大家好，我是君哥。
今天來分享 RocketMQ 中一個非常重要又不太好理解的知識點-ProcessQueue。
一句話概括，ProcessQueue 就是 MessageQueue 的消費快照。看下面這張圖：
1 ProcessQueue 構(gòu)建
RocketMQ 客戶端啟動時，會開啟一個 rebalance 線程，代碼如下:
//MQClientInstance.java
public void start() throws MQClientException {
synchronized (this) {
switch (this.serviceState) {
case CREATE_JUST:
//...
// Start rebalance service
this.rebalanceService.start();
//...
}
}
}
這個線程會不停的做重平衡操作，對 ProcessQueue 進(jìn)行維護(hù)。在重平衡線程類 RebalanceImpl 定義了一個變量 processQueueTable，數(shù)據(jù)結(jié)構(gòu)如下：
可以看到，在 processQueueTable 這個數(shù)據(jù)結(jié)構(gòu)上維護(hù)了 MessageQueue 和 ProcessQueue 的映射。
下面看一下維護(hù) processQueueTable 的代碼：
private boolean updateProcessQueueTableInRebalance(final String topic, final Set<MessageQueue> mqSet,
final boolean isOrder) {
boolean changed = false;
Iterator<Entry<MessageQueue, ProcessQueue>> it = this.processQueueTable.entrySet().iterator();
while (it.hasNext()) {
Entry<MessageQueue, ProcessQueue> next = it.next();
MessageQueue mq = next.getKey();
ProcessQueue pq = next.getValue();
if (mq.getTopic().equals(topic)) {
if (!mqSet.contains(mq)) {
//從processQueueTable上移除
} else if (pq.isPullExpired()) {
switch (this.consumeType()) {
case CONSUME_ACTIVELY://拉模式
break;
case CONSUME_PASSIVELY://推模式
//從processQueueTable上移除
break;
default:
break;
}
}
}
}
//創(chuàng)建ProcessQueue并放到processQueueTable
List<PullRequest> pullRequestList = new ArrayList<PullRequest>();
for (MessageQueue mq : mqSet) {
if (!this.processQueueTable.containsKey(mq)) {
//...
ProcessQueue pq = new ProcessQueue();
long nextOffset = -1L;
try {
nextOffset = this.computePullFromWhereWithException(mq);
} catch (Exception e) {
log.info("doRebalance, {}, compute offset failed, {}", consumerGroup, mq);
continue;
}
if (nextOffset >= 0) {
ProcessQueue pre = this.processQueueTable.putIfAbsent(mq, pq);
if (pre != null) {
log.info("doRebalance, {}, mq already exists, {}", consumerGroup, mq);
} else {
//封裝好processQueueTable后再創(chuàng)建一個PullRequest進(jìn)行消息拉取
log.info("doRebalance, {}, add a new mq, {}", consumerGroup, mq);
PullRequest pullRequest = new PullRequest();
pullRequest.setConsumerGroup(consumerGroup);
pullRequest.setNextOffset(nextOffset);
pullRequest.setMessageQueue(mq);
pullRequest.setProcessQueue(pq);
pullRequestList.add(pullRequest);
changed = true;
}
} else {
log.warn("doRebalance, {}, add new mq failed, {}", consumerGroup, mq);
}
}
}
this.dispatchPullRequest(pullRequestList);
return changed;
}
2 拉取消息
上一節(jié)中構(gòu)建 ProcessQueue 后，會再創(chuàng)建一個 PullRequest，這個 PullRequest 封裝了 MessageQueue 和 ProcessQueue，創(chuàng)建成功后被放到了 PullMessageService 中的 pullRequestQueue 變量：
//PullMessageService.java
private final LinkedBlockingQueue<PullRequest> pullRequestQueue = new LinkedBlockingQueue<PullRequest>();
public void executePullRequestImmediately(final PullRequest pullRequest) {
try {
this.pullRequestQueue.put(pullRequest);
} catch (InterruptedException e) {
log.error("executePullRequestImmediately pullRequestQueue.put", e);
}
}
這里以 RocketMQ 的推模式為例，Consumer 拉取到消息后，會進(jìn)行如下處理：
- 對拉取到的消息根據(jù) TAG 再次
進(jìn)行過濾； - 更新 PullRequest 下次拉取的偏移量 nextOffset；
- 把拉取的消息封裝到 ProcessQueue 的 msgTreeMap（
?
放到 msgTreeMap 之前首先要獲取到寫鎖 treeMapLock
?
）； - 封裝 ConsumeRequest 進(jìn)行消息消費;
- 封裝消息拉取請求再次進(jìn)行拉取。
代碼如下：
//DefaultMQPushConsumerImpl.java
public void onSuccess(PullResult pullResult) {
if (pullResult != null) {
//1. 對拉取到的消息根據(jù) TAG 再次進(jìn)行過濾
pullResult = DefaultMQPushConsumerImpl.this.pullAPIWrapper.processPullResult(pullRequest.getMessageQueue(), pullResult,
subscriptionData);
switch (pullResult.getPullStatus()) {
case FOUND:
//2. 更新 PullRequest 下次拉取的偏移量 nextOffset
pullRequest.setNextOffset(pullResult.getNextBeginOffset());
if (pullResult.getMsgFoundList() == null || pullResult.getMsgFoundList().isEmpty()) {
DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
} else {
//3. 把拉取的消息封裝到 ProcessQueue 的 msgTreeMap
boolean dispatchToConsume = processQueue.putMessage(pullResult.getMsgFoundList());
//4. 封裝 ConsumeRequest 進(jìn)行消息消費
DefaultMQPushConsumerImpl.this.consumeMessageService.submitConsumeRequest(
pullResult.getMsgFoundList(),
processQueue,
pullRequest.getMessageQueue(),
dispatchToConsume);
//5. 封裝消息拉取請求
if (DefaultMQPushConsumerImpl.this.defaultMQPushConsumer.getPullInterval() > 0) {
DefaultMQPushConsumerImpl.this.executePullRequestLater(pullRequest,
DefaultMQPushConsumerImpl.this.defaultMQPushConsumer.getPullInterval());
} else {
DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
}
}
break;
//...
}
}
}
3 消費消息
在上一節(jié)提到過，拉取到消息后，會把消息封裝成一個 ConsumeRequest,這個線程類會調(diào)用消費者定義的 MessageListener 進(jìn)行消費處理。看一下源代碼：
//ConsumeMessageConcurrentlyService.ConsumeRequest
public void run() {
if (this.processQueue.isDropped()) {
log.info("the message queue not be able to consume, because it's dropped. group={} {}", ConsumeMessageConcurrentlyService.this.consumerGroup, this.messageQueue);
return;
}
MessageListenerConcurrently listener = ConsumeMessageConcurrentlyService.this.messageListener;
ConsumeConcurrentlyContext context = new ConsumeConcurrentlyContext(messageQueue);
ConsumeConcurrentlyStatus status = null;
try {
status = listener.consumeMessage(Collections.unmodifiableList(msgs), context);
}//...
if (!processQueue.isDropped()) {
ConsumeMessageConcurrentlyService.this.processConsumeResult(status, context, this);
}
}
消息消費成功后，會調(diào)用 processConsumeResult 方法進(jìn)行結(jié)果處理。對于廣播模式，發(fā)送失敗后不會做重試，相當(dāng)于把消息丟棄，而對于集群模式，消費失敗的消息會發(fā)送到 Broker 端等待消費者重新拉取進(jìn)行重試。
消費結(jié)果處理完后，消費成功的消息會從 ProcessQueue 的 msgTreeMap 中移除（需要獲取到寫鎖 treeMapLock），同時從 msgTreeMap 中獲取最小的 Offset 來更新對應(yīng) MessageQueue 的偏移量。這個邏輯可以參考下面代碼：
public void processConsumeResult(
final ConsumeConcurrentlyStatus status,
final ConsumeConcurrentlyContext context,
final ConsumeRequest consumeRequest
) {
int ackIndex = context.getAckIndex();
switch (status) {
case CONSUME_SUCCESS:
if (ackIndex >= consumeRequest.getMsgs().size()) {
ackIndex = consumeRequest.getMsgs().size() - 1;
}
int ok = ackIndex + 1;
break;
//...
}
switch (this.defaultMQPushConsumer.getMessageModel()) {
case BROADCASTING:
//...
break;
case CLUSTERING:
List<MessageExt> msgBackFailed = new ArrayList<MessageExt>(consumeRequest.getMsgs().size());
for (int i = ackIndex + 1; i < consumeRequest.getMsgs().size(); i++) {
MessageExt msg = consumeRequest.getMsgs().get(i);
//消費失敗的，發(fā)送回Broker
boolean result = this.sendMessageBack(msg, context);
//...
}
break;
default:
break;
}
//從msgTreeMap中移除并返回msgTreeMap第一條消息的offset
long offset = consumeRequest.getProcessQueue().removeMessage(consumeRequest.getMsgs());
if (offset >= 0 && !consumeRequest.getProcessQueue().isDropped()) {
this.defaultMQPushConsumerImpl.getOffsetStore().updateOffset(consumeRequest.getMessageQueue(), offset, true);
}
}
4 消費者限流
4.1 緩存消息數(shù)量
如果消費者緩存的消息數(shù)量大于 RocketMQ 配置的閾值（默認(rèn) 1000），就會觸發(fā)延遲拉取，而消費者緩存的消息數(shù)量就來自 ProcessQueue，看下面代碼：
long cachedMessageCount = processQueue.getMsgCount().get();
if (cachedMessageCount > this.defaultMQPushConsumer.getPullThresholdForQueue()) {
this.executePullRequestLater(pullRequest, PULL_TIME_DELAY_MILLS_WHEN_FLOW_CONTROL);
return;
}
4.2 緩存的消息大小
如果消費者緩存的消息大小大于 RocketMQ 配置的閾值（默認(rèn) 100M），就會觸發(fā)延遲拉取，而消費者緩存的消息大小就來自 ProcessQueue，看下面代碼：
long cachedMessageSizeInMiB = processQueue.getMsgSize().get() / (1024 * 1024);
if (cachedMessageSizeInMiB > this.defaultMQPushConsumer.getPullThresholdSizeForQueue()) {
this.executePullRequestLater(pullRequest, PULL_TIME_DELAY_MILLS_WHEN_FLOW_CONTROL);
return;
}
4.3 消息間隔
對于普通消息，如果消費偏移量間隔大于配置的閾值（默認(rèn) 2000），就會觸發(fā)延遲拉取，而消息間隔就來自 ProcessQueue，看下面代碼：
if (!this.consumeOrderly) {
if (processQueue.getMaxSpan() > this.defaultMQPushConsumer.getConsumeConcurrentlyMaxSpan()) {
this.executePullRequestLater(pullRequest, PULL_TIME_DELAY_MILLS_WHEN_FLOW_CONTROL);
return;
}
}4.4 獲取鎖失敗
對于順序消息，如果獲取鎖失敗，也會觸發(fā)延遲拉取，而判斷獲取鎖是否成功，也是在 ProcessQueue，看下面代碼：
if (processQueue.isLocked()) {
//...
} else {
this.executePullRequestLater(pullRequest, pullTimeDelayMillsWhenException);
}5 總結(jié)
ProcessQueue 是 MessageQueue 的消費快照，可以協(xié)助消費者進(jìn)行消息拉取、消息消費、更新偏移量、限流。最后，看一下 ProcessQueue 的數(shù)據(jù)結(jié)構(gòu)：
