RocketMQ重试机制和消息幂
一.重试机制
1.由于MQ经常处于复杂的分布式系统中,考虑网络波动,服务宕机,程序异常因素,很有可能出现消息发送或者消费失败的问题。因此,消息的重试就是所有MQ中间件必须考虑到的一个关键点。如果没有消息重试,就可能产生消息丢失的问题,可能对系统产生很大的影响。所以,秉承宁可多发消息,也不可丢失消息的原则,大部分MQ都对消息重试提供了很好的支持。
2.RocketMQ为了使用者封装了消息重试的处理流程,无需开发人员手动处理。RocketMQ支持了生产端和消费端两类重试机制。
模拟异常
Consumer端消息消费两种状态:
package com.alibaba.rocketmq.client.consumer.listener; public enum ConsumeConcurrentlyStatus { CONSUME_SUCCESS, RECONSUME_LATER; private ConsumeConcurrentlyStatus() { } }
一个是成功(CONSUME_SUCCESS),一个是失败&重试(RECONSUME_LATER);
Consumer为了保证消息消费成功,只有使用方明确表示消费成功,返回CONSUME_SUCCESS,RocketMQ才会认为消息消费成功。
如果消息消费失败,只要返回ConsumeConcurrentlyStatus.RECONSUME_LATER,RocketMQ就会认为消息消费失败了,需要重新投递。
1.出现异常
package com.wn.consumer; import com.alibaba.rocketmq.client.consumer.DefaultMQPushConsumer; import com.alibaba.rocketmq.client.consumer.listener.ConsumeConcurrentlyContext; import com.alibaba.rocketmq.client.consumer.listener.ConsumeConcurrentlyStatus; import com.alibaba.rocketmq.client.consumer.listener.MessageListenerConcurrently; import com.alibaba.rocketmq.client.exception.MQClientException; import com.alibaba.rocketmq.common.message.MessageExt; import java.util.List; public class MQConsumer { public static void main(String[] args) throws MQClientException { //创建消费者 DefaultMQPushConsumer consumer=new DefaultMQPushConsumer("rmq-group"); //设置NameServer地址 consumer.setNamesrvAddr("192.168.138.187:9876;192.168.138.188:9876"); //设置消费者实例名称 consumer.setInstanceName("consumer"); //订阅topic consumer.subscribe("wn02","TagA"); //监听消息 consumer.registerMessageListener(new MessageListenerConcurrently() { @Override public ConsumeConcurrentlyStatus consumeMessage(List<MessageExt> list, ConsumeConcurrentlyContext consumeConcurrentlyContext) { //获取消息 for (MessageExt msg:list){ System.out.println(msg.getMsgId()+"---"+new String(msg.getBody())); } try { int i=1/0; }catch (Exception e){ e.printStackTrace(); //需要重试 return ConsumeConcurrentlyStatus.RECONSUME_LATER; }//消息成功 return ConsumeConcurrentlyStatus.CONSUME_SUCCESS; } }); consumer.start(); System.out.println("Consumer Started..."); } }
2.网络延迟
package com.wn.consumer; import com.alibaba.rocketmq.client.consumer.DefaultMQPushConsumer; import com.alibaba.rocketmq.client.consumer.listener.ConsumeConcurrentlyContext; import com.alibaba.rocketmq.client.consumer.listener.ConsumeConcurrentlyStatus; import com.alibaba.rocketmq.client.consumer.listener.MessageListenerConcurrently; import com.alibaba.rocketmq.client.exception.MQClientException; import com.alibaba.rocketmq.common.message.MessageExt; import java.util.List; public class MQConsumer { public static void main(String[] args) throws MQClientException { //创建消费者 DefaultMQPushConsumer consumer=new DefaultMQPushConsumer("rmq-group"); //设置NameServer地址 consumer.setNamesrvAddr("192.168.138.187:9876;192.168.138.188:9876"); //设置消费者实例名称 consumer.setInstanceName("consumer"); //订阅topic consumer.subscribe("wn03","TagA"); //监听消息 consumer.registerMessageListener(new MessageListenerConcurrently() { @Override public ConsumeConcurrentlyStatus consumeMessage(List<MessageExt> list, ConsumeConcurrentlyContext consumeConcurrentlyContext) { //获取消息 for (MessageExt msg:list){ System.out.println(msg.getMsgId()+"---"+new String(msg.getBody())); } //网络延迟 try { Thread.sleep(600000); } catch (InterruptedException e) { e.printStackTrace(); } //消息成功 return ConsumeConcurrentlyStatus.CONSUME_SUCCESS; } }); consumer.start(); System.out.println("Consumer Started..."); } }
二、消息幂等
1、在什么情况下会发生RocketMQ的消息重复消费
①、当系统的调用链路比较长的时候,比如系统A调用系统B,系统B再把消息发送到RocketMQ中,在系统A调用系统B的时候,如果系统B处理成功,但是迟迟没有将调用成功的结果返回给系统A的时候,系统A就会尝试重新发起请求给系统B,造成系统B重复处理,发起多条消息给RocketMQ造成重复消费;
②、在系统B发送给RocketMQ的时候,也有可能会发生和上面一样的问题,消息发送超时,节骨系统B重试,导致RocketMQ接收到了重读消息;
③、当RocketMQ成功接收到消息,并将消息交给消费者处理,如果消费者消费完成后还没来得及提交offset给RocketMQ,自己宕机或者重启了,那么RocketMQ没有接收到offset,就会认为消费失败了,会重发消息给消费者再次消费;
2、如何解决消息的重复消费
通过幂等性来保证,只要保证重复消息不对结果产生影响,就完美地解决这个问题。
在生产者端保证幂等性,一下两种方式:
①、RocketMQ支持消息查询的功能,只要去RocketMQ查询一下是否已经发送过该条消息就可以了,不存在则发送,存在则不发送;
②、引入Redis,在发送消息到RocketMQ成功之后,向Redis中插入一条数据,如果发送重试,则先去Redis查询一个该条消息是否已经发送过了,存在的话就不重复发送消息了;
方法一:RocketMQ消息查询的性能不是特别好,如果在高并发的场景下,每条消息在发送到RocketMQ时都去查询一下,可能会影响接口的性能;
方法二:在一些极端的场景下,Redis也无法保证消息发送成功之后,就一定能写入Redis成功,比如写入消息成功而Redis此时宕机,那么再次查询Redis判断消息是否已经发送过,是无法得到正确结果的;
3、生产者
package com.zn.idempotent; import com.alibaba.rocketmq.client.exception.MQBrokerException; import com.alibaba.rocketmq.client.exception.MQClientException; import com.alibaba.rocketmq.client.producer.DefaultMQProducer; import com.alibaba.rocketmq.client.producer.SendResult; import com.alibaba.rocketmq.common.message.Message; import com.alibaba.rocketmq.remoting.exception.RemotingException; /** * 消息幂等生产者 */ public class IdempotentProvider { public static void main(String[] args) throws MQClientException, InterruptedException, RemotingException, MQBrokerException { //创建一个生产者 DefaultMQProducer producer=new DefaultMQProducer("rmq-group"); //设置NameServer地址 producer.setNamesrvAddr("192.168.33.135:9876;192.168.33.136:9876"); //设置生产者实例名称 producer.setInstanceName("producer"); //启动生产者 producer.start(); //发送消息 for (int i=1;i<=1;i++){ //模拟网络延迟,每秒发送一次MQ Thread.sleep(1000); //创建消息,topic主题名称 tags临时值代表小分类, body代表消息体 Message message=new Message("itmayiedu-topic03","TagA",("itmayiedu-"+i).getBytes()); //消息的唯一标识 message.setKeys("订单消息:"+i); //发送消息 SendResult sendResult=producer.send(message); System.out.println("信息幂等问题来了:"+sendResult.toString()); } producer.shutdown(); } }
4、消费者
package com.zn.idempotent; import com.alibaba.rocketmq.client.consumer.DefaultMQPushConsumer; import com.alibaba.rocketmq.client.consumer.listener.ConsumeConcurrentlyContext; import com.alibaba.rocketmq.client.consumer.listener.ConsumeConcurrentlyStatus; import com.alibaba.rocketmq.client.consumer.listener.MessageListenerConcurrently; import com.alibaba.rocketmq.client.exception.MQClientException; import com.alibaba.rocketmq.common.message.MessageExt; import java.util.HashMap; import java.util.List; import java.util.Map; import java.util.logging.LogManager; import java.util.logging.Logger; /** * 消息幂等消费者 */ public class IdempotentConsumer { static private Map<String, Object> logMap = new HashMap<>(); public static void main(String[] args) throws MQClientException { //创建消费者 DefaultMQPushConsumer consumer=new DefaultMQPushConsumer("rmq-group"); //设置NameServer地址 consumer.setNamesrvAddr("192.168.33.135:9876;192.168.33.136:9876"); //设置实例名称 consumer.setInstanceName("consumer"); //订阅topic consumer.subscribe("itmayiedu-topic03","TagA"); //监听消息 consumer.registerMessageListener(new MessageListenerConcurrently() { @Override public ConsumeConcurrentlyStatus consumeMessage(List<MessageExt> list, ConsumeConcurrentlyContext consumeConcurrentlyContext) { String key=null; String msgId=null; for (MessageExt messageExt:list){ key=messageExt.getKeys(); //判读redis中有没有当前消息key if (logMap.containsKey(key)) { // 无需继续重试。 System.out.println("key:"+key+",已经消费,无需重试..."); return ConsumeConcurrentlyStatus.CONSUME_SUCCESS; } //RocketMQ由于是集群环境,所以产生的消息ID可能会重复 msgId = messageExt.getMsgId(); System.out.println("key:" + key + ",msgid:" + msgId + "---" + new String(messageExt.getBody())); //将当前key保存在redis中 logMap.put(messageExt.getKeys(),messageExt); } try { int i=5/0; }catch (Exception e){ e.printStackTrace(); //人工补偿 return ConsumeConcurrentlyStatus.RECONSUME_LATER; } return ConsumeConcurrentlyStatus.CONSUME_SUCCESS; } }); //启动消费者 consumer.start(); System.out.println("Consumer Started!"); } }