那么你期望的是不是 2*3=6 刚好6个线程;一个线程分配一个分区; 那么我们运行看看结果
看上图中,我们发现并没有按照我们的预期去做; 有三个消费者其实是闲置状态的; 只有另外的3个消费者负责了2个Topic的总共6个分区; 因为默认的分配策略是 spring.kafka.consumer.properties.partition.assignment.strategy=\ org.apache.kafka.clients.consumer.RangeAssignor
;
如果想达到我们的预期;那你可以修改策略; spring.kafka.consumer.properties.partition.assignment.strategy=\ org.apache.kafka.clients.consumer.RoundRobinAssignor
修改之后
每个线程分配一个分区
分区数3|concurrency = 1|启动一个客户端(单机)
创建了名为 SHI_TOPIC3
并且分区数为3的Topic
代码启动,设置concurrency = 1
, 只启动一个客户端;
启动日志
2020-11-18 17:14:42 o.a.k.c.c.i.ConsumerCoordinator 611 [INFO] [Consumer clientId=myClientId5-0, groupId=consumer-id5]
Finished assignment for group at generation 6: {myClientId5-0-a273480d-2370-49e5-9187-ed10fe6dcf51=
Assignment(partitions=[SHI_TOPIC3-0, SHI_TOPIC3-1, SHI_TOPIC3-2])}
2020-11-18 17:14:42 o.s.k.l.KafkaMessageListenerContainer 292 [INFO] consumer-id5:
partitions assigned: [SHI_TOPIC3-2, SHI_TOPIC3-1, SHI_TOPIC3-0]
可以看到这个客户端myClientId5-0-a273480d-2370-49e5-9187-ed10fe6dcf51
被分配了3个分区SHI_TOPIC3-0, SHI_TOPIC3-1, SHI_TOPIC3-2
;
消费日志
2020-11-18 17:14:45 c.d.b.k.KafkaListeners 109 [INFO] 线程:Thread[consumer-id5-0-C-1,5,main] consumer-id5 消费->record:ConsumerRecord(topic = SHI_TOPIC3, partition = 2, leaderEpoch = 0, offset = 0, CreateTime = 1605690882681, serialized key size = 13, serialized value size = 11, headers = RecordHeaders(headers = [], isReadOnly = false), key = 1605690882615, value = 我是data0),value:我是data0,partition:2,offset:0
2020-11-18 17:14:45 c.d.b.k.KafkaListeners 109 [INFO] 线程:Thread[consumer-id5-0-C-1,5,main] consumer-id5 消费->record:ConsumerRecord(topic = SHI_TOPIC3, partition = 2, leaderEpoch = 0, offset = 1, CreateTime = 1605690882705, serialized key size = 13, serialized value size = 11, headers = RecordHeaders(headers = [], isReadOnly = false), key = 1605690882705, value = 我是data4),value:我是data4,partition:2,offset:1
2020-11-18 17:14:45 c.d.b.k.KafkaListeners 109 [INFO] 线程:Thread[consumer-id5-0-C-1,5,main] consumer-id5 消费->record:ConsumerRecord(topic = SHI_TOPIC3, partition = 2, leaderEpoch = 0, offset = 2, CreateTime = 1605690882705, serialized key size = 13, serialized value size = 11, headers = RecordHeaders(headers = [], isReadOnly = false), key = 1605690882705, value = 我是data5),value:我是data5,partition:2,offset:2
2020-11-18 17:14:45 c.d.b.k.KafkaListeners 109 [INFO] 线程:Thread[consumer-id5-0-C-1,5,main] consumer-id5 消费->record:ConsumerRecord(topic = SHI_TOPIC3, partition = 2, leaderEpoch = 0, offset = 3, CreateTime = 1605690882706, serialized key size = 13, serialized value size = 11, headers = RecordHeaders(headers = [], isReadOnly = false), key = 1605690882705, value = 我是data6),value:我是data6,partition:2,offset:3
2020-11-18 17:14:45 c.d.b.k.KafkaListeners 109 [INFO] 线程:Thread[consumer-id5-0-C-1,5,main] consumer-id5 消费->record:ConsumerRecord(topic = SHI_TOPIC3, partition = 2, leaderEpoch = 0, offset = 4, CreateTime = 1605690882706, serialized key size = 13, serialized value size = 11, headers = RecordHeaders(headers = [], isReadOnly = false), key = 1605690882706, value = 我是data7),value:我是data7,partition:2,offset:4
…
可以看到线程都是同一个 Thread[consumer-id5-0-C-1,5,main]
; 说明的问题就是 在消费的时候是单线程消费的,并且还是一个线程去消费 3个分区的数据; 又涉及到切换消费分区的问题;
查询这个消费组的消费情况;
也证实只有一个消费者myClientId5-0-a273480d-2370-49e5-9187-ed10fe6dcf51
在消费3个分区的数据;
分区数3|concurrency = 1|启动2个客户端(分布式模式)
第一个客户端不动,继续运行, 然后启动第二个客户端
第一个客户端发生的变化
2020-11-18 17:34:24 o.a.k.c.c.i.ConsumerCoordinator 611 [INFO] [Consumer clientId=myClientId5-0, groupId=consumer-id5] Finished assignment for group at generation 9: {myClientId5-0-66a81e88-d924-4890-8b8e-2c6960ed0704=Assignment(partitions=[SHI_TOPIC3-2]), myClientId5-0-31c9a99f-5735-4a1d-b537-95bc5ab4533f=Assignment(partitions=[SHI_TOPIC3-0, SHI_TOPIC3-1])}
第一个客户端进行了 再平衡 ; 因为多了第二个可以分担压力进行消费; 可以看到把SHI_TOPIC3-2
平衡出去了
第二个客户端的日志
2020-11-18 17:34:24 o.a.k.c.Metadata 277 [INFO] [Consumer clientId=myClientId5-0, groupId=consumer-id5] Cluster ID: O304VSOeSEyporzbs5AITA
2020-11-18 17:34:24 o.a.k.c.c.i.AbstractCoordinator 797 [INFO] [Consumer clientId=myClientId5-0, groupId=consumer-id5] Discovered group coordinator xxxxxx:9092 (id: 2147483645 rack: null)
2020-11-18 17:34:24 o.a.k.c.c.i.AbstractCoordinator 552 [INFO] [Consumer clientId=myClientId5-0, groupId=consumer-id5] (Re-)joining group
2020-11-18 17:34:25 o.s.k.l.KafkaMessageListenerContainer 292 [INFO] consumer-id5: partitions assigned: [SHI_TOPIC3-2]
查询客户端消费情况
可以看到第二个客户端分配到了SHI_TOPIC3--2
的分区进行消费; 并且是单线程消费;
分区数3|concurrency = 3|启动一个客户端
客户端日志
2020-11-18 17:50:42 o.a.k.c.c.i.ConsumerCoordinator 273 [INFO] [Consumer clientId=myClientId5-1, groupId=consumer-id5] Adding newly assigned partitions: SHI_TOPIC3-1
2020-11-18 17:50:42 o.a.k.c.c.i.ConsumerCoordinator 273 [INFO] [Consumer clientId=myClientId5-0, groupId=consumer-id5] Adding newly assigned partitions: SHI_TOPIC3-0
2020-11-18 17:50:42 o.a.k.c.c.i.ConsumerCoordinator 273 [INFO] [Consumer clientId=myClientId5-2, groupId=consumer-id5] Adding newly assigned partitions: SHI_TOPIC3-2
2020-11-18 17:50:42 o.s.k.l.KafkaMessageListenerContainer 292 [INFO] consumer-id5: partitions assigned: [SHI_TOPIC3-2]
2020-11-18 17:50:42 o.s.k.l.KafkaMessageListenerContainer 292 [INFO] consumer-id5: partitions assigned: [SHI_TOPIC3-0]
2020-11-18 17:50:42 o.s.k.l.KafkaMessageListenerContainer 292 [INFO] consumer-id5: partitions assigned: [SHI_TOPIC3-1]
上面日志显示 创建了3个消费者,他们都属于同一个消费组groupId=consumer-id5
,3个分区刚好3个消费者一人一个分区平均分配;
客户端日志
2020-11-18 17:50:45 c.d.b.k.KafkaListeners 109 [INFO] 线程:Thread[consumer-id5-0-C-1,5,main] consumer-id5 消费->record:ConsumerRecord(topic = SHI_TOPIC3, partition = 0, leaderEpoch = 0, offset = 11, CreateTime = 1605693042720, serialized key size = 13, serialized value size = 11, headers = RecordHeaders(headers = [], isReadOnly = false), key = 1605693042432, value = 我是data0),value:我是data0,partition:0,offset:11
2020-11-18 17:50:45 c.d.b.k.KafkaListeners 109 [INFO] 线程:Thread[consumer-id5-2-C-1,5,main] consumer-id5 消费->record:ConsumerRecord(topic = SHI_TOPIC3, partition = 2, leaderEpoch = 0, offset = 12, CreateTime = 1605693042751, serialized key size = 13, serialized value size = 11, headers = RecordHeaders(headers = [], isReadOnly = false), key = 1605693042750, value = 我是data1),value:我是data1,partition:2,offset:12
2020-11-18 17:50:45 c.d.b.k.KafkaListeners 109 [INFO] 线程:Thread[consumer-id5-1-C-1,5,main] consumer-id5 消费->record:ConsumerRecord(topic = SHI_TOPIC3, partition = 1, leaderEpoch = 0, offset = 17, CreateTime = 1605693042757, serialized key size = 13, serialized value size = 11, headers = RecordHeaders(headers = [], isReadOnly = false), key = 1605693042757, value = 我是data7),value:我是data7,partition:1,offset:17
每个消费者都是单线程,一个线程消费一个分区
分区数3|concurrency = 3|启动2个客户端(分布式模式)
启动第一个客户端
启动第二个客户端
_**启动第二个客户端之后就发生了 再分配rebalance
; 可以看到,总共就有6个消费者, 但是其中的3个都是处于空闲状态;
因为一个分区最多只能有一个分区来进行消费;**_
/**
-
监听器工厂 批量消费
-
@return
*/
@Bean
自我介绍一下,小编13年上海交大毕业,曾经在小公司待过,也去过华为、OPPO等大厂,18年进入阿里一直到现在。
深知大多数Java工程师,想要提升技能,往往是自己摸索成长或者是报班学习,但对于培训机构动则几千的学费,着实压力不小。自己不成体系的自学效果低效又漫长,而且极易碰到天花板技术停滞不前!
因此收集整理了一份《2024年Java开发全套学习资料》,初衷也很简单,就是希望能够帮助到想自学提升又不知道该从何学起的朋友,同时减轻大家的负担。
既有适合小白学习的零基础资料,也有适合3年以上经验的小伙伴深入学习提升的进阶课程,基本涵盖了95%以上Java开发知识点,真正体系化!
由于文件比较大,这里只是将部分目录截图出来,每个节点里面都包含大厂面经、学习笔记、源码讲义、实战项目、讲解视频,并且会持续更新!
如果你觉得这些内容对你有帮助,可以扫码获取!!(备注Java获取)
最后
对于很多Java工程师而言,想要提升技能,往往是自己摸索成长,不成体系的学习效果低效漫长且无助。
整理的这些资料希望对Java开发的朋友们有所参考以及少走弯路,本文的重点是你有没有收获与成长,其余的都不重要,希望读者们能谨记这一点。
再分享一波我的Java面试真题+视频学习详解+技能进阶书籍
《一线大厂Java面试题解析+核心总结学习笔记+最新讲解视频+实战项目源码》,点击传送门即可获取!
工程师而言,想要提升技能,往往是自己摸索成长,不成体系的学习效果低效漫长且无助。**
整理的这些资料希望对Java开发的朋友们有所参考以及少走弯路,本文的重点是你有没有收获与成长,其余的都不重要,希望读者们能谨记这一点。
再分享一波我的Java面试真题+视频学习详解+技能进阶书籍
[外链图片转存中…(img-YrcpmP11-1712076542331)]
《一线大厂Java面试题解析+核心总结学习笔记+最新讲解视频+实战项目源码》,点击传送门即可获取!