RocketMQ事务消息学习及刨坑过程

一、背景

MQ组件是系统架构里必不可少的一门利器,设计层面可以降低系统耦合度,高并发场景又可以起到削峰填谷的作用,从单体应用到集群部署方案,再到现在的微服务架构,MQ凭借其优秀的性能和高可靠性,得到了广泛的认可。
随着数据量增多,系统压力变大,开始出现这种现象:数据库已经更新了,但消息没发出来,或者消息先发了,但后来数据库更新失败了,结果研发童鞋各种数据修复,这种生产问题出现的概率不大,但让人很郁闷。这个其实就是数据库事务与MQ消息的一致性问题,简单来讲,数据库的事务跟普通MQ消息发送无法直接绑定与数据库事务绑定在一起,例如上面提及的两种问题场景:

  1. 数据库事务提交后发送MQ消息;
  2. MQ消息先发,然后再提交数据库事务。

场景1的问题是数据库事务可能刚刚提交,服务器就宕机了,MQ消息没发出去,场景2的问题就是MQ消息发送出去了,但数据库事务提交失败,又没办法追加已经发出去的MQ消息,结果导致数据没更新,下游已经收到消息,最终事务出现不一致的情况。

二、事务消息的引出

我们以微服务架构的购物场景为例,参照一下RocketMQ官方的例子,用户A发起订单,支付100块钱操作完成后,能得到100积分,账户服务和会员服务是两个独立的微服务模块,有各自的数据库,按照上文提及的问题可能性,将会出现这些情况:

  • 如果先扣款,再发消息,可能钱刚扣完,宕机了,消息没发出去,结果积分没增加。
  • 如果先发消息,再扣款,可能积分增加了,但钱没扣掉,白送了100积分给人家。
  • 钱正常扣了,消息也发送成功了,但会员服务实例消费消息出现问题,结果积分没增加。
    购物场景MQ通信案例

由此引出的是数据库事务与MQ消息的事务一致性问题,rocketmq事务消息解决的问题:解决本地事务执行与消息发送的原子性问题。这里界限一定要明白,是确保MQ生产端正确无误地将消息发送出来,没有多发,也不会漏发。但至于发送后消费端有没有正常的消费掉(如上面提及的第三种情况,钱正常扣了,消息也发了,但下游消费出问题导致积分不对),这种异常场景将由MQ消息消费失败重试机制来保证,不在此次的讨论范围内。

常用的MQ组件针对此场景都有自己的实现方案,如ActiveMQ使用AMQP协议(二阶提交方式)保证消息正确发送,这里我们以RocketMQ为重点进行学习。

三、RocketMQ事务消息设计思路

根据CAP理论,RocketMQ事务消息通过异步确保方式,保证事务的最终一致性。设计流程上借鉴两阶段提交理论,流程图如下:
RocetMQ事务消息设计图

  1. 应用模块遇到要发送事务消息的场景时,先发送prepare消息给MQ。
  2. prepare消息发送成功后,应用模块执行数据库事务(本地事务)。
  3. 根据数据库事务执行的结果,再返回Commit或Rollback给MQ。
  4. 如果是Commit,MQ把消息下发给Consumer端,如果是Rollback,直接删掉prepare消息。
  5. 第3步的执行结果如果没响应,或是超时的,启动定时任务回查事务状态(最多重试15次,超过了默认丢弃此消息),处理结果同第4步。
  6. MQ消费的成功机制由MQ自己保证。

四、RocketMQ事务消息实现流程

以RocketMQ 4.5.2版本为例,事务消息有专门的一个队列RMQ_SYS_TRANS_HALF_TOPIC,所有的prepare消息都先往这里放,当消息收到Commit请求后,就把消息再塞到真实的Topic队列里,供Consumer消费,同时向RMQ_SYS_TRANS_OP_HALF_TOPIC塞一条消息。简易流程图如下:
RocketMQ事务消息实现流程

上述流程中,请允许我这样划分模块职责:

  1. RocketMQ Client即我们工程中导入的依赖jar包,RocketMQ Broker端即部署的服务端,NameServer暂未体现。
  2. 应用模块成对出现,上游为事务消息生产端,下游为事务消息消费端(事务消息对消费端是透明的,与普通消息一致)。

应用模块的事务因为中断,或是其他的网络原因,导致无法立即响应的,RocketMQ当做UNKNOW处理,RocketMQ事务消息还提供了一个补救方案:定时查询事务消息的数据库事务状态
简易流程图如下:
RocketMQ定时任务回查事务状态实现流程

五、源码剖析

讲解的思路基本上按照如下流程图,根据模块职责和流程逐一分析。

  1. 环境准备
    阅读源码前需要在IDE上获取和调试RocketMQ的源码,这部分请自行查阅方法。

  2. 应用模块(事务消息生产端)核心源码
    创建一个监听类,实现TransactionListener接口,在实现的数据库事务提交方法和回查事务状态方法模拟结果。

/**
 * @program: rocket
 * @description: 调试事务消息示例代码
 * @author: Huang
 * @create: 2019-10-16
 **/
public class SelfTransactionListener implements TransactionListener {
   private AtomicInteger transactionIndex = new AtomicInteger(0);
   private AtomicInteger checkTimes = new AtomicInteger(0);

   private ConcurrentHashMap<String, Integer> localTrans = new ConcurrentHashMap<>();
   /**
    * 执行本地事务
    *
    * @param message
    * @param o
    * @return
    */
   @Override
   public LocalTransactionState executeLocalTransaction(Message message, Object o) {
      String msgKey = message.getKeys();
      System.out.println("start execute local transaction " + msgKey);
      LocalTransactionState state;
      if (msgKey.contains("1")) {
         // 第一条消息让他通过
         state = LocalTransactionState.COMMIT_MESSAGE;
      } else if (msgKey.contains("2")) {
         // 第二条消息模拟异常,明确回复回滚操作
         state = LocalTransactionState.ROLLBACK_MESSAGE;
      } else {
         // 第三条消息无响应,让它调用回查事务方法
         state = LocalTransactionState.UNKNOW;
         // 给剩下3条消息,放1,2,3三种状态
         localTrans.put(msgKey, transactionIndex.incrementAndGet());
      }
      System.out.println("executeLocalTransaction:" + message.getKeys() + ",execute state:" + state + ",current time:" + System.currentTimeMillis());
      return state;
   }

   /**
    * 回查本地事务结果
    *
    * @param messageExt
    * @return
    */
   @Override
   public LocalTransactionState checkLocalTransaction(MessageExt messageExt) {
      String msgKey = messageExt.getKeys();
      System.out.println("start check local transaction " + msgKey);
      Integer state = localTrans.get(msgKey);
      switch (state) {
         case 1:
            System.out.println("check result unknown 回查次数" + checkTimes.incrementAndGet());
            return LocalTransactionState.UNKNOW;
         case 2:
            System.out.println("check result commit message, 回查次数" + checkTimes.incrementAndGet());
            return LocalTransactionState.COMMIT_MESSAGE;
         case 3:
            System.out.println("check result rollback message, 回查次数" + checkTimes.incrementAndGet());
            return LocalTransactionState.ROLLBACK_MESSAGE;

         default:
            return LocalTransactionState.COMMIT_MESSAGE;
      }
   }
}

事务消息生产者代码示例,共发送5条消息,基本上包含全部的场景,休眠时间设置足够的时间,保证回查事务时实例还在运行中,代码如下:

/**
 * @program: rocket
 * @description: Rocketmq事务消息
 * @author: Huang
 * @create: 2019-10-16
 **/
public class TransactionProducer {

   public static void main(String[] args) {
      try {
         TransactionMQProducer producer = new TransactionMQProducer("transactionMQProducer");
         producer.setNamesrvAddr("10.0.133.29:9876");
         producer.setTransactionListener(new SelfTransactionListener());
         producer.start();
         for (int i = 1; i < 6; i++) {
            Message message = new Message("TransactionTopic", "transactionTest","msg-" + i, ("Hello" + ":" +  i).getBytes());
            try {
               SendResult result = producer.sendMessageInTransaction(message, "Hello" + ":" +  i);
               System.out.printf("Topic:%s send success, misId is:%s%n", message.getTopic(), result.getMsgId());
            } catch (Exception e) {
               e.printStackTrace();
            }
         }
         Thread.sleep(Integer.MAX_VALUE);
         producer.shutdown();
      } catch (MQClientException e) {
         e.printStackTrace();
      } catch (InterruptedException e) {
         e.printStackTrace();
      }
   }
}
  1. RocketMQ Client端代码,代码主要逻辑可以分成三段:第一段为设置消息为prepare消息,并发送给RocketMQ服务端
SendResult sendResult = null;
MessageAccessor.putProperty(msg, MessageConst.PROPERTY_TRANSACTION_PREPARED, "true");
MessageAccessor.putProperty(msg, MessageConst.PROPERTY_PRODUCER_GROUP, this.defaultMQProducer.getProducerGroup());
try {
    sendResult = this.send(msg);
} catch (Exception e) {
    throw new MQClientException("send message Exception", e);
}

第二段:消息发送成功后,调用应用模块数据库事务方法,获取事务结果(为节省篇幅,代码有删节)

switch (sendResult.getSendStatus()) {
    case SEND_OK: {
        try {
            if (null != localTransactionExecuter) {
                localTransactionState = localTransactionExecuter.executeLocalTransactionBranch(msg, arg);
            } else if (transactionListener != null) {
                log.debug("Used new transaction API");
                localTransactionState = transactionListener.executeLocalTransaction(msg, arg);
            }
            if (null == localTransactionState) {
                localTransactionState = LocalTransactionState.UNKNOW;
            }
        } catch (Throwable e) {
            log.info("executeLocalTransactionBranch exception", e);
            log.info(msg.toString());
            localException = e;
        }
    }
    break;
    case FLUSH_DISK_TIMEOUT:
    case FLUSH_SLAVE_TIMEOUT:
    case SLAVE_NOT_AVAILABLE:
        localTransactionState = LocalTransactionState.ROLLBACK_MESSAGE;
        break;
    default:
        break;
}

第三段:发送事务结果到RocketMQ端,结束事务,并响应结果给应用模块

try {
    this.endTransaction(sendResult, localTransactionState, localException);
} catch (Exception e) {
    log.warn("local transaction execute " + localTransactionState + ", but end broker transaction failed", e);
}
  1. RocketMQ Broker端事务提交/回滚操作(这里取endTransaction部分)
    代码入口:org.apache.rocketmq.broker.processor.EndTransactionProcessor
OperationResult result = new OperationResult();
if (MessageSysFlag.TRANSACTION_COMMIT_TYPE == requestHeader.getCommitOrRollback()) {
    result = this.brokerController.getTransactionalMessageService().commitMessage(requestHeader);
    if (result.getResponseCode() == ResponseCode.SUCCESS) {
        RemotingCommand res = checkPrepareMessage(result.getPrepareMessage(), requestHeader);
        if (res.getCode() == ResponseCode.SUCCESS) {
            // 修改消息的Topic为由RMQ_SYS_TRANS_HALF_TOPIC改为真实Topic
            MessageExtBrokerInner msgInner = endMessageTransaction(result.getPrepareMessage());
            msgInner.setSysFlag(MessageSysFlag.resetTransactionValue(msgInner.getSysFlag(), requestHeader.getCommitOrRollback()));
            msgInner.setQueueOffset(requestHeader.getTranStateTableOffset());
            msgInner.setPreparedTransactionOffset(requestHeader.getCommitLogOffset());
            msgInner.setStoreTimestamp(result.getPrepareMessage().getStoreTimestamp());
            // 将消息存储到真实Topic中,供Consumer消费
            RemotingCommand sendResult = sendFinalMessage(msgInner);
            if (sendResult.getCode() == ResponseCode.SUCCESS) {
                // 将消息存储到RMQ_SYS_TRANS_OP_HALF_TOPIC,标记为删除状态,事务消息回查的定时任务中会做处理
                this.brokerController.getTransactionalMessageService().deletePrepareMessage(result.getPrepareMessage());
            }
            return sendResult;
        }
        return res;
    }
} else if (MessageSysFlag.TRANSACTION_ROLLBACK_TYPE == requestHeader.getCommitOrRollback()) {
    result = this.brokerController.getTransactionalMessageService().rollbackMessage(requestHeader);
    if (result.getResponseCode() == ResponseCode.SUCCESS) {
        RemotingCommand res = checkPrepareMessage(result.getPrepareMessage(), requestHeader);
        if (res.getCode() == ResponseCode.SUCCESS) {
            this.brokerController.getTransactionalMessageService().deletePrepareMessage(result.getPrepareMessage());
        }
        return res;
    }
}
  1. RocketMQ Broker端定时任务回查数据库事务部分
    方法入口:org.apache.rocketmq.broker.transaction.TransactionalMessageCheckService
@Override
protected void onWaitEnd() {
    long timeout = brokerController.getBrokerConfig().getTransactionTimeOut();
    // 超过15次的回查事务状态失败后,默认是丢弃此消息
    int checkMax = brokerController.getBrokerConfig().getTransactionCheckMax();
    long begin = System.currentTimeMillis();
    log.info("Begin to check prepare message, begin time:{}", begin);
    this.brokerController.getTransactionalMessageService().check(timeout, checkMax, this.brokerController.getTransactionalMessageCheckListener());
    log.info("End to check prepare message, consumed time:{}", System.currentTimeMillis() - begin);
}

回查事务调用入口:

// 此段代码为TransactionalMessageServiceImpl类中的check方法
List<MessageExt> opMsg = pullResult.getMsgFoundList();
boolean isNeedCheck = (opMsg == null && valueOfCurrentMinusBorn > checkImmunityTime)
    || (opMsg != null && (opMsg.get(opMsg.size() - 1).getBornTimestamp() - startTime > transactionTimeout))
    || (valueOfCurrentMinusBorn <= -1
);

if (isNeedCheck) {
    if (!putBackHalfMsgQueue(msgExt, i)) {
        continue;
    }
    // 调用AbstractTransactionalMessageCheckListener的
    listener.resolveHalfMsg(msgExt);
} else {
    pullResult = fillOpRemoveMap(removeMap, opQueue, pullResult.getNextBeginOffset(), halfOffset, doneOpOffset);
    log.info("The miss offset:{} in messageQueue:{} need to get more opMsg, result is:{}", i,
        messageQueue, pullResult);
    continue;
}

// 此方法在AbstractTransactionalMessageCheckListener类中
public void resolveHalfMsg(final MessageExt msgExt) {
    executorService.execute(new Runnable() {
        @Override
        public void run() {
            try {
                sendCheckMessage(msgExt);
            } catch (Exception e) {
                LOGGER.error("Send check message error!", e);
            }
        }
    });
}

// 此方法在AbstractTransactionalMessageCheckListener类中
public void sendCheckMessage(MessageExt msgExt) throws Exception {
    CheckTransactionStateRequestHeader checkTransactionStateRequestHeader = new CheckTransactionStateRequestHeader();
    checkTransactionStateRequestHeader.setCommitLogOffset(msgExt.getCommitLogOffset());
    checkTransactionStateRequestHeader.setOffsetMsgId(msgExt.getMsgId());
    checkTransactionStateRequestHeader.setMsgId(msgExt.getUserProperty(MessageConst.PROPERTY_UNIQ_CLIENT_MESSAGE_ID_KEYIDX));
    checkTransactionStateRequestHeader.setTransactionId(checkTransactionStateRequestHeader.getMsgId());
    checkTransactionStateRequestHeader.setTranStateTableOffset(msgExt.getQueueOffset());
    msgExt.setTopic(msgExt.getUserProperty(MessageConst.PROPERTY_REAL_TOPIC));
    msgExt.setQueueId(Integer.parseInt(msgExt.getUserProperty(MessageConst.PROPERTY_REAL_QUEUE_ID)));
    msgExt.setStoreSize(0);
    String groupId = msgExt.getProperty(MessageConst.PROPERTY_PRODUCER_GROUP);
    Channel channel = brokerController.getProducerManager().getAvaliableChannel(groupId);
    if (channel != null) {
        // 通过Netty发送请求到RocketMQ Client端,执行checkTransactionState方法
        brokerController.getBroker2Client().checkProducerTransactionState(groupId, channel, checkTransactionStateRequestHeader, msgExt);
    } else {
        LOGGER.warn("Check transaction failed, channel is null. groupId={}", groupId);
    }
}

RocketMQ Client接收到服务端的请求后,重新调用回查数据库事务方法,并将事务结果再次提交到RocketMQ Broker端
方法入口:org.apache.rocketmq.client.impl.producer.DefaultMQProducerImpl类的方法

try {
    if (transactionCheckListener != null) {
        localTransactionState = transactionCheckListener.checkLocalTransactionState(message);
    } else if (transactionListener != null) {
        log.debug("Used new check API in transaction message");
        localTransactionState = transactionListener.checkLocalTransaction(message);
    } else {
        log.warn("CheckTransactionState, pick transactionListener by group[{}] failed", group);
    }
} catch (Throwable e) {
    log.error("Broker call checkTransactionState, but checkLocalTransactionState exception", e);
    exception = e;
}

this.processTransactionState(
    localTransactionState,
    group,
    exception);

六、补充一个问题

官网有提及,事务消息是不支持延迟消息和批量消息,我手贱试了一下延迟消息,事务消息设置一个DelayTimeLevel,结果这条消息就一直无法从RMQ_SYS_TRANS_HALF_TOPIC移除掉了,应用模块的日志发现在反复地尝试回查事务,Console界面上RMQ_SYS_TRANS_HALF_TOPIC的消息查询列表很快就超过2000条记录了,为什么?

我们回到代码层面进行分析,过程如下:

1.设置了DelayTimeLevel后,数据事务提交后(或是回查数据库事务完成后),将消息写入目标Topic时,由于DelayTimeLevel的干扰,目标Topic将变成SCHEDULE_TOPIC_XXXX,同时REAL_TOPIC变成RMQ_SYS_TRANS_HALF_TOPIC,真实的Topic在这个环节已经丢失。

// RocketMQ Broker端接受事务提交后的处理
org.apache.rocketmq.broker.processor.EndTransactionProcessor类
OperationResult result = new OperationResult();
if (MessageSysFlag.TRANSACTION_COMMIT_TYPE == requestHeader.getCommitOrRollback()) {
    // 这里调用CommitLog的putMessage方法
    result = this.brokerController.getTransactionalMessageService().commitMessage(requestHeader);
    if (result.getResponseCode() == ResponseCode.SUCCESS) {
        RemotingCommand res = checkPrepareMessage(result.getPrepareMessage(), requestHeader);
        if (res.getCode() == ResponseCode.SUCCESS) {
            // 修改消息的Topic为由RMQ_SYS_TRANS_HALF_TOPIC改为真实Topic
            MessageExtBrokerInner msgInner = endMessageTransaction(result.getPrepareMessage());
            msgInner.setSysFlag(MessageSysFlag.resetTransactionValue(msgInner.getSysFlag(), requestHeader.getCommitOrRollback()));
            msgInner.setQueueOffset(requestHeader.getTranStateTableOffset());
            msgInner.setPreparedTransactionOffset(requestHeader.getCommitLogOffset());
            msgInner.setStoreTimestamp(result.getPrepareMessage().getStoreTimestamp());
            // 将消息存储到真实Topic中,此时Topic已经变成SCHEDULE_TOPIC_XXXX
            RemotingCommand sendResult = sendFinalMessage(msgInner);
            if (sendResult.getCode() == ResponseCode.SUCCESS) {
                // 将消息存储到RMQ_SYS_TRANS_OP_HALF_TOPIC,标记为删除状态,事务消息回查的定时任务中会做处理
                this.brokerController.getTransactionalMessageService().deletePrepareMessage(result.getPrepareMessage());
            }
            return sendResult;
        }
        return res;
    }
}

// 此段代码在org.apache.rocketmq.store.CommitLog类的putMessage方法中
// 由于DelayTimeLevel的干扰,目标Topic将变成SCHEDULE_TOPIC_XXXX
final int tranType = MessageSysFlag.getTransactionValue(msg.getSysFlag());
if (tranType == MessageSysFlag.TRANSACTION_NOT_TYPE
    || tranType == MessageSysFlag.TRANSACTION_COMMIT_TYPE) {
    // Delay Delivery
    if (msg.getDelayTimeLevel() > 0) {
        if (msg.getDelayTimeLevel() > this.defaultMessageStore.getScheduleMessageService().getMaxDelayLevel()) {
            msg.setDelayTimeLevel(this.defaultMessageStore.getScheduleMessageService().getMaxDelayLevel());
        }

        topic = ScheduleMessageService.SCHEDULE_TOPIC;
        queueId = ScheduleMessageService.delayLevel2QueueId(msg.getDelayTimeLevel());

        // Backup real topic, queueId
        MessageAccessor.putProperty(msg, MessageConst.PROPERTY_REAL_TOPIC, msg.getTopic());
        MessageAccessor.putProperty(msg, MessageConst.PROPERTY_REAL_QUEUE_ID, String.valueOf(msg.getQueueId()));
        msg.setPropertiesString(MessageDecoder.messageProperties2String(msg.getProperties()));

        msg.setTopic(topic);
        msg.setQueueId(queueId);
    }
}

打印的日志示例如下:

2019-10-17 14\:41\:05 INFO EndTransactionThread_4 - Transaction op message write successfully. messageId=0A00851D00002A9F0000000000000E09, queueId=0 
msgExt:MessageExt [queueId=0, storeSize=335, queueOffset=5, sysFlag=8, bornTimestamp=1571293959305, bornHost=/10.0.133.29:54634, storeTimestamp=1571294460555, 
storeHost=/10.0.133.29:10911, msgId=0A00851D00002A9F0000000000000E09, commitLogOffset=3593, bodyCRC=1849408413, reconsumeTimes=0, preparedTransactionOffset=0, 
toString()=Message{topic='SCHEDULE_TOPIC_XXXX', flag=0, properties={REAL_TOPIC=RMQ_SYS_TRANS_HALF_TOPIC, TRANSACTION_CHECK_TIMES=3, KEYS=msg-test-3, 
TRAN_MSG=true, UNIQ_KEY=0A00851D422C18B4AAC25584B0880000, WAIT=false, DELAY=1, PGROUP=transactionMQProducer, TAGS=transactionTest, REAL_QID=0}, 
body=[72, 101, 108, 108, 111, 84, 105, 109, 101, 58, 51], transactionId='null'}]

2.延迟消息是定时任务触发的,我刚刚设置的延迟是1秒,定时任务又把消息重新放回RMQ_SYS_TRANS_HALF_TOPIC中,注意此时只有RMQ_SYS_TRANS_HALF_TOPIC有消息,RMQ_SYS_TRANS_OP_HALF_TOPIC队列是没有这条消息的,如下代码:

// 此段代码在org.apache.rocketmq.store.schedule.ScheduleMessageService类executeOnTimeup方法内
try {
    // 消息重新回到RMQ_SYS_TRANS_HALF_TOPIC队列中
    MessageExtBrokerInner msgInner = this.messageTimeup(msgExt);
    PutMessageResult putMessageResult =
        ScheduleMessageService.this.writeMessageStore
            .putMessage(msgInner);

    if (putMessageResult != null
        && putMessageResult.getPutMessageStatus() == PutMessageStatus.PUT_OK) {
        continue;
    } else {
        log.error(
            "ScheduleMessageService, a message time up, but reput it failed, topic: {} msgId {}",
            msgExt.getTopic(), msgExt.getMsgId());
        ScheduleMessageService.this.timer.schedule(
            new DeliverDelayedMessageTimerTask(this.delayLevel,
                nextOffset), DELAY_FOR_A_PERIOD);
        ScheduleMessageService.this.updateOffset(this.delayLevel,
            nextOffset);
        return;
    }
} catch (Exception e) {
    log.error(
        "ScheduleMessageService, messageTimeup execute error, drop it. msgExt="
            + msgExt + ", nextOffset=" + nextOffset + ",offsetPy="
            + offsetPy + ",sizePy=" + sizePy, e);
}

3.事务消息定时任务启动,查RMQ_SYS_TRANS_HALF_TOPIC有消息,但RMQ_SYS_TRANS_OP_HALF_TOPIC没有消息,为了保证消息顺序写入,又将此消息重新填入RMQ_SYS_TRANS_OP_HALF_TOPIC中,并且触发一次回查事务操作。示例代码如上文回查事务调用入口相同:

// 此段代码为TransactionalMessageServiceImpl类中的check方法
List<MessageExt> opMsg = pullResult.getMsgFoundList();
boolean isNeedCheck = (opMsg == null && valueOfCurrentMinusBorn > checkImmunityTime)
    || (opMsg != null && (opMsg.get(opMsg.size() - 1).getBornTimestamp() - startTime > transactionTimeout))
    || (valueOfCurrentMinusBorn <= -1
);

if (isNeedCheck) {
    if (!putBackHalfMsgQueue(msgExt, i)) {
        continue;
    }
    listener.resolveHalfMsg(msgExt);
} else {
    pullResult = fillOpRemoveMap(removeMap, opQueue, pullResult.getNextBeginOffset(), halfOffset, doneOpOffset);
    log.info("The miss offset:{} in messageQueue:{} need to get more opMsg, result is:{}", i,
        messageQueue, pullResult);
    continue;
}

这样构成了一个死循环,直到尝试到15次才丢弃此消息(默认最大尝试次数是15次),这个代价有点大。针对此问题的优化,已经提交PR到RocketMQ社区,新版本发布后,事务消息将屏蔽DelayTimeLevel,这个问题就不会再出现了。

在新版本发布之前,我们的解决办法:

  1. 明确研发过程中事务消息禁止设置DelayTimeLevel。
    感觉有风险,毕竟新来的童鞋,不是特别了解此部分功能的可能会手抖加上(像我最早那样)。
  2. 对RocketMQ Client做一次简单的封装,比如提供一个rocketmq-spring-boot-starter,在提供发送事务消息的方法里不提供设置的入口,如下示例:
/**
 * 事务消息发送
 * 不支持延迟发送和批量发送
 */
public void sendMessageInTransaction(String topic, String tag, Object message, String requestId) throws Exception {
   TransactionMQProducer producer = annotationScan.getProducer(topic + "_" + tag);
   producer.sendMessageInTransaction(MessageBuilder.of(topic, tag, message, requestId).build(), message);
}

应该靠谱一些,毕竟从源头杜绝了DelayTimeLevel参数的设置。

七、结束语

本篇简单介绍了事务消息的解决的场景和职责的界限,基本的设计思路和流程,在此借鉴学习了RocketMQ作者的图稿,然后挑了部分代码作简要的讲解,还是自己的刨坑过程,文章内有任何不正确或不详尽之处请留言指导,谢谢。

posted @ 2020-04-26 14:15  大数据从业者FelixZh  阅读(742)  评论(0编辑  收藏  举报