实现简易rp框架(完结)-实现服务注销和负载均衡

在RpcServer关闭的时候，要清除nacos中对应的服务信息。由于不知道服务器什么什么时候关闭，所以需要钩子函数。执行在某些事件之后的函数就叫钩子函数，把注销服务的方法写入系统关闭的钩子函数就可以起到善后的工作。实现nacos的一个工具类：NacosUtil,今后nacos服务注册和服务发现都要借助这个工具类中，所有的服务名都会被存在NacosUtil的成员变量serviceNames中，当RpcServer关闭的时候，遍历这个set，调用 namingService.deregisterInstance，清除nacos中所有的服务

public class NacosUtil {

    private static final Logger logger = LoggerFactory.getLogger(NacosUtil.class);

    private static final NamingService namingService;
    private static final Set<String> serviceNames = new HashSet<>();
    private static InetSocketAddress address;

    private static final String SERVER_ADDR = "127.0.0.1:8848";

    static {
        namingService = getNacosNamingService();
    }

    public static NamingService getNacosNamingService() {
        try {
            return NamingFactory.createNamingService(SERVER_ADDR);
        } catch (NacosException e) {
            logger.error("连接到Nacos时有错误发生: ", e);
            throw new RpcException(RpcError.FAILED_TO_CONNECT_TO_SERVICE_REGISTRY);
        }
    }

    public static void registerService(String serviceName, InetSocketAddress address) throws NacosException {
        namingService.registerInstance(serviceName, address.getHostName(), address.getPort());
        NacosUtil.address = address;
        serviceNames.add(serviceName);

    }

    public static List<Instance> getAllInstance(String serviceName) throws NacosException {
        return namingService.getAllInstances(serviceName);
    }

    public static void clearRegistry() {
        if(!serviceNames.isEmpty() && address != null) {
            String host = address.getHostName();
            int port = address.getPort();
            Iterator<String> iterator = serviceNames.iterator();
            while(iterator.hasNext()) {
                String serviceName = iterator.next();
                try {
                    namingService.deregisterInstance(serviceName, host, port);
                } catch (NacosException e) {
                    logger.error("注销服务 {} 失败", serviceName, e);
                }
            }
        }
    }
}

接下来就是钩子，钩子在RpcServer关闭的时候调用，在 addClearAllHook 中，Runtime 对象是 JVM 虚拟机的运行时环境，调用其 addShutdownHook 方法增加一个钩子函数，创建一个新线程调用 clearRegistry 方法完成注销工作。这个钩子函数会在 JVM 关闭之前被调用。

public class ShutdownHook {

    private static final Logger logger = LoggerFactory.getLogger(ShutdownHook.class);

    private static final ShutdownHook shutdownHook = new ShutdownHook();

    public static ShutdownHook getShutdownHook() {
        return shutdownHook;
    }

    public void addClearAllHook() {
        logger.info("关闭后将自动注销所有服务");
        Runtime.getRuntime().addShutdownHook(new Thread(() -> {
            NacosUtil.clearRegistry();
            ThreadPoolFactory.shutDownAll();
        }));
    }

}

在Rpcserver启动的时候注册这个钩子：

@Override
public void start() {
    ShutdownHook.getShutdownHook().addClearAllHook();
    EventLoopGroup bossGroup = new NioEventLoopGroup();
    EventLoopGroup workerGroup = new NioEventLoopGroup();
    try {

        ServerBootstrap serverBootstrap = new ServerBootstrap();
        serverBootstrap.group(bossGroup, workerGroup)
                .channel(NioServerSocketChannel.class)
                .handler(new LoggingHandler(LogLevel.INFO))
                .option(ChannelOption.SO_BACKLOG, 256)
                .option(ChannelOption.SO_KEEPALIVE, true)
                .childOption(ChannelOption.TCP_NODELAY, true)
                .childHandler(new ChannelInitializer<SocketChannel>() {
                    @Override
                    protected void initChannel(SocketChannel ch) throws Exception {
                        ChannelPipeline pipeline = ch.pipeline();
                        pipeline.addLast(new IdleStateHandler(30, 0, 0, TimeUnit.SECONDS))
                                .addLast(new CommonEncoder(serializer))
                                .addLast(new CommonDecoder())
                                .addLast(new NettyServerHandler());
                    }
                });
        ChannelFuture future = serverBootstrap.bind(host, port).sync();
        future.channel().closeFuture().sync();

    } catch (InterruptedException e) {
        logger.error("启动服务器时有错误发生: ", e);
    } finally {
        bossGroup.shutdownGracefully();
        workerGroup.shutdownGracefully();
    }
}

负载均衡则是 则是Rpcclient从nacos中获取所有的服务列表的时候选择服务的标准，最简单的负载均衡就是轮询，定义负载均衡算法的接口：

public interface LoadBalancer {

    Instance select(List<Instance> instances);

}

轮询实现：

public class RoundRobinLoadBalancer implements LoadBalancer {

    private int index = 0;

    @Override
    public Instance select(List<Instance> instances) {
        if(index >= instances.size()) {
            index %= instances.size();
        }
        return instances.get(index++);
    }

}

 

集成到服务发现类中：

public class NacosServiceDiscovery implements ServiceDiscovery {

    private static final Logger logger = LoggerFactory.getLogger(NacosServiceDiscovery.class);

    private final LoadBalancer loadBalancer;

    public NacosServiceDiscovery(LoadBalancer loadBalancer) {
        if(loadBalancer == null) this.loadBalancer = new RandomLoadBalancer();
        else this.loadBalancer = loadBalancer;
    }

    @Override
    public InetSocketAddress lookupService(String serviceName) {
        try {
            List<Instance> instances = NacosUtil.getAllInstance(serviceName);
            Instance instance = loadBalancer.select(instances);
            return new InetSocketAddress(instance.getIp(), instance.getPort());
        } catch (NacosException e) {
            logger.error("获取服务时有错误发生:", e);
        }
        return null;
    }

}