RocketMQ客户端PUSH消费--负载均衡

瓜农老梁

发布于 2019-08-06 16:04:27

9710

发布于 2019-08-06 16:04:27

文章被收录于专栏：瓜农老梁瓜农老梁

一、问题思考

1.主题队列是如何分配的？ 2.什么时候会进行负载均衡？ 3.负载均衡后是否会导致消息重复消费？

二、调用链条

1.初始化链条

@1 DefaultMQPushConsumerImpl#start
this.mQClientFactory
= MQClientManager.getInstance().getAndCreateMQClientInstance
@2 MQClientManager#getAndCreateMQClientInstance
instance = new MQClientInstance
@3 MQClientInstance#MQClientInstance
this.rebalanceService = new RebalanceService

2.启动链条

@1 DefaultMQPushConsumerImpl#start
mQClientFactory.start()
@2 MQClientInstance#start
this.rebalanceService.start

小结：从初始化链和调用链可以看出RebalanceService为线程类，随着消费启动时而启动，消费不退出则一直运行着。

三、负载均衡流程

1.负载均衡链条

@1 RebalanceService#run
mqClientFactory.doRebalance()
@2 MQClientInstance#doRebalance
impl.doRebalance()
@3 DefaultMQPushConsumerImpl#doRebalance
this.rebalanceImpl.doRebalance
@4 RebalanceImpl#doRebalance
rebalanceByTopic

2.负载均衡流程

小结：在负载均衡时，会循环该消费组订阅的所有Topic都会执行负载均衡。

3.更新缓存processQueue流程

小结： 1. 更新缓存时如果消费组订阅的队列不在新分配的队列集合中或者队列拉取时间超时失效，则将快照ProcessQueue设置为丢弃。 2. 消费拉取时判断ProcessQueue为丢弃，则不再对该队列拉取。 3. 顺序消费时如果获取消费锁成功，表明此队列空闲没有被消费，此时向Broker发起解锁请求，解锁成功后将该队列从缓存（processQueueTable）移除。 4. 顺序消费时获取锁失败，表明正在消费则不从processQueueTable移除，由于ProcessQueue设置为丢弃，在顺序消费下次拉取时会退出该队列的拉取请求。

4.向Broker发送心跳流程

5.队列分配算法

负载均衡流程图中对clientId和分区队列的分配提交给分区算法执行，那该算法是如何运作的呢？接口AllocateMessageQueueStrategy队列分配策略提供五种分配算法实现：

1.平均分配策略 AllocateMessageQueueAveragely

2.环形分配策略 AllocateMessageQueueAveragelyByCircle

3.机房分配策略 AllocateMessageQueueByMachineRoom

4.一致性Hash分配策略

AllocateMessageQueueConsistentHash

5.配置文件分配策略 AllocateMessageQueueByConfig 除此之外可以自定义分配算法，实现接口接口即可，默认使用平均分配算法，也是最常用的，下面以该算法看看如何工作的。

public List<MessageQueue> allocate
(String consumerGroup, String currentCID, List<MessageQueue> mqAll,
List<String> cidAll) {
List<MessageQueue> result = new ArrayList<MessageQueue>();
int index = cidAll.indexOf(currentCID);
int mod = mqAll.size() % cidAll.size();
int averageSize = mqAll.size() <= cidAll.size() ?
1 : (mod > 0 && index < mod ?
mqAll.size() / cidAll.size() + 1 : mqAll.size() / cidAll.size());
int startIndex = (mod > 0 && index < mod) ? index * averageSize : index * averageSize + mod;
int range = Math.min(averageSize, mqAll.size() - startIndex);
for (int i = 0; i < range; i++) {
result.add(mqAll.get((startIndex + i) % mqAll.size()));
}
return result;
}

代码不是很好阅读，看下面验证结果即可。

6.平均分配算法验证

只有一个clientId时分配情况会把1个Broker的16个分区全部分配给该客户端，每隔20秒触发一次负载均衡。 currentCID=2.0.1.138@consumer01分到的队列为0～15

----------2019-08-04 22:10:15-----------
currentCID=2.0.1.138@consumer01
index=0
mod=0
averageSize=16
startIndex=0
range=16
result=[MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=0], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=1], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=2], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=3], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=4], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=5], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=6], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=7], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=8], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=9], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=10], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=11], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=12], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=13], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=14], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=15]]
----------2019-08-04 22:10:35-----------
currentCID=2.0.1.138@consumer01
index=0
mod=0
averageSize=16
startIndex=0
range=16
result=[MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=0], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=1], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=2], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=3], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=4], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=5], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=6], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=7], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=8], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=9], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=10], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=11], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=12], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=13], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=14], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=15]]

新加入第二个client时此时有两个clinetId分别为2.0.1.138@consumer01和2.0.1.138@consumer02，1个 Broker16个分区的分配情况。 currentCID=2.0.1.138@consumer01分到的分区为0～7 currentCID=2.0.1.138@consumer02分到的分区为8～16

----------2019-08-04 22:12:25-----------
currentCID=2.0.1.138@consumer01
index=0
mod=0
averageSize=8
startIndex=0
range=8
result=[MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=0], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=1], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=2], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=3], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=4], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=5], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=6], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=7]]
----------2019-08-04 22:12:45-----------
currentCID=2.0.1.138@consumer02
index=1
mod=0
averageSize=8
startIndex=8
range=8
result=[MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=8], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=9], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=10], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=11], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=12], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=13], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=14], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=15]]

新加入第三个client时此时有三个客户端2.0.1.138@consumer01、2.0.1.138@consumer02、2.0.1.138@consumer03，1个Broker的16个队列的分配情况。 currentCID=2.0.1.138@consumer01分到的队列0～5 currentCID=2.0.1.138@consumer02分到的队列6～10 currentCID=2.0.1.138@consumer03分到的队列11～15

----------2019-08-04 22:13:58-----------
currentCID=2.0.1.138@consumer01
index=0
mod=1
averageSize=6
startIndex=0
range=6
result=[MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=0], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=1], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=2], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=3], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=4], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=5]]
----------2019-08-04 22:14:18-----------
currentCID=2.0.1.138@consumer02
index=1
mod=1
averageSize=5
startIndex=6
range=5
result=[MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=6], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=7], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=8], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=9], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=10]]
----------2019-08-04 22:14:39-----------
currentCID=2.0.1.138@consumer03
index=2
mod=1
averageSize=5
startIndex=11
range=5
result=[MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=11], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=12], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=13], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=14], MessageQueue [topic=topic_test, brokerName=liangyongdeMacBook-Pro.local, queueId=15]]

四、总结

1.主题队列是如何分配的？

备注：见队列分配算法，通常使用平均分配算法。

2.什么时候会进行负载均衡？

备注：负载均衡线程每隔20秒执行一次，当有新客户端退出或者加入或者新的Broker加入或掉线都会触发重新负载均衡。

3.负载均衡后是否会导致消息重复消费？

备注：

情况1: 并发消费可能导致消息被重复消费，看以下代码。

//并发消费对结果的处理

//ConsumeMessageConcurrentlyService#ConsumeRequest

if (!processQueue.isDropped()) { ConsumeMessageConcurrentlyService.this.processConsumeResult(status, context, this);

} else { //被丢弃，消费进度不会更新

log.warn("processQueue is dropped without process consume result. messageQueue= {}, msgs={}", messageQueue, msgs);

}

如果负载均衡前已分配的队列不在负载均衡后的新队列集合中，会丢弃该队列即：processQueue.isDropped()。而该队列可能已经被消费完了，在处理结果时被丢弃了，消费进度没有更新。别的消费客户端重新拉取该队列时造成重复消费。

情况2: 顺序消费不会导致消息被重复消费

本文参与腾讯云自媒体分享计划，分享自微信公众号。

原始发表：2019-08-04，如有侵权请联系 cloudcommunity@tencent.com 删除

负载均衡

负载均衡缓存

编程算法

缓存

本文分享自瓜农老梁微信公众号，前往查看

如有侵权，请联系 cloudcommunity@tencent.com 删除。

本文参与腾讯云自媒体分享计划，欢迎热爱写作的你一起参与！

登录后参与评论

0 条评论

热度