PushConsumer 消费消息

CLUSTERING 模式下，消费者会订阅 retry topic

// DefaultMQPushConsumerImpl#copySubscription
private void copySubscription() throws MQClientException {
    try {
        Map<String, String> sub = this.defaultMQPushConsumer.getSubscription();
        if (sub != null) {
            for (final Map.Entry<String, String> entry : sub.entrySet()) {
                final String topic = entry.getKey();
                final String subString = entry.getValue();
                SubscriptionData subscriptionData = FilterAPI.buildSubscriptionData(this.defaultMQPushConsumer.getConsumerGroup(),
                    topic, subString);
                this.rebalanceImpl.getSubscriptionInner().put(topic, subscriptionData);
            }
        }

        if (null == this.messageListenerInner) {
            this.messageListenerInner = this.defaultMQPushConsumer.getMessageListener();
        }

        switch (this.defaultMQPushConsumer.getMessageModel()) {
            case BROADCASTING:
                break;
            case CLUSTERING:
                final String retryTopic = MixAll.getRetryTopic(this.defaultMQPushConsumer.getConsumerGroup());
                SubscriptionData subscriptionData = FilterAPI.buildSubscriptionData(this.defaultMQPushConsumer.getConsumerGroup(),
                    retryTopic, SubscriptionData.SUB_ALL);
                this.rebalanceImpl.getSubscriptionInner().put(retryTopic, subscriptionData);
                break;
            default:
                break;
        }
    } catch (Exception e) {
        throw new MQClientException("subscription exception", e);
    }
}

拉取消息

// org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl#pullMessage
// 拉取消息的回调 PullCallback：
// 消息放入 processQueue 的红黑树中，以分区 offset 作 key
boolean dispatchToConsume = processQueue.putMessage(pullResult.getMsgFoundList());
// 为拉取到的消息创建 ConsumeRequest 任务
DefaultMQPushConsumerImpl.this.consumeMessageService.submitConsumeRequest(
    pullResult.getMsgFoundList(),
    processQueue,
    pullRequest.getMessageQueue(),
    dispatchToConsume);

存放拉取的消息

// ProcessQueue#putMessage
// 消息放入 TreeMap 中，按 offset 排序
public boolean putMessage(final List<MessageExt> msgs) {
    boolean dispatchToConsume = false;
    try {
        this.lockTreeMap.writeLock().lockInterruptibly();
        try {
            int validMsgCnt = 0;
            for (MessageExt msg : msgs) {
                MessageExt old = msgTreeMap.put(msg.getQueueOffset(), msg);
                if (null == old) {
                    validMsgCnt++;
                    this.queueOffsetMax = msg.getQueueOffset();
                    msgSize.addAndGet(msg.getBody().length);
                }
            }
            msgCount.addAndGet(validMsgCnt);

            if (!msgTreeMap.isEmpty() && !this.consuming) {
                dispatchToConsume = true;
                this.consuming = true;
            }

            if (!msgs.isEmpty()) {
                MessageExt messageExt = msgs.get(msgs.size() - 1);
                String property = messageExt.getProperty(MessageConst.PROPERTY_MAX_OFFSET);
                if (property != null) {
                    long accTotal = Long.parseLong(property) - messageExt.getQueueOffset();
                    if (accTotal > 0) {
                        this.msgAccCnt = accTotal;
                    }
                }
            }
        } finally {
            this.lockTreeMap.writeLock().unlock();
        }
    } catch (InterruptedException e) {
        log.error("putMessage exception", e);
    }

    return dispatchToConsume;
}

提交 ConsumeRequest

// ConsumeMessageConcurrentlyService#submitConsumeRequest
// 按照 consumeBatchSize 对消息分批
public void submitConsumeRequest(
    final List<MessageExt> msgs,
    final ProcessQueue processQueue,
    final MessageQueue messageQueue,
    final boolean dispatchToConsume) {
    final int consumeBatchSize = this.defaultMQPushConsumer.getConsumeMessageBatchMaxSize();
    if (msgs.size() <= consumeBatchSize) {
        ConsumeRequest consumeRequest = new ConsumeRequest(msgs, processQueue, messageQueue);
        try {
            this.consumeExecutor.submit(consumeRequest);
        } catch (RejectedExecutionException e) {
            this.submitConsumeRequestLater(consumeRequest);
        }
    } else {
        for (int total = 0; total < msgs.size(); ) {
            List<MessageExt> msgThis = new ArrayList<MessageExt>(consumeBatchSize);
            for (int i = 0; i < consumeBatchSize; i++, total++) {
                if (total < msgs.size()) {
                    msgThis.add(msgs.get(total));
                } else {
                    break;
                }
            }

            ConsumeRequest consumeRequest = new ConsumeRequest(msgThis, processQueue, messageQueue);
            try {
                this.consumeExecutor.submit(consumeRequest);
            } catch (RejectedExecutionException e) {
                for (; total < msgs.size(); total++) {
                    msgThis.add(msgs.get(total));
                }

                this.submitConsumeRequestLater(consumeRequest);
            }
        }
    }
}

并行消费和顺序消费

ConsumeMessageConcurrentlyService.ConsumeRequest#run
ConsumeMessageOrderlyService.ConsumeRequest#run

按批消费消息，消费完一批消息返回状态。消费成功，则删除这一批消息，并更新 offsetTable。

ConsumeMessageConcurrentlyService 和 ConsumeMessageOrderlyService 的线程池是一样的，顺序消费体现在，同一时刻，一个分区的消息只有一个线程在消费。分区消息的到达是有先后顺序，但消费是在多个线程中进行。 一个 MessageQueue 对应一把锁。

重点看下消费失败的情形

public void processConsumeResult(
    final ConsumeConcurrentlyStatus status,
    final ConsumeConcurrentlyContext context,
    final ConsumeRequest consumeRequest
) {
    //  Integer.MAX_VALUE
    int ackIndex = context.getAckIndex();

    if (consumeRequest.getMsgs().isEmpty())
        return;

    switch (status) {
        case CONSUME_SUCCESS:
            if (ackIndex >= consumeRequest.getMsgs().size()) {
                ackIndex = consumeRequest.getMsgs().size() - 1;
            }
            int ok = ackIndex + 1;
            int failed = consumeRequest.getMsgs().size() - ok;
            this.getConsumerStatsManager().incConsumeOKTPS(consumerGroup, consumeRequest.getMessageQueue().getTopic(), ok);
            this.getConsumerStatsManager().incConsumeFailedTPS(consumerGroup, consumeRequest.getMessageQueue().getTopic(), failed);
            break;
        case RECONSUME_LATER:
            ackIndex = -1;
            this.getConsumerStatsManager().incConsumeFailedTPS(consumerGroup, consumeRequest.getMessageQueue().getTopic(),
                consumeRequest.getMsgs().size());
            break;
        default:
            break;
    }

    switch (this.defaultMQPushConsumer.getMessageModel()) {
        case BROADCASTING:
            for (int i = ackIndex + 1; i < consumeRequest.getMsgs().size(); i++) {
                MessageExt msg = consumeRequest.getMsgs().get(i);
                log.warn("BROADCASTING, the message consume failed, drop it, {}", msg.toString());
            }
            break;
        case CLUSTERING:
            List<MessageExt> msgBackFailed = new ArrayList<MessageExt>(consumeRequest.getMsgs().size());
            // 消费失败，ackIndex = -1
            for (int i = ackIndex + 1; i < consumeRequest.getMsgs().size(); i++) {
                MessageExt msg = consumeRequest.getMsgs().get(i);
                // 消息重新发送回队列
                boolean result = this.sendMessageBack(msg, context);
                // 如果发送失败，则放入 msgBackFailed 重新消费
                if (!result) {
                    msg.setReconsumeTimes(msg.getReconsumeTimes() + 1);
                    msgBackFailed.add(msg);
                }
            }

            if (!msgBackFailed.isEmpty()) {
                consumeRequest.getMsgs().removeAll(msgBackFailed);
                // 根据 msgBackFailed 重新生成 ConsumeRequest
                this.submitConsumeRequestLater(msgBackFailed, consumeRequest.getProcessQueue(), consumeRequest.getMessageQueue());
            }
            break;
        default:
            break;
    }

    // 发送回 broker 的消息相当于是处理掉了，从红黑树中删除，返回最小的 offset
    long offset = consumeRequest.getProcessQueue().removeMessage(consumeRequest.getMsgs());
    if (offset >= 0 && !consumeRequest.getProcessQueue().isDropped()) {
        this.defaultMQPushConsumerImpl.getOffsetStore().updateOffset(consumeRequest.getMessageQueue(), offset, true);
    }
}

broker 处理 retry 消息

SendMessageProcessor#consumerSendMsgBack

消费失败的消息，会发送到 %RETRY% + consumerGroup，消费者重新拉取消费。

如果重复消费的次数达到 16 次，则 broker 把消息发送到死信队列 %DLQ% + consumerGroup。 

位移提交

// MQClientInstance#startScheduledTask
// 定时任务提交 offset
this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {

    @Override
    public void run() {
        try {
            MQClientInstance.this.persistAllConsumerOffset();
        } catch (Exception e) {
            log.error("ScheduledTask persistAllConsumerOffset exception", e);
        }
    }
}, 1000 * 10, this.clientConfig.getPersistConsumerOffsetInterval(), TimeUnit.MILLISECONDS);