Kafka Broker（一）

1、Kafka Broker 工作流程

1.1、Zookeeper 存储的 Kafka 信息

[hui@hadoop103 zookeeper-3.4.10]$ bin/zkCli.sh 
[zk: localhost:2181(CONNECTED) 0] ls /
[zookeeper, spark, kafka, hbase]
[zk: localhost:2181(CONNECTED) 1] ls /kafka
[cluster, controller_epoch, controller, brokers, feature, admin, isr_change_notification, consumers, log_dir_event_notification, latest_producer_id_block, config]
[zk: localhost:2181(CONNECTED) 2] 

1.2、KafkaBroker总体工作流程

 

1. broker启动后在zk中注册
2. controller谁先注册，谁说了算
3. 由选举出来的Controller监听brokers节点变化
4. Controller决定Leader选举 选举规则：在isr中存活为前提，按照AR中排在前面的优先。例如ar[1,0,2], isr[1，0，2]，那么leader就会按照1，0，2的顺序轮询
5. Controller将节点信息上传到ZK
6. 其他contorller从zk同步相关信息
7. 假设Broker1中Leader挂了
8. Controller监听到节点变化 获取ISR
9. 选举新的Leader（在isr中存活为前提，按照AR中排在前面的优先）
10. 更新Leader及ISR

模拟Kafka上下线，Zookeeper中数据变化

1、查看/kafka/brokers/ids路径上的节点。

[zk: localhost:2181(CONNECTED) 2] ls /kafka/brokers/ids
[0, 1, 2]

2、查看/kafka/controller路径上的数据

[zk: localhost:2181(CONNECTED) 3] get /kafka/controller
{"version":1,"brokerid":2,"timestamp":"1640209253641"}
cZxid = 0x640000003b
ctime = Thu Dec 23 05:40:53 CST 2021
mZxid = 0x640000003b
mtime = Thu Dec 23 05:40:53 CST 2021
pZxid = 0x640000003b
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x57de415196e0001
dataLength = 54
numChildren = 0

3、查看/kafka/brokers/topics/first/partitions/0/state路径上的数据。 

[zk: localhost:2181(CONNECTED) 4] get /kafka/brokers/topics/first/partitions/0/state
{"controller_epoch":16,"leader":1,"version":1,"leader_epoch":12,"isr":[1]}
cZxid = 0x5e00000006
ctime = Mon Mar 28 13:07:18 CST 2022
mZxid = 0x64000000cb
mtime = Thu Dec 23 05:42:03 CST 2021
pZxid = 0x5e00000006
cversion = 0
dataVersion = 12
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 74
numChildren = 0

4、停止hadoop105上的kafka。

[hui@hadoop105 ~]$ jps
2005 QuorumPeerMain
3159 Jps
2399 Kafka
[hui@hadoop105 ~]$ kill 2399
[hui@hadoop105 ~]$ jps
2005 QuorumPeerMain
3177 Jps

5、再次查看/kafka/brokers/ids路径上的节点 

[zk: localhost:2181(CONNECTED) 5] ls /kafka/brokers/ids
[0, 1]

刚才停止了hadoop105 节点上的kafka，对应的brokerI=2，现在zk已经不显示brokerid=2的信息了。

6、再次查看/kafka/controller路径上的数据 

[zk: localhost:2181(CONNECTED) 6]  get /kafka/controller
{"version":1,"brokerid":1,"timestamp":"1648591708926"}
cZxid = 0x6500000006
ctime = Wed Mar 30 06:08:28 CST 2022
mZxid = 0x6500000006
mtime = Wed Mar 30 06:08:28 CST 2022
pZxid = 0x6500000006
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x37fd7975d1c0002
dataLength = 54
numChildren = 0

现在的leader已经是 hadoop104 了

7、再次查看/kafka/brokers/topics/first/partitions/0/state路径上的数据 

[zk: localhost:2181(CONNECTED) 7] get /kafka/brokers/topics/first/partitions/0/state
{"controller_epoch":16,"leader":1,"version":1,"leader_epoch":12,"isr":[1]}
cZxid = 0x5e00000006
ctime = Mon Mar 28 13:07:18 CST 2022
mZxid = 0x64000000cb
mtime = Thu Dec 23 05:42:03 CST 2021
pZxid = 0x5e00000006
cversion = 0
dataVersion = 12
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 74
numChildren = 0

现在我们再次启动 hadoop105 上的kafka

[hui@hadoop103 kafka]$ jps.sh
------------------- hui@hadoop103 --------------
3168 Kafka
2016 QuorumPeerMain
3412 ZooKeeperMain
3918 Jps
------------------- hui@hadoop104 --------------
3570 Jps
2023 QuorumPeerMain
2830 Kafka
------------------- hui@hadoop105 --------------
2005 QuorumPeerMain
3562 Kafka
3661 Jps
[zk: localhost:2181(CONNECTED) 8] ls /kafka/brokers/ids                             
[0, 1, 2]
[zk: localhost:2181(CONNECTED) 9]  get /kafka/controller                            
{"version":1,"brokerid":1,"timestamp":"1648591708926"}
cZxid = 0x6500000006
ctime = Wed Mar 30 06:08:28 CST 2022
mZxid = 0x6500000006
mtime = Wed Mar 30 06:08:28 CST 2022
pZxid = 0x6500000006
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x37fd7975d1c0002
dataLength = 54
numChildren = 0
[zk: localhost:2181(CONNECTED) 10] get /kafka/brokers/topics/first/partitions/0/state
{"controller_epoch":16,"leader":1,"version":1,"leader_epoch":12,"isr":[1]}
cZxid = 0x5e00000006
ctime = Mon Mar 28 13:07:18 CST 2022
mZxid = 0x64000000cb
mtime = Thu Dec 23 05:42:03 CST 2021
pZxid = 0x5e00000006
cversion = 0
dataVersion = 12
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 74
numChildren = 0

再次启动后 leader 未重新选取。

1.3、Broker重要参数

 

2、节点服役和退役

2,.1  服役新节点

1、新节点准备

为了方便起见，停止原有 kafka 和 zk 集群，关闭hadoop105客户机，进行完整克隆，命名为 hadoop102;

配置主机IP

[hui@hadoop102 ~]$ less /etc/sysconfig/network-scripts/ifcfg-eth0 
DEVICE=eth0
HWADDR=00:0c:29:c9:1d:9c
TYPE=Ethernet
UUID=65e5ce7d-7df7-400b-b519-3c986d9dd3c9
ONBOOT=yes
NM_CONTROLLED=yes
BOOTPROTO=static
IPADDR=192.168.124.131
GATEWAY=192.168.14.2
DNS1=192.168.124.2
BOOTPROTO=dhcp

配置主机名

[hui@hadoop102 ~]$ cat /etc/sysconfig/network
NETWORKING=yes
HOSTNAME=hadoop102

重启hadoop102,开启hadoop105

群起 zk & kafka

[hui@hadoop103 kafka]$ jps.sh
------------------- hui@hadoop103 --------------
3168 Kafka
2016 QuorumPeerMain
4310 Jps
------------------- hui@hadoop104 --------------
3987 Jps
2023 QuorumPeerMain
2830 Kafka
------------------- hui@hadoop105 --------------
2005 QuorumPeerMain
3562 Kafka
4078 Jps

修改 hadoop102 kafka 跑配置文件的 broker_id = 3

less config/server.properties 
broker.id=3

删除 kafka 目录下的data 和 log 目录

hadoop102 kafka]$ rm -rf datas/ logs/

单独启动 hadoop102 上的kafka

[hui@hadoop102 kafka]$ bin/kafka-server-start.sh -daemon ./config/server.properties
[hui@hadoop102 kafka]$ jps
2795 Kafka
2812 Jps

2、执行负载均衡

创建一个要均衡的主题

[hui@hadoop103 kafka]$ vim topics-to-move.json
{
        "topics": [{
                "topic": "first"
        }],
        "version": 1
}

生成负载均衡计划

[hui@hadoop103 kafka]$ bin/kafka-reassign-partitions.sh --bootstrap-server hadoop103:9092 --topics-to-move-json-file topics-to-move.json --broker-list "0,1,2,3" --generate
Current partition replica assignment
{"version":1,"partitions":[{"topic":"first","partition":0,"replicas":[1],"log_dirs":["any"]},{"topic":"first","partition":1,"replicas":[2],"log_dirs":["any"]},{"topic":"first","partition":2,"replicas":[0],"log_dirs":["any"]}]}

Proposed partition reassignment configuration
{"version":1,"partitions":[{"topic":"first","partition":0,"replicas":[0],"log_dirs":["any"]},{"topic":"first","partition":1,"replicas":[1],"log_dirs":["any"]},{"topic":"first","partition":2,"replicas":[2],"log_dirs":["any"]}]}
[hui@hadoop103 kafka]$ 

创建副本存储计划（所有副本存储在broker0、broker1、broker2、broker3中） 

[hui@hadoop103 kafka]$ vim increase-replication-factor.json
{"version":1,"partitions":[{"topic":"first","partition":0,"replicas":[2,3,0],"log_dirs":["any","any","any"]},{"topic":"first","partition":1,"replicas":[3,0,1],"log_dirs":["any","any","any"]},{"topic":"first","partition":2,"replicas":[0,1,2],"log_dirs":["any","any","any"]}]}

执行副本存储计划

[hui@hadoop103 kafka]$ bin/kafka-reassign-partitions.sh --bootstrap-server hadoop103:9092 --reassignment-json-file increase-replication-factor.json --execute
Current partition replica assignment

验证副本存储计划 

[hui@hadoop103 kafka]$ bin/kafka-reassign-partitions.sh --bootstrap-server hadoop103:9092 --reassignment-json-file increase-replication-factor.json --verify

2,.2  退役旧节点

先按照退役一台节点，生成执行计划，然后按照服役时操作流程执行负载均衡。

创建一个要均衡的主题

[hui@hadoop103 kafka]$ cat topics-to-move.json
{
        "topics": [{
                "topic": "first"
        }],
        "version": 1
}

创建执行计划。

bin/kafka-reassign-partitions.sh --bootstrap-server hadoop103:9092 --topics-to-move-json-file topics-to-move.json --broker-list "0,1,2" --generate

创建副本存储计划（所有副本存储在broker0、broker1、broker2中） 

vim increase-replication-factor.json
{"version":1,"partitions":[{"topic":"first","partition":0,"replicas":[2,0,1],"log_dirs":["any","any","any"]},{"topic":"first","partition":1,"replicas":[0,1,2],"log_dirs":["any","any","any"]},{"topic":"first","partition":2,"replicas":[1,2,0],"log_dirs":["any","any","any"]}]}

执行副本存储计划

bin/kafka-reassign-partitions.sh --bootstrap-server hadoop102:9092 --reassignment-json-file increase-replication-factor.json --execute

验证副本存储计划

bin/kafka-reassign-partitions.sh --bootstrap-server hadoop103:9092 --reassignment-json-file increase-replication-factor.json --verify

 最后停止hadoop102 上的 kafka 即可；

[hui@hadoop102 ~]$ jps
3510 Jps
2795 Kafka
[hui@hadoop102 ~]$ kill -9 2795
[hui@hadoop102 ~]$ jps
3522 Jps