【探索SpringCloud】服务发现-Nacos服务端数据结构和模型

前言

上一文中,我们从官方的图示了解到Nacos的服务数据结构。但我关心的是,Nacos2.x不是重构了吗?怎么还是这种数据结构?我推测,必然是为了对Nacos1.x的兼容,实际存储应该不是这样的。于是,沿着这个问题出发我们一起来翻一下源码。

从NamingService的使用开始

在扎入源码之前,我们需要回忆一下,我们是怎么使用Nacos的?

  1. 构建NamingService
    NamingService serviceRegistry = NacosFactory.createNamingService(properties);
    实际上,这个动作的背后,意味着我们连接了Nacos服务端。
  2. 注册服务
    serviceRegistry.registerInstance(serviceName, groupName, instance);
  3. 查询服务
    serviceRegistry.getAllInstances(serviceName, groupName, List.of(clusterName));
    因此,我们就沿着这几个操作,摸一摸源码。

!!!高能警告!!!

没有耐心看源码的同学,可以直接翻到总结,直接看结论。

构建NamingService

客户端

java 复制代码
    // com.alibaba.nacos.client.naming.NacosNamingService
    /**
     * 初始化方法
     * <p>由NacosNamingService构造器调用,用于初始NamingService</p>
     */
    private void init(Properties properties) throws NacosException {
        final NacosClientProperties nacosClientProperties = NacosClientProperties.PROTOTYPE.derive(properties);
        // 省略...
        // 创建客户端
        this.clientProxy = new NamingClientProxyDelegate(this.namespace, serviceInfoHolder, nacosClientProperties, changeNotifier);
    }

    // com.alibaba.nacos.client.naming.remote.NamingClientProxyDelegate
    /**
     * NamingClientProxyDelegate构造器
     */    
    public NamingClientProxyDelegate(String namespace, ServiceInfoHolder serviceInfoHolder, NacosClientProperties properties,
        InstancesChangeNotifier changeNotifier) throws NacosException {
        // 省略...
        // 初始化了两个客户端,一个是Http,另一个是Grpc。不过,在注册实例时,如果该实例为临时实例,则使用Grpc,因此我们重点关注Grpc
        this.httpClientProxy = new NamingHttpClientProxy(namespace, securityProxy, serverListManager, properties);
        this.grpcClientProxy = new NamingGrpcClientProxy(namespace, securityProxy, serverListManager, properties,
                serviceInfoHolder);
    }
    
    // com.alibaba.nacos.client.naming.remote.gprc.NamingGrpcClientProxy
    /**
     * NamingGrpcClientProxy构造器
     */    
    public NamingGrpcClientProxy(String namespaceId, SecurityProxy securityProxy, ServerListFactory serverListFactory,
            NacosClientProperties properties, ServiceInfoHolder serviceInfoHolder) throws NacosException {
        // 省略...
        // 创建RPC客户端
        this.rpcClient = RpcClientFactory.createClient(uuid, ConnectionType.GRPC, labels);
        this.redoService = new NamingGrpcRedoService(this);
        // 启动客户端
        start(serverListFactory, serviceInfoHolder);
    }
    // com.alibaba.nacos.client.naming.remote.gprc.NamingGrpcClientProxy
    private void start(ServerListFactory serverListFactory, ServiceInfoHolder serviceInfoHolder) throws NacosException {
        rpcClient.serverListFactory(serverListFactory);
        rpcClient.registerConnectionListener(redoService);
        rpcClient.registerServerRequestHandler(new NamingPushRequestHandler(serviceInfoHolder));
        // 启动客户端
        rpcClient.start();
        NotifyCenter.registerSubscriber(this);
    }
    
    // com.alibaba.nacos.common.remote.client.RpcClient#start
    /**
     * 启动客户端
     */
    public final void start() throws NacosException {
        // 控制只启动一次
        boolean success = rpcClientStatus.compareAndSet(RpcClientStatus.INITIALIZED, RpcClientStatus.STARTING);
        if (!success) {
            return;
        }
        // 创建一个只有2个线程的定时任务线程池
        clientEventExecutor = new ScheduledThreadPoolExecutor(2, r -> {
            Thread t = new Thread(r);
            t.setName("com.alibaba.nacos.client.remote.worker");
            t.setDaemon(true);
            return t;
        });
        
        // 提交-处理连接事件的Task
        clientEventExecutor.submit(() -> {
            while (!clientEventExecutor.isTerminated() && !clientEventExecutor.isShutdown()) {
                ConnectionEvent take;
                take = eventLinkedBlockingQueue.take();
                if (take.isConnected()) {
                    notifyConnected();
                } else if (take.isDisConnected()) {
                    notifyDisConnected();
                }
            }
        });
        // 提交-心跳任务
        clientEventExecutor.submit(() -> {
            while (true) {
            // 由于这里有一大堆逻辑,省略。
            // 1. 超过时间间隔,发起心跳请求
            // 1.1 心跳请求失败,记录当前状态为不健康,并记录上下文。
            // 1.2 检查当前配置的推荐的Nacos服务器是否在服务器列表中。在,则尝试重新连接推荐的服务器。
        });
        
        // connect to server, try to connect to server sync retryTimes times, async starting if failed.
        // 连接服务端,尝试retryTimes次,同步地连接服务端,如果依然失败,则改为异步连接。
        Connection connectToServer = null;
        rpcClientStatus.set(RpcClientStatus.STARTING);
        
        int startUpRetryTimes = rpcClientConfig.retryTimes();
        while (startUpRetryTimes > 0 && connectToServer == null) {
            try {
                startUpRetryTimes--;
                ServerInfo serverInfo = nextRpcServer();
                // 连接服务器
                connectToServer = connectToServer(serverInfo);
            } catch (Throwable e) {
                LoggerUtils.printIfWarnEnabled(LOGGER,
                        "[{}] Fail to connect to server on start up, error message = {}, start up retry times left: {}",
                        rpcClientConfig.name(), e.getMessage(), startUpRetryTimes, e);
            }
            
        }
        
        if (connectToServer != null) {
            this.currentConnection = connectToServer;
            rpcClientStatus.set(RpcClientStatus.RUNNING);
            eventLinkedBlockingQueue.offer(new ConnectionEvent(ConnectionEvent.CONNECTED));
        } else {
            switchServerAsync();
        }
        
        registerServerRequestHandler(new ConnectResetRequestHandler());
        
        // register client detection request.
        // 注册客户端检测请求处理器,用于响应服务端的探测
        registerServerRequestHandler(request -> {
            if (request instanceof ClientDetectionRequest) {
                return new ClientDetectionResponse();
            }
            
            return null;
        });
        
    }

服务端-处理连接请求

服务端的源码首先我们得找到GrpcServer

java 复制代码
    @Override
    public void startServer() throws Exception {
    	// 1. 创建请求处理器注册器
        final MutableHandlerRegistry handlerRegistry = new MutableHandlerRegistry();
        // 2. 注册请求处理器,并封装拦截器器。封装后,有点类似于SpringMVC的HandlerAdapter
        addServices(handlerRegistry, new GrpcConnectionInterceptor());
        NettyServerBuilder builder = NettyServerBuilder.forPort(getServicePort()).executor(getRpcExecutor());
		// 省略
		
        server = builder.maxInboundMessageSize(getMaxInboundMessageSize()).fallbackHandlerRegistry(handlerRegistry)
                .compressorRegistry(CompressorRegistry.getDefaultInstance())
                .decompressorRegistry(DecompressorRegistry.getDefaultInstance())
                .addTransportFilter(new AddressTransportFilter(connectionManager))
                .keepAliveTime(getKeepAliveTime(), TimeUnit.MILLISECONDS)
                .keepAliveTimeout(getKeepAliveTimeout(), TimeUnit.MILLISECONDS)
                .permitKeepAliveTime(getPermitKeepAliveTime(), TimeUnit.MILLISECONDS)
                .build();
		// 启动服务
        server.start();
    }


    private void addServices(MutableHandlerRegistry handlerRegistry, ServerInterceptor... serverInterceptor) {

        // unary common call register.
        // 通用调用注册
        final MethodDescriptor<Payload, Payload> unaryPayloadMethod = MethodDescriptor.<Payload, Payload>newBuilder()
                .setType(MethodDescriptor.MethodType.UNARY)
                .setFullMethodName(MethodDescriptor.generateFullMethodName(GrpcServerConstants.REQUEST_SERVICE_NAME,
                        GrpcServerConstants.REQUEST_METHOD_NAME))
                .setRequestMarshaller(ProtoUtils.marshaller(Payload.getDefaultInstance()))
                .setResponseMarshaller(ProtoUtils.marshaller(Payload.getDefaultInstance())).build();
		// 定义服务器调用处理器。核心处理逻辑可就在这lambda表达式定义的匿名内部类里了。也只有一个方法:
		// grpcCommonRequestAcceptor.request(request, responseObserver)
        final ServerCallHandler<Payload, Payload> payloadHandler = ServerCalls
                .asyncUnaryCall((request, responseObserver) -> grpcCommonRequestAcceptor.request(request, responseObserver));

        final ServerServiceDefinition serviceDefOfUnaryPayload = ServerServiceDefinition.builder(
                        GrpcServerConstants.REQUEST_SERVICE_NAME)
                .addMethod(unaryPayloadMethod, payloadHandler).build();
        handlerRegistry.addService(ServerInterceptors.intercept(serviceDefOfUnaryPayload, serverInterceptor));

        // bi stream register.
        // bi流式调用服务,主要是连接请求、连接断开
        // 核心处理逻辑:
		// grpcBiStreamRequestAcceptor.requestBiStream(responseObserver)
        final ServerCallHandler<Payload, Payload> biStreamHandler = ServerCalls.asyncBidiStreamingCall(
                (responseObserver) -> grpcBiStreamRequestAcceptor.requestBiStream(responseObserver));

        final MethodDescriptor<Payload, Payload> biStreamMethod = MethodDescriptor.<Payload, Payload>newBuilder()
                .setType(MethodDescriptor.MethodType.BIDI_STREAMING).setFullMethodName(MethodDescriptor
                        .generateFullMethodName(GrpcServerConstants.REQUEST_BI_STREAM_SERVICE_NAME,
                                GrpcServerConstants.REQUEST_BI_STREAM_METHOD_NAME))
                .setRequestMarshaller(ProtoUtils.marshaller(Payload.newBuilder().build()))
                .setResponseMarshaller(ProtoUtils.marshaller(Payload.getDefaultInstance())).build();

        final ServerServiceDefinition serviceDefOfBiStream = ServerServiceDefinition
                .builder(GrpcServerConstants.REQUEST_BI_STREAM_SERVICE_NAME).addMethod(biStreamMethod, biStreamHandler).build();
        handlerRegistry.addService(ServerInterceptors.intercept(serviceDefOfBiStream, serverInterceptor));

    }

处理连接请求:

java 复制代码
	// com.alibaba.nacos.core.remote.grpc.GrpcBiStreamRequestAcceptor
    @Override
    public StreamObserver<Payload> requestBiStream(StreamObserver<Payload> responseObserver) {
        
        StreamObserver<Payload> streamObserver = new StreamObserver<Payload>() {
            
            final String connectionId = GrpcServerConstants.CONTEXT_KEY_CONN_ID.get();
            
            final Integer localPort = GrpcServerConstants.CONTEXT_KEY_CONN_LOCAL_PORT.get();
            
            final int remotePort = GrpcServerConstants.CONTEXT_KEY_CONN_REMOTE_PORT.get();
            
            String remoteIp = GrpcServerConstants.CONTEXT_KEY_CONN_REMOTE_IP.get();
            
            String clientIp = "";
            
            @Override
            public void onNext(Payload payload) {
                // 处理连接请求
                clientIp = payload.getMetadata().getClientIp();
                traceDetailIfNecessary(payload);
                
                Object parseObj;
                // 省略...
				// 检查
				
                if (parseObj instanceof ConnectionSetupRequest) {
                    ConnectionSetupRequest setUpRequest = (ConnectionSetupRequest) parseObj;
                    // 设置label,省略
                    
                    // 构建Connection
                    ConnectionMeta metaInfo = new ConnectionMeta(connectionId, payload.getMetadata().getClientIp(),
                            remoteIp, remotePort, localPort, ConnectionType.GRPC.getType(),
                            setUpRequest.getClientVersion(), appName, setUpRequest.getLabels());
                    metaInfo.setTenant(setUpRequest.getTenant());
                    // 第三个参数Channel,是发生网路数据的关键
                    Connection connection = new GrpcConnection(metaInfo, responseObserver, GrpcServerConstants.CONTEXT_KEY_CHANNEL.get());
                    connection.setAbilities(setUpRequest.getAbilities());
                    boolean rejectSdkOnStarting = metaInfo.isSdkSource() && !ApplicationUtils.isStarted();
                    // 注册连接, 重点在 "或" 条件上
                    // connectionManager.register
                    if (rejectSdkOnStarting || !connectionManager.register(connectionId, connection)) {
                        //Not register to the connection manager if current server is over limit or server is starting.
                        // 如果当前服务器已超限制,或者服务器还在启动过程中,则注册失败。
                            connection.request(new ConnectResetRequest(), 3000L);
                            connection.close();
                    }
                }            
            	// 省略。。。
            }
            // 省略。。。
        };
        
        return streamObserver;
    }

这里出现了我们接触到的第一个概念:Connection-连接,他有个属性ConnectionMeta,记录连接相关的信息。当需要发起请求时,他会将这些信息设置到Request中,然后通过GrpcUtils转换成Payload发出请求

继续看com.alibaba.nacos.core.remote.ConnectionManager#register

java 复制代码
    public synchronized boolean register(String connectionId, Connection connection) {
        
        if (connection.isConnected()) {
            String clientIp = connection.getMetaInfo().clientIp;
            // 省略入参检查
            // 注册客户端
            connections.put(connectionId, connection);
            // 登记客户端IP
            if (!connectionForClientIp.containsKey(clientIp)) {
                connectionForClientIp.put(clientIp, new AtomicInteger(0));
            }
            connectionForClientIp.get(clientIp).getAndIncrement();
            
            // 通知客户端连接Listener
            clientConnectionEventListenerRegistry.notifyClientConnected(connection);
            
            return true;
            
        }
        return false;
        
    }

此处出现第一个Manager:ConnectionManager。用来管理所有客户端的连接。登记连接后,调用了所有的Listener的clientConnected方法。其中,有个ConnectionBasedClientManager,看名字就知道,可能是负责管理客户端的。

java 复制代码
	// > ConnectionBasedClientManager#clientConnected(com.alibaba.nacos.core.remote.Connection)
	//  > ConnectionBasedClientManager#clientConnected(java.lang.String, com.alibaba.nacos.naming.core.v2.client.ClientAttributes)
	// ConnectionBasedClientManager
	@Override
    public boolean clientConnected(String clientId, ClientAttributes attributes) {
        String type = attributes.getClientAttribute(ClientConstants.CONNECTION_TYPE);
        ClientFactory clientFactory = ClientFactoryHolder.getInstance().findClientFactory(type);
        // 通过ClientFactory创建客户端
        // 从以上的两行代码,我们通过ClientConstants.CONNECTION_TYPE就知道工厂是ConnectionBasedClientFactory,对应的客户端自然是ConnectionBasedClient
        return clientConnected(clientFactory.newClient(clientId, attributes));
    }
    @Override
    public boolean clientConnected(final Client client) {
    	// 登记客户端
        clients.computeIfAbsent(client.getClientId(), s -> {
            return (ConnectionBasedClient) client;
        });
        return true;
    }

至此,我们又发现一个新概念:Client-客户端。由Grpc连接的客户端,都由ConnectionBasedClientManager进行管理。

小结

概念 管理者
连接 com.alibaba.nacos.core.remote.Connection ConnectionManager
客户端 com.alibaba.nacos.naming.core.v2.client.Client ClientManager

注册实例

客户端

我们重点看看

java 复制代码
    public void doRegisterService(String serviceName, String groupName, Instance instance) throws NacosException {
    	// 创建请求。每个Request在Nacos服务端都由对应的Handler
        InstanceRequest request = new InstanceRequest(namespaceId, serviceName, groupName,
                NamingRemoteConstants.REGISTER_INSTANCE, instance);
        requestToServer(request, Response.class);
        redoService.instanceRegistered(serviceName, groupName);
    }

服务端

我们前面说服务端启动时,说这个是负责处理通用请求的:

final ServerCallHandler<Payload, Payload> payloadHandler = ServerCalls.asyncUnaryCall((request, responseObserver) -> grpcCommonRequestAcceptor.request(request, responseObserver));

我们就顺着往下看

java 复制代码
	// com.alibaba.nacos.core.remote.grpc.GrpcRequestAcceptor#request
    @Override
    public void request(Payload grpcRequest, StreamObserver<Payload> responseObserver) {
        
        String type = grpcRequest.getMetadata().getType();
        // 省略如下内容:
        // 检查服务是否已启动
        // 如果是客户端对服务端的健康检查,则直接响应
        // ----------------------------
        // 从对应的请求处理器
        RequestHandler requestHandler = requestHandlerRegistry.getByRequestType(type);
        // 省略:no handler found. 的异常处理
        // ----------------------------
        String connectionId = GrpcServerConstants.CONTEXT_KEY_CONN_ID.get();
        
        // 省略:检查连接是否正常.
        Object parseObj = null;
        parseObj = GrpcUtils.parse(grpcRequest);
		// 省略:转换异常、无效请求异常
        
        Request request = (Request) parseObj;
			// 从ConnectionManager获取到对应的Connection
            Connection connection = connectionManager.getConnection(GrpcServerConstants.CONTEXT_KEY_CONN_ID.get());
            // 组装RequestMeta
            RequestMeta requestMeta = new RequestMeta();
            requestMeta.setClientIp(connection.getMetaInfo().getClientIp());
            requestMeta.setConnectionId(GrpcServerConstants.CONTEXT_KEY_CONN_ID.get());
            requestMeta.setClientVersion(connection.getMetaInfo().getVersion());
            requestMeta.setLabels(connection.getMetaInfo().getLabels());
            connectionManager.refreshActiveTime(requestMeta.getConnectionId());
            // 调用requestHandler处理请求
            Response response = requestHandler.handleRequest(request, requestMeta);
            Payload payloadResponse = GrpcUtils.convert(response);
            traceIfNecessary(payloadResponse, false);
            responseObserver.onNext(payloadResponse);
            responseObserver.onCompleted();
    }

这些便是通用请求处理的核心逻辑。现在我们便来看InstanceRequest的处理com.alibaba.nacos.naming.remote.rpc.handler.InstanceRequestHandler

java 复制代码
    @Override
    @Secured(action = ActionTypes.WRITE)
    public InstanceResponse handle(InstanceRequest request, RequestMeta meta) throws NacosException {
        Service service = Service
                .newService(request.getNamespace(), request.getGroupName(), request.getServiceName(), true);
        switch (request.getType()) {
            case NamingRemoteConstants.REGISTER_INSTANCE:
                return registerInstance(service, request, meta);
            case NamingRemoteConstants.DE_REGISTER_INSTANCE:
                return deregisterInstance(service, request, meta);
            default:
                throw new NacosException
        }
    }
    private InstanceResponse registerInstance(Service service, InstanceRequest request, RequestMeta meta)
            throws NacosException {
        // 1. 注册实例
        clientOperationService.registerInstance(service, request.getInstance(), meta.getConnectionId());
        // 2. 发布事件:RegisterInstanceTraceEvent
        NotifyCenter.publishEvent(new RegisterInstanceTraceEvent(System.currentTimeMillis(),
                meta.getClientIp(), true, service.getNamespace(), service.getGroup(), service.getName(),
                request.getInstance().getIp(), request.getInstance().getPort()));
        return new InstanceResponse(NamingRemoteConstants.REGISTER_INSTANCE);
    }

// 注册实例:
// com.alibaba.nacos.naming.core.v2.service.impl.EphemeralClientOperationServiceImpl#registerInstance

    @Override
    public void registerInstance(Service service, Instance instance, String clientId) throws NacosException {
        NamingUtils.checkInstanceIsLegal(instance);
    	// 从ServiceManager获取已注册服务。而我们当前是要注册实例,所以,这个方法肯定还内含玄机
        Service singleton = ServiceManager.getInstance().getSingleton(service);
        // 省略:如果获取到的是持久化实例,意味着当前注册临时实例冲突了,返回异常。
        Client client = clientManager.getClient(clientId);

        InstancePublishInfo instanceInfo = getPublishInfo(instance);
        // 记录当前客户端发布的实例
        client.addServiceInstance(singleton, instanceInfo);
        client.setLastUpdatedTime();
        client.recalculateRevision();
        // 发布服务注册事件
        NotifyCenter.publishEvent(new ClientOperationEvent.ClientRegisterServiceEvent(singleton, clientId));
        NotifyCenter
                .publishEvent(new MetadataEvent.InstanceMetadataEvent(singleton, instanceInfo.getMetadataId(), false));
    }

	// com.alibaba.nacos.naming.core.v2.ServiceManager
    /**
     * Get singleton service. Put to manager if no singleton.
     * 获取单例服务(单例意味着整个应用只有一个对象),如果不存在,则注册到Manager
     */
    public Service getSingleton(Service service) {
    	// 如果不存在就注册
        singletonRepository.computeIfAbsent(service, key -> {
        	// 发布服务元信息数据事件。不过该事件对于持久实例才有用处。
            NotifyCenter.publishEvent(new MetadataEvent.ServiceMetadataEvent(service, false));
            return service;
        });
        Service result = singletonRepository.get(service);
        // 将服务登记到namespace
        namespaceSingletonMaps.computeIfAbsent(result.getNamespace(), namespace -> new ConcurrentHashSet<>());
        namespaceSingletonMaps.get(result.getNamespace()).add(result);
        return result;
    }
	// 再看看ClientOperationEvent.ClientRegisterServiceEvent
	// > com.alibaba.nacos.naming.core.v2.index.ClientServiceIndexesManager#onEvent
	//  > com.alibaba.nacos.naming.core.v2.index.ClientServiceIndexesManager#handleClientOperation
	//   > com.alibaba.nacos.naming.core.v2.index.ClientServiceIndexesManager#addPublisherIndexes
	// 登记发布服务的客户端
    private void addPublisherIndexes(Service service, String clientId) {
        publisherIndexes.computeIfAbsent(service, key -> new ConcurrentHashSet<>());
        publisherIndexes.get(service).add(clientId);
        NotifyCenter.publishEvent(new ServiceEvent.ServiceChangedEvent(service, true));
    }

小结

我们总结一下,以上涉及到的概念。

概念 管理者 描述
服务 com.alibaba.nacos.naming.core.v2.pojo.Service ServiceManager

除了这个概念,实际上我们还看到Client的内部结构:

AbstractClient:

  • 记录客户端发布的服务:ConcurrentHashMap<Service, InstancePublishInfo> publishers
  • 记录客户端订阅的服务:ConcurrentHashMap<Service, Subscriber> subscribers 这个点其实要到订阅服务请求才会分析到,但为了信息不会太分散,所以就放到一起了。

ClientServiceIndexesManager

  • 客户端索引管理者。这里的索引指的是,通过Service快速找到客户端,只是客户端有ClientManager,如果这里再存一份也不合适,不利于数据维护。因此这里存的是clientId。估计也是如此,他才叫客户端索引管理者。

查询和订阅实例

java 复制代码
> com.alibaba.nacos.client.naming.NacosNamingService#getAllInstances(java.lang.String, java.lang.String, java.util.List<java.lang.String>)
 > com.alibaba.nacos.client.naming.NacosNamingService#getAllInstances(java.lang.String, java.lang.String, java.util.List<java.lang.String>, boolean)
  > com.alibaba.nacos.client.naming.NacosNamingService#getAllInstances(String serviceName, String groupName, List<String> clusters, boolean subscribe)
   > com.alibaba.nacos.client.naming.remote.gprc.NamingGrpcClientProxy#subscribe
    > com.alibaba.nacos.client.naming.remote.gprc.NamingGrpcClientProxy#doSubscribe
    public ServiceInfo doSubscribe(String serviceName, String groupName, String clusters) throws NacosException {
    	// 重点SubscribeServiceRequest,看服务端代码需要知道是什么请求
        SubscribeServiceRequest request = new SubscribeServiceRequest(namespaceId, groupName, serviceName, clusters, true);
        SubscribeServiceResponse response = requestToServer(request, SubscribeServiceResponse.class);
        redoService.subscriberRegistered(serviceName, groupName, clusters);
        return response.getServiceInfo();
    }
    

服务端

java 复制代码
	// com.alibaba.nacos.naming.remote.rpc.handler.SubscribeServiceRequestHandler
    @Override
    @Secured(action = ActionTypes.READ)
    public SubscribeServiceResponse handle(SubscribeServiceRequest request, RequestMeta meta) throws NacosException {
        String namespaceId = request.getNamespace();
        String serviceName = request.getServiceName();
        String groupName = request.getGroupName();
        String app = request.getHeader("app", "unknown");
        String groupedServiceName = NamingUtils.getGroupedName(serviceName, groupName);
        Service service = Service.newService(namespaceId, groupName, serviceName, true);
        // 订阅者
        Subscriber subscriber = new Subscriber(meta.getClientIp(), meta.getClientVersion(), app, meta.getClientIp(),
                namespaceId, groupedServiceName, 0, request.getClusters());
        // 服务信息,这里有几个参数是需要通过方法来获取的
        // 重点是:serviceStorage.getData(service)
        // 而这个方法也是个重要的方法,过滤cluster、健康实例,并执行自动保护机制,都是他实现的
        ServiceInfo serviceInfo = ServiceUtil.selectInstancesWithHealthyProtection(serviceStorage.getData(service),
                metadataManager.getServiceMetadata(service).orElse(null), subscriber.getCluster(), false,
                true, subscriber.getIp());
        if (request.isSubscribe()) {
        	// 订阅服务
            clientOperationService.subscribeService(service, subscriber, meta.getConnectionId());
            // 发布订阅事件
            NotifyCenter.publishEvent(new SubscribeServiceTraceEvent(System.currentTimeMillis(),
                    meta.getClientIp(), service.getNamespace(), service.getGroup(), service.getName()));
        } else {
        	// 取消订阅
            clientOperationService.unsubscribeService(service, subscriber, meta.getConnectionId());
            NotifyCenter.publishEvent(new UnsubscribeServiceTraceEvent(System.currentTimeMillis(),
                    meta.getClientIp(), service.getNamespace(), service.getGroup(), service.getName()));
        }
        return new SubscribeServiceResponse(ResponseCode.SUCCESS.getCode(), "success", serviceInfo);
    }
	
	// > com.alibaba.nacos.naming.core.v2.index.ServiceStorage#getData
	//  > com.alibaba.nacos.naming.core.v2.index.ServiceStorage#getPushData
	//   > com.alibaba.nacos.naming.core.v2.index.ServiceStorage#getAllInstancesFromIndex
	// > com.alibaba.nacos.naming.utils.ServiceUtil#selectInstancesWithHealthyProtection(com.alibaba.nacos.api.naming.pojo.ServiceInfo, com.alibaba.nacos.naming.core.v2.metadata.ServiceMetadata, java.lang.String, boolean, boolean, java.lang.String)
	//  > com.alibaba.nacos.naming.utils.ServiceUtil#doSelectInstances
	// 上面是调用路径,这里把生产数据的方法重点捞出来
	// ServiceStorage的数据生产
    public ServiceInfo getPushData(Service service) {
        ServiceInfo result = emptyServiceInfo(service);
        if (!ServiceManager.getInstance().containSingleton(service)) {
            return result;
        }
        Service singleton = ServiceManager.getInstance().getSingleton(service);
        result.setHosts(getAllInstancesFromIndex(singleton));
        // 从ServiceManager拿到服务的实例信息,并登记到ServiceStorage#serviceDataIndexes中
        serviceDataIndexes.put(singleton, result);
        return result;
    }
    private List<Instance> getAllInstancesFromIndex(Service service) {
        Set<Instance> result = new HashSet<>();
        Set<String> clusters = new HashSet<>();
        
        for (String each : serviceIndexesManager.getAllClientsRegisteredService(service)) {
        	// 获取实例信息
            Optional<InstancePublishInfo> instancePublishInfo = getInstanceInfo(each, service);
            if (instancePublishInfo.isPresent()) {
                InstancePublishInfo publishInfo = instancePublishInfo.get();
                //If it is a BatchInstancePublishInfo type, it will be processed manually and added to the instance list
                if (publishInfo instanceof BatchInstancePublishInfo) {
                    BatchInstancePublishInfo batchInstancePublishInfo = (BatchInstancePublishInfo) publishInfo;
                    List<Instance> batchInstance = parseBatchInstance(service, batchInstancePublishInfo, clusters);
                    result.addAll(batchInstance);
                } else {
                    Instance instance = parseInstance(service, instancePublishInfo.get());
                    result.add(instance);
                    clusters.add(instance.getClusterName());
                }
            }
        }
        // cache clusters of this service
        // 缓存集群信息
        serviceClusterIndex.put(service, clusters);
        return new LinkedList<>(result);
    }
        
    private Optional<InstancePublishInfo> getInstanceInfo(String clientId, Service service) {
    	// 通过客户端ID,获取到Client,进而从其获取客户端发布的服务。
        Client client = clientManager.getClient(clientId);
        if (null == client) {
            return Optional.empty();
        }
        return Optional.ofNullable(client.getInstancePublishInfo(service));
    }

从查询实例这里,我们看到有个数据存储:ServiceStorage。重要的是,这个虽然叫存储,但是实际上里面的数据却是从别处获取来的。来源于:ServiceManager、ServiceIndexesManager、ClientManager。从这个角度说,更像是个缓存。

总结

上面的整了一堆源代码,容易看烦了。感兴趣的,可以根据上面的源码深入看看。为了方便大家,我画了图给大家:

为了让大家重点看到数据生产过程:

从图中,我们可以看到,nacos2.x的数据结构并不像官方的Service->Group->Instance。而是按照Connection、Client、Service分别通过对应的管理器进行管理。此外,为了避免数据多处存储,还有ClientServiceIndexesManager作为Client和Service的桥梁。

除此之外,还有ServiceStorage,作为数据缓存。不过,当我们深入了解ServiceStorage时,会发现他的数据一致性/数据的更新,是在给订阅服务的客户端定时推送时通过调用com.alibaba.nacos.naming.core.v2.index.ServiceStorage#getPushData来实现的。个人认为这是有可以优化空间的,他们完全可以通过各种事件来监听实例的生死来更新数据。

总而言之,如果不算ServiceStorage这个缓存,那么数据主要存在于一下的Manager中:

ConnectionManager、ClientManager、ServiceManager、ClientServiceIndexesManager。

到这里,可能有同学就有疑问了。那么Group、Cluster这些概念去哪了呢?

这些概念都成为了属性/字段了。
com.alibaba.nacos.naming.core.v2.pojo.Service#group
com.alibaba.nacos.api.naming.pojo.Instance#clusterName

即使在ServiceStorage封装ServiceInfo时,他们也是作为属性来存储的。通过ServiceUtil来过滤目标实例。

最后,提醒大家一下,我们这里只是分析了临时实例。是最常用的场景。当然,如果我们用Nacos的持久实例,SpringCloud也就自然支持了持久实例。不过,咱们不深究了,感兴趣的同学,可以顺着往下挖一挖持久实例。

后记

这种深度刨析源码、深挖一个技术细节的实现,太费时间、也太费篇幅了。我自己都感觉差点把整个nacos的源码都搬上来了。莫见怪。。。

关于nacos的一致性协议,就不在这里聊了,这个东西得单独倒腾,还要与其他分布式中间件相互对比,还有理论。。

下次,咱们先往后聊OpenFeign。

推荐

Nacos的实现原理在官网有一电子书《Nacos架构&原理》,想要了解顶层设计原理的同学,建议看看。

相关推荐
aloha_7891 小时前
从零记录搭建一个干净的mybatis环境
java·笔记·spring·spring cloud·maven·mybatis·springboot
茶馆大橘3 小时前
微服务系列五:避免雪崩问题的限流、隔离、熔断措施
java·jmeter·spring cloud·微服务·云原生·架构·sentinel
荆州克莱8 小时前
[FE] React 初窥门径(四):React 组件的加载过程(render 阶段)
spring boot·spring·spring cloud·css3·技术
Genius Kim10 小时前
SpringCloud Sentinel 服务治理详解
spring cloud·sentinel·php
为美好的生活献上中指11 小时前
Java学习Day60:微服务总结!(有经处无火,无火处无经)
java·spring boot·spring cloud·微服务·sentinel·jetty
BUG指挥官17 小时前
深度解析阿里的Sentinel
spring boot·spring·spring cloud·sentinel
WANT_如初17 小时前
SpringCloud-Eureka注册中心
java·spring cloud
白总Server1 天前
OpenHarmony
后端·spring cloud·华为·微服务·ribbon·架构·系统架构
荆州克莱1 天前
VUE2升级成VUE3的优化与区别
spring boot·spring·spring cloud·css3·技术