spring集成rabbitmq实现rpc

阅读更多

 

public Object convertSendAndReceive(final String routingKey, final Object message) throws AmqpException {
		return this.convertSendAndReceive(this.exchange, routingKey, message, null);
	}

 一

 

 spring整合Rabbit MQ提供了Reply来实现RPC，AMQP协议定义了14中消息的属性，其中两项，一项是Replyto，表示返回消息的队列，一个是correlationId 用来表示发送消息和返回消息的标志，来区分是否是一个调用

下面一步步来实现RPC

首先贴出spring配置文件代码

 其中消息转换器类：

 

 

package com.temp.rabbit;

import org.springframework.amqp.core.Message;
import org.springframework.amqp.core.MessageProperties;
import org.springframework.amqp.support.converter.AbstractJsonMessageConverter;
import org.springframework.amqp.support.converter.MessageConversionException;
import org.springframework.util.SerializationUtils;

public class BytesMessageConverter extends AbstractJsonMessageConverter{

	@Override
	protected Message createMessage(Object msg, MessageProperties msgPro) {
		byte[] data = SerializationUtils.serialize(msg);
		msgPro.setContentLength(data.length);
		System.out.println("create message "+msg.getClass());
		return new Message(data , msgPro);
	}

	@Override
	public Object fromMessage(Message msg) throws MessageConversionException {
		byte[] data = msg.getBody() ;
		Object result = SerializationUtils.deserialize(data);
		System.out.println("create obj "+result.getClass());
		return result;
	}

}

 消费者处理handler

 

 

package com.temp.rabbit.receive;

import java.lang.reflect.Method;

import com.temp.rabbit.bean.RpcRequest;
import com.temp.rabbit.bean.RpcResponse;
import com.temp.rabbit.bean.TempServiceImp;

public class MessageHandler {

    //没有设置默认的处理方法的时候，方法名是handleMessage
    public RpcResponse handleMessage(RpcRequest message){
    	Class clazz = message.getClassName() ;
    	RpcResponse response = new RpcResponse();
    	Method method;
		try {
			method = clazz.getMethod(message.getMethodName(), message.getParamType());
			Object result = method.invoke(new TempServiceImp(), message.getParams());
			response.setResult(result);
		} catch (Exception e) {
			e.printStackTrace();
		}
        return response ;
    }
}

 服务提供：

 

 

package com.temp.rabbit.bean;

public class TempServiceImp implements TempService {

	public String sayHello(){
		return "TempServiceImp hello ... " ;
	}
	
}

 代理类：

 

 

package com.temp.rabbit.receive.proxy;

import java.io.Serializable;
import java.lang.reflect.Method;

import org.springframework.beans.factory.annotation.Autowired;

import com.temp.rabbit.bean.RpcRequest;
import com.temp.rabbit.bean.RpcResponse;
import com.temp.rabbit.send.SendRabbitMsgImp;

import net.sf.cglib.proxy.Enhancer;
import net.sf.cglib.proxy.MethodInterceptor;
import net.sf.cglib.proxy.MethodProxy;

public class ServiceProxy implements MethodInterceptor{

	private Enhancer enhancer = new Enhancer();
	
	private T t ;
	
	public void setT(T t){
		this.t = t ;
	}
	@Autowired
	private SendRabbitMsgImp rabbitMsg ;
	
	public Object getProxy(){
		enhancer.setSuperclass(t.getClass());
		enhancer.setCallback(this);
		return enhancer.create();
	}
	
	
	@Override
	public Object intercept(Object obj, Method method, Object[] param, MethodProxy proxy) throws Throwable {
		RpcRequest request = new RpcRequest();
		request.setMethodName(method.getName());
		request.setClassName(t.getClass());
		Class[] paramType = new Class[param.length];
		Serializable[] para = new Serializable[param.length];
		for(int i = 0 ; i < param.length ; i ++){
			paramType[i] = param[i].getClass();
			para[i] = (Serializable)param[i];
		}
		request.setParams(para);
		request.setParamType(paramType);
		RpcResponse result = (RpcResponse)rabbitMsg.sendAdcReceive("dengwei.goods", request) ; 
		return result.getResult();
	}

}

 主程序

 

 

package com.temp;

import org.springframework.context.ApplicationContext;
import org.springframework.context.support.ClassPathXmlApplicationContext;

import com.temp.rabbit.bean.TempService;

public class Main {

	public static void main(String[] args) {
		ApplicationContext app = new ClassPathXmlApplicationContext("classpath:spring.xml");  
		TempService proxy = (TempService)app.getBean("proxyService");
		System.out.println("main result " + proxy.sayHello()) ;
	}
	
} 

 

消息发送实现：

 

package com.temp.rabbit.send;

import org.springframework.amqp.rabbit.core.RabbitTemplate;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Component;

@Component("sendMsg")
public class SendRabbitMsgImp implements SendRabbitMsg{

	@Autowired
	private RabbitTemplate template ;
	
	@Override
	public void sendData2Queue(String queueKey, Object msg) {
		try {
			template.convertAndSend(queueKey, msg);
		} catch (Exception e) {
			e.printStackTrace();
			System.out.println("send data 2 msg erro ");
		}
		System.out.println("消息已发送");
	}
	
	@Override
	public Object sendAdcReceive(String queueKey , Object msg){
		try {
			return template.convertSendAndReceive(queueKey, msg);
		} catch (Exception e) {
			e.printStackTrace();
			System.out.println("send data 2 msg erro ");
		}
		System.out.println("消息已发送");
		return null ; 
	}
}

 

 

这里面的RpcRequest和RpcResponse就不贴代码了

这里讲一下原理实现，我们可以跟着源码看一下

首先调用的是RabbitTemplate的

 

public Object convertSendAndReceive(final String routingKey, final Object message) throws AmqpException {
		return this.convertSendAndReceive(this.exchange, routingKey, message, null);
	}

 然后一路走下去

@Override
	public Object convertSendAndReceive(final String exchange, final String routingKey, final Object message,
			final MessagePostProcessor messagePostProcessor) throws AmqpException {
		Message requestMessage = convertMessageIfNecessary(message);
		if (messagePostProcessor != null) {
			requestMessage = messagePostProcessor.postProcessMessage(requestMessage);
		}
		Message replyMessage = this.doSendAndReceive(exchange, routingKey, requestMessage);
		if (replyMessage == null) {
			return null;
		}
		return this.getRequiredMessageConverter().fromMessage(replyMessage);
	}

protected Message doSendAndReceive(final String exchange, final String routingKey, final Message message) {
		if (this.replyQueue == null) {
			return doSendAndReceiveWithTemporary(exchange, routingKey, message);
		}
		else {
			return doSendAndReceiveWithFixed(exchange, routingKey, message);
		}
	}

 到这里我们会看到，有一个分支如果replyqueue不为空则是走另外的一个方法，因为之前没有设置replyqueue所以，这里会

 

走第一步方法，也就是doSendAndReceiveWithTemporary

来看一下这个方法源码

 

protected Message doSendAndReceiveWithTemporary(final String exchange, final String routingKey, final Message message) {
		return this.execute(new ChannelCallback() {

			@Override
			public Message doInRabbit(Channel channel) throws Exception {
				final ArrayBlockingQueue replyHandoff = new ArrayBlockingQueue(1);

				Assert.isNull(message.getMessageProperties().getReplyTo(),
						"Send-and-receive methods can only be used if the Message does not already have a replyTo property.");
				DeclareOk queueDeclaration = channel.queueDeclare();
				String replyTo = queueDeclaration.getQueue();
				message.getMessageProperties().setReplyTo(replyTo);

				String consumerTag = UUID.randomUUID().toString();
				DefaultConsumer consumer = new DefaultConsumer(channel) {

					@Override
					public void handleDelivery(String consumerTag, Envelope envelope, AMQP.BasicProperties properties,
											   byte[] body) throws IOException {
						MessageProperties messageProperties = messagePropertiesConverter.toMessageProperties(
								properties, envelope, encoding);
						Message reply = new Message(body, messageProperties);
						if (logger.isTraceEnabled()) {
							logger.trace("Message received " + reply);
						}
						try {
							replyHandoff.put(reply);
						}
						catch (InterruptedException e) {
							Thread.currentThread().interrupt();
						}
					}
				};
				channel.basicConsume(replyTo, true, consumerTag, true, true, null, consumer);
				doSend(channel, exchange, routingKey, message, null);
				Message reply = (replyTimeout < 0) ? replyHandoff.take() : replyHandoff.poll(replyTimeout,
						TimeUnit.MILLISECONDS);
				channel.basicCancel(consumerTag);
				return reply;
			}
		});
	}

 这里流程就是申明一个大小为1的临时队列，然后发送消息，然后监听返回的消息，放到临时队列，然后取出返回消息。

 

那么因为每次都会创建临时队列，所以对性能是个考验那么有第二种方式，在rabbitmq中申明一个返回队列，用来存放该服务的返回消息。

那么需要在spring配置文件中配置一个reply队列

然后在消息监听容器中再配置一个发送消息的模板template为消费者

 

 

 

最后再发送消息的实现中即SendRabbitMsgImp类中注入队列

 

 

@Autowired
	@Qualifier("reply")
	private Queue reply ;

 

 

然后设置template的replyqueue为reply     ；template.setReplyQueue(reply);

这个设置代码可以再初始化方法中，也可以再发送消息之前，其实最好的实在spring中设置

那么该说原理了，我们可以看最开始发送消息的第二个方法

protected Message doSendAndReceive(final String exchange, final String routingKey, final Message message) {
		if (this.replyQueue == null) {
			return doSendAndReceiveWithTemporary(exchange, routingKey, message);
		}
		else {
			return doSendAndReceiveWithFixed(exchange, routingKey, message);
		}
	}

 

protected Message doSendAndReceiveWithFixed(final String exchange, final String routingKey, final Message message) {
		return this.execute(new ChannelCallback() {

			@Override
			public Message doInRabbit(Channel channel) throws Exception {
				final PendingReply pendingReply = new PendingReply();
				String messageTag = UUID.randomUUID().toString();
				RabbitTemplate.this.replyHolder.put(messageTag, pendingReply);
				// Save any existing replyTo and correlation data
				String savedReplyTo = message.getMessageProperties().getReplyTo();
				pendingReply.setSavedReplyTo(savedReplyTo);
				if (StringUtils.hasLength(savedReplyTo) && logger.isDebugEnabled()) {
					logger.debug("Replacing replyTo header:" + savedReplyTo
							+ " in favor of template's configured reply-queue:"
							+ RabbitTemplate.this.replyQueue.getName());
				}
				message.getMessageProperties().setReplyTo(RabbitTemplate.this.replyQueue.getName());
				String savedCorrelation = null;
				if (RabbitTemplate.this.correlationKey == null) { // using standard correlationId property
					byte[] correlationId = message.getMessageProperties().getCorrelationId();
					if (correlationId != null) {
						savedCorrelation = new String(correlationId,
								RabbitTemplate.this.encoding);
					}
				}
				else {
					savedCorrelation = (String) message.getMessageProperties()
							.getHeaders().get(RabbitTemplate.this.correlationKey);
				}
				pendingReply.setSavedCorrelation(savedCorrelation);
				if (RabbitTemplate.this.correlationKey == null) { // using standard correlationId property
					message.getMessageProperties().setCorrelationId(messageTag
							.getBytes(RabbitTemplate.this.encoding));
				}
				else {
					message.getMessageProperties().setHeader(
							RabbitTemplate.this.correlationKey, messageTag);
				}

				if (logger.isDebugEnabled()) {
					logger.debug("Sending message with tag " + messageTag);
				}
				doSend(channel, exchange, routingKey, message, null);
				LinkedBlockingQueue replyHandoff = pendingReply.getQueue();
				Message reply = (replyTimeout < 0) ? replyHandoff.take() : replyHandoff.poll(replyTimeout,
						TimeUnit.MILLISECONDS);
				RabbitTemplate.this.replyHolder.remove(messageTag);
				return reply;
			}
		});
	}

 这个方法并没有申请临时队列，发送消息后直接再pendingReply中的队列中取，那么怎么放到pendingReply的队列中区的呢，可以看到，RabbitTemplate是实现了MessageLIstener,那么看他实现的onMessage方法

public void onMessage(Message message) {
		try {
			String messageTag;
			if (this.correlationKey == null) { // using standard correlationId property
				messageTag = new String(message.getMessageProperties().getCorrelationId(), this.encoding);
			}
			else {
				messageTag = (String) message.getMessageProperties()
						.getHeaders().get(this.correlationKey);
			}
			if (messageTag == null) {
				logger.error("No correlation header in reply");
				return;
			}

			PendingReply pendingReply = this.replyHolder.get(messageTag);
			if (pendingReply == null) {
				if (logger.isWarnEnabled()) {
					logger.warn("Reply received after timeout for " + messageTag);
				}
			}
			else {
				// Restore the inbound correlation data
				String savedCorrelation = pendingReply.getSavedCorrelation();
				if (this.correlationKey == null) {
					if (savedCorrelation == null) {
						message.getMessageProperties().setCorrelationId(null);
					}
					else {
						message.getMessageProperties().setCorrelationId(
								savedCorrelation.getBytes(this.encoding));
					}
				}
				else {
					if (savedCorrelation != null) {
						message.getMessageProperties().setHeader(this.correlationKey,
							savedCorrelation);
					}
					else {
						message.getMessageProperties().getHeaders().remove(this.correlationKey);
					}
				}
				// Restore any inbound replyTo
				String savedReplyTo = pendingReply.getSavedReplyTo();
				message.getMessageProperties().setReplyTo(savedReplyTo);
				LinkedBlockingQueue queue = pendingReply.getQueue();
				queue.add(message);
				if (logger.isDebugEnabled()) {
					logger.debug("Reply received for " + messageTag);
					if (savedReplyTo != null) {
						logger.debug("Restored replyTo to " + savedReplyTo);
					}
				}
			}
		}
		catch (UnsupportedEncodingException e) {
			throw new AmqpIllegalStateException("Invalid Character Set:" + this.encoding, e);
		}
	}

 这里就明白了，根据唯一id也就是前面说的correlationId找到消息的pendingReply，然后将返回的消息放到pendingReply的队列中，这样就实现了RPC的调用，下面附上工程代码

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

• taobao-temp-order.rar (81.7 KB)
• 下载次数: 1