RocketMQ 5.x延时消息源码分析(不包含时间轮)

这里是weihubeats ,觉得文章不错可以关注公众号小奏技术，文章首发。拒绝营销号，拒绝标题党

RocketMQ版本

5.1.0

背景

首先说明本次源码分析仅分析时间轮之前的延时消息设计

现在的RocketMQ已经支持基于时间轮的任意级别延时消息

延时消息基础知识

默认RocketMQ延时消息共有18个等级

1s 5s 10s 30s 1m 2m 3m 4m 5m 6m 7m 8m 9m 10m 20m 30m 1h 2h

延时消息的简单使用

发送延时消息

java 复制代码

        DefaultMQProducer producer = new DefaultMQProducer(producerGroup);
        producer.setNamesrvAddr(namesrvAddr);
        try {
            producer.start();
        } catch (MQClientException e) {
            throw new RuntimeException(e);
        }

        Message msg = new Message(TOPIC /* Topic */,
                TAG /* Tag */,
                ("Hello RocketMQ " + i).getBytes(RemotingHelper.DEFAULT_CHARSET) /* Message body */
        );

        msg.setDelayTimeLevel(2);
        producer.send(msg);

和普通消息不同的我们新增了延时消息等级的设置

java 复制代码

        msg.setDelayTimeLevel(2);

源码分析

消息发送本身和普通消息发送没有太大差别，但是新增了一个DELAY属性

java 复制代码

public static final String PROPERTY_DELAY_TIME_LEVEL = "DELAY";

消息发送的请求code是

java 复制代码

public static final int SEND_BATCH_MESSAGE = 320;

源码入口

这里的我们先从消息的发送开始

消息发送最终的请求状态码是

java 复制代码

public static final int SEND_MESSAGE_V2 = 310;

可以看到实际消息发送和普通消息没有什么区别，所以我们还是要去broker那边看看有没有什么特殊处理

broker处理请求

消息处理方法 org.apache.rocketmq.broker.processor.SendMessageProcessor#processRequest

实际的消息发送逻辑在方法

java 复制代码

this.sendMessage(ctx, request, sendMessageContext, requestHeader, mappingContext,
                        (ctx12, response12) -> executeSendMessageHookAfter(response12, ctx12));

这里我们进去这个方法看看

由于是分析延时消息，所以我们不关注消息发送的一些其他细节。主要关注延时消息和普通消息的处理区别我们看方法

java 复制代码

org.apache.rocketmq.store.DefaultMessageStore#asyncPutMessage

这里可以看到现在的topic还是我们正常指定的xiao-zou-topic

但是这里会执行多个消息的后置处理器putMessageHookList

我们看看putMessageHookList有哪些

java 复制代码

    public void registerMessageStoreHook() {
        List<PutMessageHook> putMessageHookList = messageStore.getPutMessageHookList();

        putMessageHookList.add(new PutMessageHook() {
            @Override
            public String hookName() {
                return "checkBeforePutMessage";
            }

            @Override
            public PutMessageResult executeBeforePutMessage(MessageExt msg) {
                return HookUtils.checkBeforePutMessage(BrokerController.this, msg);
            }
        });

        putMessageHookList.add(new PutMessageHook() {
            @Override
            public String hookName() {
                return "innerBatchChecker";
            }

            @Override
            public PutMessageResult executeBeforePutMessage(MessageExt msg) {
                if (msg instanceof MessageExtBrokerInner) {
                    return HookUtils.checkInnerBatch(BrokerController.this, msg);
                }
                return null;
            }
        });

        putMessageHookList.add(new PutMessageHook() {
            @Override
            public String hookName() {
                return "handleScheduleMessage";
            }

            @Override
            public PutMessageResult executeBeforePutMessage(MessageExt msg) {
                if (msg instanceof MessageExtBrokerInner) {
                    return HookUtils.handleScheduleMessage(BrokerController.this, (MessageExtBrokerInner) msg);
                }
                return null;
            }
        });

        SendMessageBackHook sendMessageBackHook = new SendMessageBackHook() {
            @Override
            public boolean executeSendMessageBack(List<MessageExt> msgList, String brokerName, String brokerAddr) {
                return HookUtils.sendMessageBack(BrokerController.this, msgList, brokerName, brokerAddr);
            }
        };

        if (messageStore != null) {
            messageStore.setSendMessageBackHook(sendMessageBackHook);
        }
    }

这里可以看到主要是三个

checkBeforePutMessage
innerBatchChecker
handleScheduleMessage

通过方法名我们可以很确定的看到主要是handleScheduleMessage这个方法出处理延时消息，所以我们重点看看handleScheduleMessage

handleScheduleMessage

org.apache.rocketmq.broker.util.HookUtils#handleScheduleMessage

java 复制代码

 public static PutMessageResult handleScheduleMessage(BrokerController brokerController,
        final MessageExtBrokerInner msg) {
        final int tranType = MessageSysFlag.getTransactionValue(msg.getSysFlag());
        // 事务消息 暂时不管
        if (tranType == MessageSysFlag.TRANSACTION_NOT_TYPE
            || tranType == MessageSysFlag.TRANSACTION_COMMIT_TYPE) {
            if (!isRolledTimerMessage(msg)) {
                if (checkIfTimerMessage(msg)) {
                    if (!brokerController.getMessageStoreConfig().isTimerWheelEnable()) {
                        //wheel timer is not enabled, reject the message
                        return new PutMessageResult(PutMessageStatus.WHEEL_TIMER_NOT_ENABLE, null);
                    }
                    PutMessageResult transformRes = transformTimerMessage(brokerController, msg);
                    if (null != transformRes) {
                        return transformRes;
                    }
                }
            }
            // Delay Delivery 延时消息
            if (msg.getDelayTimeLevel() > 0) {
                transformDelayLevelMessage(brokerController, msg);
            }
        }
        return null;
    }

这里的核心逻辑还是

java 复制代码

if (msg.getDelayTimeLevel() > 0) {
                transformDelayLevelMessage(brokerController, msg);
            }

我们看看transformDelayLevelMessage方法

java 复制代码

public static void transformDelayLevelMessage(BrokerController brokerController, MessageExtBrokerInner msg) {
        // 判断是否超过最大延时级别18 如果超过则设置为最大延时等级
        if (msg.getDelayTimeLevel() > brokerController.getScheduleMessageService().getMaxDelayLevel()) {
            msg.setDelayTimeLevel(brokerController.getScheduleMessageService().getMaxDelayLevel());
        }

        // Backup real topic, queueId 备份真实的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()));
        // 设置延时消息为SCHEDULE_TOPIC_XXXX 这个固定的topic
        msg.setTopic(TopicValidator.RMQ_SYS_SCHEDULE_TOPIC);    
        // 设置延时消息的queueID delayLevel - 1 
        msg.setQueueId(ScheduleMessageService.delayLevel2QueueId(msg.getDelayTimeLevel()));
    }

备份前的message信息

备份后

总结就是

将原始topic替换为延迟消息固定的topic:SCHEDULE_TOPIC_XXXX
将原始queueid替换为delayLevel - 1
备份原始topic/queueid, 保存到原始消息的properties属性中

延时消息消费

我们通过查看SCHEDULE_TOPIC_XXXX的调用发现有一个方法

org.apache.rocketmq.broker.schedule.ScheduleMessageService.DeliverDelayedMessageTimerTask#executeOnTimeup

我们查看调用链

发现很明显应该是使用了一个定时器去执行的，我们可以看看

这里可以看到把18个延时等级的任务都加进去了

deliverExecutorService的核心线程数也是18个

java 复制代码

            this.deliverExecutorService = new ScheduledThreadPoolExecutor(this.maxDelayLevel, new ThreadFactoryImpl("ScheduleMessageTimerThread_"));

我们这里看看丢进去的DeliverDelayedMessageTimerTask的任务里面里面的逻辑

org.apache.rocketmq.broker.schedule.ScheduleMessageService.DeliverDelayedMessageTimerTask#executeOnTimeup

核心逻辑都被封装在executeOnTimeup

我们进去看看executeOnTimeup

代码逻辑有点长我们，慢慢看

java 复制代码

public void executeOnTimeup() {
            // 根据延迟topic和延迟queueid 去获取Consumequeue
            ConsumeQueueInterface cq =
                ScheduleMessageService.this.brokerController.getMessageStore().getConsumeQueue(TopicValidator.RMQ_SYS_SCHEDULE_TOPIC,
                    delayLevel2QueueId(delayLevel));
            // 如果 Consumequeue 为空则新增一个 DeliverDelayedMessageTimerTask 丢到 deliverExecutorService再定时执行，延时时间默认为100毫秒
            // Consumequeue存在但是没有消费文件 一样延时100毫秒后继续重复执行
            if (cq == null) {
                this.scheduleNextTimerTask(this.offset, DELAY_FOR_A_WHILE);
                return;
            }

            ReferredIterator<CqUnit> bufferCQ = cq.iterateFrom(this.offset);
            if (bufferCQ == null) {
                long resetOffset;
                if ((resetOffset = cq.getMinOffsetInQueue()) > this.offset) {
                    log.error("schedule CQ offset invalid. offset={}, cqMinOffset={}, queueId={}",
                        this.offset, resetOffset, cq.getQueueId());
                } else if ((resetOffset = cq.getMaxOffsetInQueue()) < this.offset) {
                    log.error("schedule CQ offset invalid. offset={}, cqMaxOffset={}, queueId={}",
                        this.offset, resetOffset, cq.getQueueId());
                } else {
                    resetOffset = this.offset;
                }

                this.scheduleNextTimerTask(resetOffset, DELAY_FOR_A_WHILE);
                return;
            }

            long nextOffset = this.offset;
            try {
                while (bufferCQ.hasNext() && isStarted()) {
                    CqUnit cqUnit = bufferCQ.next();
                    long offsetPy = cqUnit.getPos();
                    int sizePy = cqUnit.getSize();
                    // 这里的 tagCode实际是投递时间
                    long tagsCode = cqUnit.getTagsCode();

                    if (!cqUnit.isTagsCodeValid()) {
                        //can't find ext content.So re compute tags code.
                        log.error("[BUG] can't find consume queue extend file content!addr={}, offsetPy={}, sizePy={}",
                            tagsCode, offsetPy, sizePy);
                        long msgStoreTime = ScheduleMessageService.this.brokerController.getMessageStore().getCommitLog().pickupStoreTimestamp(offsetPy, sizePy);
                        tagsCode = computeDeliverTimestamp(delayLevel, msgStoreTime);
                    }

                    long now = System.currentTimeMillis();
                    // 延时消息时间校验 如果超过deliverTimestamp now+delayLevel 时间则矫正为 now+delayLevel
                    long deliverTimestamp = this.correctDeliverTimestamp(now, tagsCode);

                    long currOffset = cqUnit.getQueueOffset();
                    assert cqUnit.getBatchNum() == 1;
                    nextOffset = currOffset + cqUnit.getBatchNum();

                    long countdown = deliverTimestamp - now;
                    // 如果延时时间超过当前时间，证明还未到消息处理时间
                    if (countdown > 0) {
                        this.scheduleNextTimerTask(currOffset, DELAY_FOR_A_WHILE);      
                        ScheduleMessageService.this.updateOffset(this.delayLevel, currOffset);
                        return;
                    }
                    // 加载出延时消息
                    MessageExt msgExt = ScheduleMessageService.this.brokerController.getMessageStore().lookMessageByOffset(offsetPy, sizePy);
                    if (msgExt == null) {
                        continue;
                    }
                    // 构建新的消息体，将原来的消息信息设置到这里，并将topic和queueid设置为原始的topic和queueid(前面备份过）
                    MessageExtBrokerInner msgInner = ScheduleMessageService.this.messageTimeup(msgExt);
                    if (TopicValidator.RMQ_SYS_TRANS_HALF_TOPIC.equals(msgInner.getTopic())) {
                        log.error("[BUG] the real topic of schedule msg is {}, discard the msg. msg={}",
                            msgInner.getTopic(), msgInner);
                        continue;
                    }

                    boolean deliverSuc;
                    if (ScheduleMessageService.this.enableAsyncDeliver) {
                        deliverSuc = this.asyncDeliver(msgInner, msgExt.getMsgId(), currOffset, offsetPy, sizePy);
                    } else {
                        // 将消息写入到 commitlog中
                        deliverSuc = this.syncDeliver(msgInner, msgExt.getMsgId(), currOffset, offsetPy, sizePy);
                    }

                    if (!deliverSuc) {
                        this.scheduleNextTimerTask(nextOffset, DELAY_FOR_A_WHILE);
                        return;
                    }
                }
            } catch (Exception e) {
                log.error("ScheduleMessageService, messageTimeup execute error, offset = {}", nextOffset, e);
            } finally {
                bufferCQ.release();
            }

            this.scheduleNextTimerTask(nextOffset, DELAY_FOR_A_WHILE);
        }

上面的代码我们将核心逻辑做了注释,我们总结一下大致流程

校验对应的延时等级消息是否存在ConsumeQueue，如果不存在则延时100ms继续轮训
校验ConsumeQueue中的索引是否存在，如果不存在则延时100ms继续轮训
校验索引里面的tagsCode 实际是投递时间是否到了需要投递的时间，如果是则取出延时消息
将延时消息还原为原始消息，投递到commitLog中继续消费

这里我们通过debug看看消息转换前后的参数

转换前的延时消息

转换后的原始消息

消息的重新投递是通过syncDeliver方法

java 复制代码

deliverSuc = this.syncDeliver(msgInner, msgExt.getMsgId(), currOffset, offsetPy, sizePy);

java 复制代码

private boolean syncDeliver(MessageExtBrokerInner msgInner, String msgId, long offset, long offsetPy,
            int sizePy) {
            PutResultProcess resultProcess = deliverMessage(msgInner, msgId, offset, offsetPy, sizePy, false);
            PutMessageResult result = resultProcess.get();
            boolean sendStatus = result != null && result.getPutMessageStatus() == PutMessageStatus.PUT_OK;
            if (sendStatus) {
                ScheduleMessageService.this.updateOffset(this.delayLevel, resultProcess.getNextOffset());
            }
            return sendStatus;
        }

如果成功则更新延时消息的消费进度，注意延时消息的消费进度是存储在

java 复制代码

    private final ConcurrentMap<Integer /* level */, Long/* offset */> offsetTable =
        new ConcurrentHashMap<>(32);

实际的配置文件是delayOffset.json

自此RocketMQ的延时消息我们就分析完了

总结

总的来说延时消息就是默认分为18个队列，然后启动18个线程一直去扫描这18个队列。如果没有消息就过100毫秒继续重复扫描，周而复始如果扫描到消息则将延时消息Topic：SCHEDULE_TOPIC_XXXX中的消息取出来,然后重新转换为原始消息，包括原始消息的Topic和queueId，然后重新写入到commitLog中