，顺便拿到io事件。
7. 何时会select阻塞，阻塞多久？
   long currentTimeNanos = System.nanoTime();
   long selectDeadLineNanos = currentTimeNanos + delayNanos(cuurrentTimeNanos);
   for(;😉{
   long timeoutMillis = (selectDeadLineNanos - currentTimeNanos + 500000L) / 1000000L;
   int selectedKeys = selector.select(timeoutMillis);
   }
   没有定时任务的情况
   selectDeadLineNanos：截至时间 = 当前时间 + 1s
   timeoutMillis:超时时间 = 1s + 0.5ms
8. nio空轮询bug在哪体现，如何解决？
   for(;😉{
   long timeoutMillis = (selectDeadLineNanos - currentTimeNanos + 500000L) / 1000000L;
   int selectedKeys = selector.select(timeoutMillis);
   }
   在select中没有阻塞，一直在死循环
   解决：引入selectCnt，每循环一次++，当超过设置的阈值（默认512），selectRebuildSelector(selectCnt)重新创建一个selector，替换旧的。
9. ioRatio控制什么，设置为100有何作用？
   if(ioRatio == 100) {
   processSelectedKeys(); //处理所有ioio事件
   runAllTasks();
   } else {
   long ioStartTime = System.nanoTime();
   processSelectedKeys();
   long ioTime = System.nanoTime() - ioStartTime;
   runAllTasks(ioTime * (100 - ioRatio) / ioRatio); //避免普通事件执行时间太长
   }
   ioRatio控制处理io事件所占用的事件比例50%，ioTime代表执行io事件处理耗时。
10. selectedKeys优化，在哪区分不同事件类型。
    selectedKeys由hashset集合替换为数组实现。
    private void processSelectedKeys() [
    if(selectedKeys != null) {
    processSelectedKeysOptimized(); //优化后的
    } else {
    processSelectedKeysPlain(selector.selectedKeys()); //原始的
    }
    }
    private void processSelectedKeysOptimized() {
    for(int i = 0;i < selectedKeys.size; ++i) {
    SelectionKey k = selectedKeys.keys[i];
    selectedKeys.keys[i] = null;
    Objected a = k.attachment();
    if(a instanceof AbstractNioChannel) {
    processSelectedKey(k, (AbstractNioChannel) a); //处理具体的事件类型
    }
    }
    }

accept流程

1. selector.select()阻塞直到事件发生
2. 遍历处理selectedKeys
3. 拿到一个key，判断事件类型是否为accpet
4. 创建socketChannel，设置非阻塞
5. 将socketChannel注册到selector
6. 关注selectionKey的read事件。

read流程

1. selector.select()阻塞直到事件发生
2. 遍历处理selectedKeys
3. 拿到一个key，判断事件类型是否为read
4. 读取操作