使用 Netty、ZooKeeper 和 Spring Boot 手撸一个微服务框架。
详细信息可参考:RPC 实战与原理
项目可以分为调用方(client)和提供方(server),client 端只需要调用接口即可,最终调用信息会通过网络传输到 server,server 通过解码后反射调用对应的方法,并将结果通过网络返回给 client。对于 client 端可以完全忽略网络的存在,就像调用本地方法一样调用 rpc 服务。
整个项目的 model 结构如下:
启动 Docker,并下载 ZooKeeper 镜像。详见 https://hub.docker.com/_/zookeeper
启动命令如下,容器的名称是zookeeper-rpc-demo
,同时向本机暴露 8080、2181、2888 和 3888 端口:
docker run --name zookeeper-rpc-demo --restart always -p 8080:8080 -p 2181:2181 -p 2888:2888 -p 3888:3888 -d zookeeper
This image includes EXPOSE 2181 2888 3888 8080 (the zookeeper client port, follower port, election port, AdminServer port respectively), so standard container linking will make it automatically available to the linked containers. Since the Zookeeper "fails fast" it's better to always restart it.
可以通过下面的命令进入容器,其中fb6f95cde6ba
是我本机的 Docker ZooKeeper 容器 id。
docker exec -it fb6f95cde6ba /bin/bash
在容器中进入目录:/apache-zookeeper-3.7.0-bin/bin,执行命令 zkCli.sh -server 0.0.0.0:2181
链接 zk 服务。
本示例提供了两个接口:HelloService 和 HiService,里面分别有一个接口方法,客户端仅需引用 rpc-sample-api,只知道接口定义,并不知道里面的具体实现。
public interface HelloService {
String hello(String msg);
}
public interface HiService {
String hi(String msg);
}
启动一个 Server 服务,实现上面的两个 RPC 接口,并向 ZooKeeper 进行服务注册。
/**
* @author yano
* GitHub 项目: https://github.com/LjyYano/Thinking_in_Java_MindMapping
* @date 2021-05-07
*/
@RpcServer(cls = HelloService.class)
public class HelloServiceImpl implements HelloService {
@Override
public String hello(String msg) {
return "hello echo: " + msg;
}
}
/**
* @author yano
* GitHub 项目: https://github.com/LjyYano/Thinking_in_Java_MindMapping
* @date 2021-05-07
*/
@RpcServer(cls = HiService.class)
public class HiServiceImpl implements HiService {
public String hi(String msg) {
return "hi echo: " + msg;
}
}
这里涉及到了两个问题:
本示例 Server 使用 Spring Boot,但是我们并不需要启动一个 Web 服务,只需要保持后台运行就可以,所以将 web 设置成 WebApplicationType.NONE
@SpringBootApplication
public class RpcServerApplication {
public static void main(String[] args) {
new SpringApplicationBuilder(RpcServerApplication.class)
.web(WebApplicationType.NONE)
.run(args);
}
}
NettyApplicationContextAware 是一个 ApplicationContextAware 的实现类,程序在启动时,将带有 RpcServer(下面会讲解)注解的实现类注册到 ZooKeeper 上。
@Component
public class NettyApplicationContextAware implements ApplicationContextAware {
private static final Logger logger = LoggerFactory.getLogger(NettyApplicationContextAware.class);
@Value("${zk.address}")
private String zookeeperAddress;
@Value("${zk.port}")
private int zookeeperPort;
public void setApplicationContext(ApplicationContext applicationContext) throws BeansException {
Map<String, Object> rpcBeanMap = new HashMap<>();
for (Object object : applicationContext.getBeansWithAnnotation(RpcServer.class).values()) {
rpcBeanMap.put("/" + object.getClass().getAnnotation(RpcServer.class).cls().getName(), object);
}
try {
NettyServer.start(zookeeperAddress, zookeeperPort, rpcBeanMap);
} catch (Exception e) {
logger.error("register error !", e);
}
}
}
RpcServer 注解的定义如下:
@Retention(RetentionPolicy.RUNTIME)
@Documented
@Target(ElementType.TYPE)
@Component
public @interface RpcServer {
/**
* 接口类,用以接口注册
*/
Class<?> cls();
}
applicationContext.getBeansWithAnnotation(RpcServer.class).values()
就是获取项目中带有 RpcServer 注解的类,并将其放入一个 rpcBeanMap 中,其中 key 就是待注册到 ZooKeeper 中的路径。注意路径使用接口的名字,而不是类的名字。
使用注解的好处是,Server A 可以仅提供 HelloService,Server B 仅提供 HiService,不会相互影响且更加灵活。
服务注册主要在 com.yano.server.NettyServer#start 中。
public class NettyServer {
private static final Logger logger = LoggerFactory.getLogger(NettyServer.class);
public static void start(String ip, int port, Map<String, Object> params) throws Exception {
EventLoopGroup bossGroup = new NioEventLoopGroup();
EventLoopGroup workerGroup = new NioEventLoopGroup();
try {
ServerBootstrap serverBootstrap = new ServerBootstrap();
serverBootstrap.group(bossGroup, workerGroup)
.channel(NioServerSocketChannel.class)
.option(ChannelOption.SO_BACKLOG, 1024)
.option(ChannelOption.SO_KEEPALIVE, true)
.childHandler(new ChannelInitializer<SocketChannel>() {
protected void initChannel(SocketChannel socketChannel) {
socketChannel.pipeline()
.addLast(new RpcDecoder(Request.class))
.addLast(new RpcEncoder(Response.class))
.addLast(new RpcServerInboundHandler(params));
}
});
ChannelFuture future = serverBootstrap.bind(ip, port).sync();
if (future.isSuccess()) {
params.keySet().forEach(key -> ZooKeeperOp.register(key, ip + ":" + port));
}
future.channel().closeFuture().sync();
} finally {
bossGroup.shutdownGracefully();
workerGroup.shutdownGracefully();
}
}
}
这个类的作用是:
params.keySet().forEach(key -> ZooKeeperOp.register(key, ip + ":" + port));
主要就是维护 zk 连接,并将 Server 的 ip 和 port 注册到对应的 ZooKeeper 中。这里使用 Ephemeral node,这样 Server 在下线丢失连接之后,ZooKeeper 能够自动删除节点,这样 Client 就不会获取到下线的 Server 地址了。
public class ZooKeeperOp {
private static final String zkAddress = "localhost:2181";
private static final ZkClient zkClient = new ZkClient(zkAddress);
public static void register(String serviceName, String serviceAddress) {
if (!zkClient.exists(serviceName)) {
zkClient.createPersistent(serviceName);
}
zkClient.createEphemeral(serviceName + "/" + serviceAddress);
System.out.printf("create node %s \n", serviceName + "/" + serviceAddress);
}
public static String discover(String serviceName) {
List<String> children = zkClient.getChildren(serviceName);
if (CollectionUtils.isEmpty(children)) {
return "";
}
return children.get(ThreadLocalRandom.current().nextInt(children.size()));
}
}
Netty RPC Client 主要是像调用本地方法一样调用上述的两个接口,验证能够正常返回即可。
public class RpcClientApplication {
public static void main(String[] args) {
HiService hiService = RpcProxy.create(HiService.class);
String msg = hiService.hi("msg");
System.out.println(msg);
HelloService helloService = RpcProxy.create(HelloService.class);
msg = helloService.hello("msg");
System.out.println(msg);
}
}
运行上述代码,最终控制台会输出:
hi echo: msg
hello echo: msg
HiService hiService = RpcProxy.create(HiService.class);
String msg = hiService.hi("msg");
Client 需要通过 com.yano.RpcProxy#create 创建代理,之后就可以调用 hiService 的 hi 方法了。
public class RpcProxy {
public static <T> T create(final Class<?> cls) {
return (T) Proxy.newProxyInstance(cls.getClassLoader(), new Class<?>[] {cls}, (o, method, objects) -> {
Request request = new Request();
request.setInterfaceName("/" + cls.getName());
request.setRequestId(UUID.randomUUID().toString());
request.setParameter(objects);
request.setMethodName(method.getName());
request.setParameterTypes(method.getParameterTypes());
Response response = new NettyClient().client(request);
return response.getResult();
});
}
}
Server 端要想能够通过反射调用 Client 端请求的方法,至少需要:
@Data
public class Request {
private String requestId;
private String interfaceName;
private String methodName;
private Class<?>[] parameterTypes;
private Object parameter[];
}
最终是通过下面这段代码远程调用的,其中 request 包含了调用方法的所有信息。
Response response = new NettyClient().client(request);
/**
* @author yano
* GitHub 项目: https://github.com/LjyYano/Thinking_in_Java_MindMapping
* @date 2021-05-07
*/
public class NettyClient extends SimpleChannelInboundHandler<Response> {
private Response response;
@Override
public void exceptionCaught(ChannelHandlerContext ctx, Throwable cause) {
ctx.close();
}
@Override
protected void channelRead0(ChannelHandlerContext channelHandlerContext, Response response) {
this.response = response;
}
public Response client(Request request) throws Exception {
EventLoopGroup group = new NioEventLoopGroup();
try {
// 创建并初始化 Netty 客户端 Bootstrap 对象
Bootstrap bootstrap = new Bootstrap();
bootstrap.group(group)
.channel(NioSocketChannel.class)
.option(ChannelOption.TCP_NODELAY, true)
.handler(new ChannelInitializer<SocketChannel>() {
@Override
public void initChannel(SocketChannel channel) {
channel.pipeline()
.addLast(new RpcDecoder(Response.class))
.addLast(new RpcEncoder(Request.class))
.addLast(NettyClient.this);
}
});
// 连接 RPC 服务器
String[] discover = ZooKeeperOp.discover(request.getInterfaceName()).split(":");
ChannelFuture future = bootstrap.connect(discover[0], Integer.parseInt(discover[1])).sync();
// 写入 RPC 请求数据并关闭连接
Channel channel = future.channel();
channel.writeAndFlush(request).sync();
channel.closeFuture().sync();
return response;
} finally {
group.shutdownGracefully();
}
}
}
这段代码是核心,主要做了两件事:
channel.pipeline()
.addLast(new RpcDecoder(Response.class))
.addLast(new RpcEncoder(Request.class))
.addLast(NettyClient.this);
Client 和 Server 都需要对 Request、Response 编解码。本示例采用了最简单的 Json 格式。Netty 的消息编解码具体不详细讲解,具体代码如下。
RpcDecoder 是一个 ChannelInboundHandler,在 Client 端是对 Response 解码。
public class RpcDecoder extends MessageToMessageDecoder<ByteBuf> {
private final Class<?> genericClass;
public RpcDecoder(Class<?> genericClass) {
this.genericClass = genericClass;
}
@Override
public void decode(ChannelHandlerContext ctx, ByteBuf msg, List<Object> out) {
if (msg.readableBytes() < 4) {
return;
}
msg.markReaderIndex();
int dataLength = msg.readInt();
if (msg.readableBytes() < dataLength) {
msg.resetReaderIndex();
return;
}
byte[] data = new byte[dataLength];
msg.readBytes(data);
out.add(JSON.parseObject(data, genericClass));
}
}
RpcEncoder 是一个 ChannelOutboundHandler,在 Client 端是对 Request 编码。
public class RpcEncoder extends MessageToByteEncoder {
private final Class<?> genericClass;
public RpcEncoder(Class<?> genericClass) {
this.genericClass = genericClass;
}
@Override
public void encode(ChannelHandlerContext ctx, Object msg, ByteBuf out) {
if (genericClass.isInstance(msg)) {
byte[] data = JSON.toJSONBytes(msg);
out.writeInt(data.length);
out.writeBytes(data);
}
}
}
这个是 Server 反射调用的核心,这里单独拿出来讲解。Netty Server 在启动时,已经在 pipeline 中加入了 RpcServerInboundHandler。
socketChannel.pipeline()
.addLast(new RpcDecoder(Request.class))
.addLast(new RpcEncoder(Response.class))
.addLast(new RpcServerInboundHandler(params));
@Override
public void channelRead(ChannelHandlerContext ctx, Object msg) throws Exception {
Request request = (Request) msg;
logger.info("request data {}", JSON.toJSONString(request));
// jdk 反射调用
Object bean = handle.get(request.getInterfaceName());
Method method = bean.getClass().getMethod(request.getMethodName(), request.getParameterTypes());
method.setAccessible(true);
Object result = method.invoke(bean, request.getParameter());
Response response = new Response();
response.setRequestId(request.getRequestId());
response.setResult(result);
// client 接收到信息后主动关闭掉连接
ctx.writeAndFlush(response).addListener(ChannelFutureListener.CLOSE);
}
Server 启动后的输出如下:
其中有 2 行 log:
create node /com.yano.service.HelloService/127.0.0.1:3000
create node /com.yano.service.HiService/127.0.0.1:3000
在 ZooKeeper 中查看节点,发现服务已经注册上去了。
[zk: 0.0.0.0:2181(CONNECTED) 0] ls /com.yano.service.HelloService
[127.0.0.1:3000]
[zk: 0.0.0.0:2181(CONNECTED) 1] ls /com.yano.service.HiService
[127.0.0.1:3000]
使用 Netty、ZooKeeper 和 Spring Boot 手撸一个微服务 RPC 框架。这个 demo 只能作为一个实例,手动实现能加深理解,切勿在生产环境使用。
本文代码均可在 GitHub 源码地址 中找到,欢迎大家 star 和 fork。
原创声明:本文系作者授权腾讯云开发者社区发表,未经许可,不得转载。
如有侵权,请联系 cloudcommunity@tencent.com 删除。
原创声明:本文系作者授权腾讯云开发者社区发表,未经许可,不得转载。
如有侵权,请联系 cloudcommunity@tencent.com 删除。