spring Kafka中的@KafkaListener源碼分析

本篇內容主要講解“spring Kafka中的@KafkaListener源碼分析”，感興趣的朋友不妨來看看。本文介紹的方法操作簡單快捷，實用性強。下面就讓小編來帶大家學習“spring Kafka中的@KafkaListener源碼分析”吧!

一、總體流程

從spring啟動開始處理@KafkaListener，到start消息監聽整體流程圖

二、源碼解讀

1、postProcessAfterInitialization

KafkaListenerAnnotationBeanPostProcessor#postProcessAfterInitialization

	public Object postProcessAfterInitialization(final Object bean, final String beanName) throws BeansException {
		if (!this.nonAnnotatedClasses.contains(bean.getClass())) {
		    Class<?> targetClass = AopUtils.getTargetClass(bean);
		    
		    // 掃描@KafkaListener注解
			Collection<KafkaListener> classLevelListeners = findListenerAnnotations(targetClass);
			
			......
			
			if (annotatedMethods.isEmpty()) {
				this.nonAnnotatedClasses.add(bean.getClass());
				this.logger.trace(() -> "No @KafkaListener annotations found on bean type: " + bean.getClass());
			}
			else {
				// Non-empty set of methods
				for (Map.Entry<Method, Set<KafkaListener>> entry : annotatedMethods.entrySet()) {
					Method method = entry.getKey();
					// 遍歷掃描到的所有@KafkaListener注解并開始處理
					for (KafkaListener listener : entry.getValue()) {
						processKafkaListener(listener, method, bean, beanName);
					}
				}
				this.logger.debug(() -> annotatedMethods.size() + " @KafkaListener methods processed on bean '"
							+ beanName + "': " + annotatedMethods);
			}
			// 處理在類上的@KafkaListener注解
			if (hasClassLevelListeners) {
				processMultiMethodListeners(classLevelListeners, multiMethods, bean, beanName);
			}
		}
		return bean;
	}

1.1、processKafkaListener

KafkaListenerAnnotationBeanPostProcessor#processKafkaListener

	protected void processKafkaListener(KafkaListener kafkaListener, Method method, Object bean, String beanName) {
		Method methodToUse = checkProxy(method, bean);
		MethodKafkaListenerEndpoint<K, V> endpoint = new MethodKafkaListenerEndpoint<>();
		endpoint.setMethod(methodToUse);
		processListener(endpoint, kafkaListener, bean, methodToUse, beanName);
	}

1.2、processListener

KafkaListenerAnnotationBeanPostProcessor#processListener

將每個kafkaListener轉變成MethodKafkaListenerEndpoint并注冊到KafkaListenerEndpointRegistrar容器，方便后續統一啟動監聽

	protected void processListener(MethodKafkaListenerEndpoint<?, ?> endpoint, KafkaListener kafkaListener,
			Object bean, Object adminTarget, String beanName) {

		String beanRef = kafkaListener.beanRef();
		if (StringUtils.hasText(beanRef)) {
			this.listenerScope.addListener(beanRef, bean);
		}
		endpoint.setBean(bean);
		endpoint.setMessageHandlerMethodFactory(this.messageHandlerMethodFactory);
		endpoint.setId(getEndpointId(kafkaListener));
		endpoint.setGroupId(getEndpointGroupId(kafkaListener, endpoint.getId()));
		endpoint.setTopicPartitions(resolveTopicPartitions(kafkaListener));
		endpoint.setTopics(resolveTopics(kafkaListener));
		endpoint.setTopicPattern(resolvePattern(kafkaListener));
		endpoint.setClientIdPrefix(resolveExpressionAsString(kafkaListener.clientIdPrefix(), "clientIdPrefix"));
		String group = kafkaListener.containerGroup();

        ......
      
        // 注冊已經封裝好的消費端-endpoint
		this.registrar.registerEndpoint(endpoint, factory);
		
		if (StringUtils.hasText(beanRef)) {
			this.listenerScope.removeListener(beanRef);
		}
	}

1.3、registerEndpoint

KafkaListenerEndpointRegistrar#registerEndpoint

	public void registerEndpoint(KafkaListenerEndpoint endpoint, KafkaListenerContainerFactory<?> factory) {
		
	    ......
		
		KafkaListenerEndpointDescriptor descriptor = new KafkaListenerEndpointDescriptor(endpoint, factory);
		synchronized (this.endpointDescriptors) {
		    // 如果到了需要立即啟動監聽的階段就直接注冊并監聽(也就是創建消息監聽容器并啟動)
			if (this.startImmediately) { // Register and start immediately
				this.endpointRegistry.registerListenerContainer(descriptor.endpoint,
						resolveContainerFactory(descriptor), true);
			}
			else {
			    // 一般情況都先走這一步，添加至此列表，待bean后續的生命周期 統一注冊并啟動
				this.endpointDescriptors.add(descriptor);
			}
		}
	}

	public void registerListenerContainer(KafkaListenerEndpoint endpoint, KafkaListenerContainerFactory<?> factory,
			boolean startImmediately) {

        ......
        
		synchronized (this.listenerContainers) {
		
			......
			
			// 1.創建消息監聽容器
			MessageListenerContainer container = createListenerContainer(endpoint, factory);
			this.listenerContainers.put(id, container);
			if (StringUtils.hasText(endpoint.getGroup()) && this.applicationContext != null) {
				List<MessageListenerContainer> containerGroup;
				if (this.applicationContext.containsBean(endpoint.getGroup())) {
					containerGroup = this.applicationContext.getBean(endpoint.getGroup(), List.class);
				}
				else {
					containerGroup = new ArrayList<MessageListenerContainer>();
					this.applicationContext.getBeanFactory().registerSingleton(endpoint.getGroup(), containerGroup);
				}
				containerGroup.add(container);
			}
            
            // 2.是否立即啟動消息監聽
			if (startImmediately) {
				startIfNecessary(container);
			}
		}
	}

1.4、startIfNecessary

KafkaListenerEndpointRegistry#startIfNecessary
啟動消息監聽

	private void startIfNecessary(MessageListenerContainer listenerContainer) {
		if (this.contextRefreshed || listenerContainer.isAutoStartup()) {
		    // 啟動消息監聽
		    // 到這一步之后，消息監聽以及處理都是KafkaMessageListenerContainer的邏輯
		    // 到此也就打通了@KafkaListener到MessageListenerContainer消息監聽容器的邏輯
			listenerContainer.start();
		}
	}

2、afterSingletonsInstantiated

這一步是實例化(此處的實例化是已經創建對象并完成了初始化操作)之后，緊接著的操作

KafkaListenerAnnotationBeanPostProcessor#afterSingletonsInstantiated

	public void afterSingletonsInstantiated() {
		this.registrar.setBeanFactory(this.beanFactory);

        // 對"注冊員"信息的完善
		if (this.beanFactory instanceof ListableBeanFactory) {
			Map<String, KafkaListenerConfigurer> instances =
					((ListableBeanFactory) this.beanFactory).getBeansOfType(KafkaListenerConfigurer.class);
			for (KafkaListenerConfigurer configurer : instances.values()) {
				configurer.configureKafkaListeners(this.registrar);
			}
		}

		if (this.registrar.getEndpointRegistry() == null) {
			if (this.endpointRegistry == null) {
				Assert.state(this.beanFactory != null,
						"BeanFactory must be set to find endpoint registry by bean name");
				this.endpointRegistry = this.beanFactory.getBean(
						KafkaListenerConfigUtils.KAFKA_LISTENER_ENDPOINT_REGISTRY_BEAN_NAME,
						KafkaListenerEndpointRegistry.class);
			}
			this.registrar.setEndpointRegistry(this.endpointRegistry);
		}

		......

		// Actually register all listeners
		// 整個方法這里才是關鍵
		// 創建MessageListenerContainer并注冊
		this.registrar.afterPropertiesSet();
	}

2.1、afterPropertiesSet

KafkaListenerEndpointRegistrar#afterPropertiesSet

	public void afterPropertiesSet() {
		registerAllEndpoints();
	}

2.2、registerAllEndpoints

KafkaListenerEndpointRegistrar#registerAllEndpoints

	protected void registerAllEndpoints() {
		synchronized (this.endpointDescriptors) {
			for (KafkaListenerEndpointDescriptor descriptor : this.endpointDescriptors) {
			    // 這里是真正的創建ListenerContainer監聽對象并注冊
				this.endpointRegistry.registerListenerContainer(
						descriptor.endpoint, resolveContainerFactory(descriptor));
			}
			// 啟動時所有消息監聽對象都注冊之后，便將參數置為true
			this.startImmediately = true;  // trigger immediate startup
		}
	}

到此，相信大家對“spring Kafka中的@KafkaListener源碼分析”有了更深的了解，不妨來實際操作一番吧！這里是億速云網站，更多相關內容可以進入相關頻道進行查詢，關注我們，繼續學習！