• 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!");
        }
    }
     
  • 相关阅读:
    二叉树——Java实现
    Java实现单链表的增删查改及逆置打印
    常见排序——Java实现
    [导入]Interesting Finds: 2007.12.10
    [导入]Interesting Finds: 2007.12.09
    [导入]Interesting Finds: 2007.12.12
    [导入]Interesting Finds: 2007.12.06
    [导入]Interesting Finds: 2007.12.07
    [导入]Interesting Finds: 2007.12.04
    [导入]Interesting Finds: 2007.12.05
  • 原文地址:https://www.cnblogs.com/lowerma/p/12327612.html
Copyright © 2020-2023  润新知