wget http://mirror.bit.edu.cn/apache/kafka/2.1.0/kafka_2.12-2.1.0.tgz  #下载kafka_2.12-2.1.0.tgz

mkdir /usr/local/kafka   #kafka安装文件放置目录

tar xzf kafka_2.-2.1..tgz          #解压kafka二进制文件

mv kafka_2.-2.1. /usr/local/kafka  #移动解压后的kafka到/usr/local/kafka

cd /usr/local/kafka/kafka_2.-2.10/config   #定位到kafka_2.-.10/config目录，以便后面进行参数配置

 ls -l /usr/local/kafka/kafka_2.-2.1./config

total

-rw-r--r--  root root   Nov  : connect-console-sink.properties

-rw-r--r--  root root   Nov  : connect-console-source.properties

-rw-r--r--  root root  Nov  : connect-distributed.properties

-rw-r--r--  root root   Nov  : connect-file-sink.properties

-rw-r--r--  root root   Nov  : connect-file-source.properties

-rw-r--r--  root root  Nov  : connect-log4j.properties

-rw-r--r--  root root  Nov  : connect-standalone.properties

-rw-r--r--  root root  Nov  : consumer.properties

-rw-r--r--  root root  Nov  : log4j.properties

-rw-r--r--  root root  Nov  : producer.properties

-rw-r--r--  root root  Nov  : server.properties

-rw-r--r--  root root  Nov  : tools-log4j.properties

-rw-r--r--  root root  Nov  : trogdor.conf

-rw-r--r--  root root  Nov  : zookeeper.properties

vim /usr/local/kafka/kafka_2.-2.1./config/server.properties

#当前机器在集群中的唯一标识，和zookeeper的myid性质一样

broker.id=  

#当前kafka对外提供服务的端口默认是9092

port= 

#这个参数默认是关闭的，在0..1有个bug，DNS解析问题，失败率的问题。

host.name=192.168.1.101 

#这个是borker进行网络处理的线程数

num.network.threads= 

#这个是borker进行I/O处理的线程数

num.io.threads= 

#消息存放的目录，这个目录可以配置为“，”逗号分割的表达式，上面的num.io.threads要大于这个目录的个数这个目录，如果配置多个目录，新创建的topic他把消息持久化的地方是，当前以逗号分割的目录中，那个分区数最少就放那一个

log.dirs=/usr/local/kafka/kafka_2.-2.1./kafkalogs/ 

#发送缓冲区buffer大小，数据不是一下子就发送的，先回存储到缓冲区了到达一定的大小后在发送，能提高性能

socket.send.buffer.bytes= 

#kafka接收缓冲区大小，当数据到达一定大小后在序列化到磁盘

socket.receive.buffer.bytes= 

#这个参数是向kafka请求消息或者向kafka发送消息的请请求的最大数，这个值不能超过java的堆栈大小

socket.request.max.bytes= 

#默认的分区数，一个topic默认1个分区数

num.partitions= 

#默认消息的最大持久化时间，168小时，7天

log.retention.hours= 

#消息保存的最大值5M

message.max.byte=  

#kafka保存消息的副本数，如果一个副本失效了，另一个还可以继续提供服务

default.replication.factor=

#取消息的最大直接数

replica.fetch.max.bytes=  

#这个参数是：因为kafka的消息是以追加的形式落地到文件，当超过这个值的时候，kafka会新起一个文件

log.segment.bytes= 

#每隔300000毫秒去检查上面配置的log失效时间（log.retention.hours= ），到目录查看是否有过期的消息如果有，删除

log.retention.check.interval.ms= 

#是否启用log压缩，一般不用启用，启用的话可以提高性能

log.cleaner.enable=false 

#设置zookeeper的连接端口

zookeeper.connect=192.168.1.101:,192.168.1.102:,192.168.1.103: 

#broker.id=  每台服务器的broker.id都不能相同

#hostname

host.name=192.168.1.101

#在log.retention.hours= 下面新增下面三项

message.max.byte=

default.replication.factor=

replica.fetch.max.bytes=

#设置zookeeper的连接端口

zookeeper.connect=192.168.1.101:,192.168.1.102:,192.168.1.103: 

scp -r /usr/local/kafka/ 192.168.1.102:/usr/local #将目录/usr/local/kafka/复制到192.168.1.102远程机器上去

scp -r /usr/local/kafka/ 192.168.1.103:/usr/local #将目录/usr/local/kafka/复制到192.168.1.103远程机器上去

vim  /etc/profile

#kafka

export KAFKA_HOME=/usr/local/kafka/kafka_2.-2.1.

export PATH=$KAFKA_HOME/bin:$PATH

source  /etc/profile #使环境变量生效

cd /usr/local/kafka/kafka_2.-2.1.  #定位到kafka的根目录

./bin/kafka-server-start.sh -daemon  ./config/server.properties #启动kafka服务,-daemon表示在后台运行kafka服务器

#或者直接使用绝对地址启动
$KAFKA_HOME/bin/kafka-server-start.sh -daemon  $KAFKA_HOME/config/server.properties

[root@ kafka_2.-2.1.]# jps

 NameNode

 NodeManager

 Kafka

 PaloFe

 QuorumPeerMain

 BrokerBootstrap

 Jps

 DataNode

[root@ kafka_2.-2.1.]# jps

 QuorumPeerMain

 NodeManager

 ResourceManager

 Kafka

 PaloFe

 Jps

 BrokerBootstrap

 SecondaryNameNode

 DataNode

[root@ kafka_2.-2.1.]# jps

 Jps

 RunJar

 RunJar

 PaloFe

 BrokerBootstrap

 NodeManager

 QuorumPeerMain

 ManagerBootStrap

 Kafka

 DataNode

#创建Topic

$KAFKA_HOME/bin/kafka-topics.sh --create --zookeeper 192.168.1.101: --replication-factor  --partitions  --topic lenmom

#解释

--replication-factor    #复制两份

--partitions  #创建1个分区

--topic #主题为shuaige

#在一台服务器（192.168.1.101)上创建一个发布者

#创建一个broker，发布者

$KAFKA_HOME/bin/kafka-console-producer.sh --broker-list 192.168.1.101: --topic lenmom

#在一台服务器(192.168.1.102）上创建一个订阅者

#kafka 1.2.2以前版本使用     kafka-console-consumer.sh --zookeeper localhost: --topic lenmom --from-beginning

#kafka 1.2.2及之后版本是使用 kafka-console-consumer.sh --bootstrap-server localhost: --topic lenmom --from-beginning

$KAFKA_HOME/bin/kafka-console-consumer.sh --bootstrap-server localhost: --topic lenmom --from-beginning

[root@ kafka_2.-2.1.]# $KAFKA_HOME/bin/kafka-topics.sh --create --zookeeper 192.168.1.101: --replication-factor  --partitions  --topic lenmom

Created topic "lenmom".

[root@ kafka_2.-2.1.]# $KAFKA_HOME/bin/kafka-console-producer.sh --broker-list 192.168.1.101: --topic lenmom

>

>

>

>

>

>

>

>

>

>

[root@ kafka_2.-2.1.]# $KAFKA_HOME/bin/kafka-console-consumer.sh --bootstrap-server localhost: --topic lenmom --from-beginning

$KAFKA_HOME/bin/kafka-topics.sh --list --zookeeper localhost:2181   #显示所有主题

__consumer_offsets

lenmom

$KAFKA_HOME/bin/kafka-topics.sh --describe --zookeeper localhost: --topic lenmom

Topic:lenmom    PartitionCount:        ReplicationFactor:     Configs:

        Topic: lenmom   Partition:     Leader:        Replicas: ,   Isr: ,

#分区为为1  复制因子为2   他的  shuaige的分区为0

#Replicas: ,   复制的为0，

$KAFKA_HOME/bin/kafka-topics.sh --zookeeper localhost: --delete --topic lenmom

Topic lenmom is marked for deletion.

Note: This will have no impact if delete.topic.enable is not set to true.

$KAFKA_HOME/bin/kafka-producer-perf-test.sh --topic lenmom --num-record  --record-size  --throughput  --producer-props  bootstrap.servers=192.168.1.101:,192.168.1.102:,192.168.1.103: acks=-

 records sent, 3.4 records/sec (0.00 MB/sec), 27.6 ms avg latency, 357.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 5.8 ms avg latency, 7.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 5.3 ms avg latency, 6.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 5.3 ms avg latency, 6.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 5.2 ms avg latency, 9.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 5.3 ms avg latency, 9.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 4.9 ms avg latency, 6.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 4.9 ms avg latency, 5.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 5.0 ms avg latency, 6.0 max latency.

 records sent, 3.0 records/sec (0.00 MB/sec), 4.5 ms avg latency, 5.0 max latency.

server.log #kafka的运行日志

state-change.log  #kafka他是用zookeeper来保存状态，所以他可能会进行切换，切换的日志就保存在这里

controller.log #kafka选择一个节点作为“controller”,当发现有节点down掉的时候它负责在游泳分区的所有节点中选择新的leader,这使得Kafka可以批量的高效的管理所有分区节点的主从关系。如果controller down掉了，活着的节点中的一个会备切换为新的controller.

#使用客户端进入zk

./zkCli.sh -server 127.0.0.1:

#查看目录情况 执行“ls /”

[zk: 127.0.0.1:(CONNECTED) ] ls /

#显示结果：[cluster, controller_epoch, controller, brokers, zookeeper, admin, isr_change_notification, consumers, log_dir_event_notification, latest_producer_id_block, config]

#上面的显示结果中：只有zookeeper是，zookeeper原生的，其他都是Kafka创建的

#标注一个重要的

[zk: 127.0.0.1:(CONNECTED) ] get /brokers/ids/

{"listener_security_protocol_map":{"PLAINTEXT":"PLAINTEXT"},"endpoints":["PLAINTEXT://192.168.1.103:9092"],"jmx_port":-,"host":"192.168.1.103","timestamp":"","port":,"version":}

cZxid = 0x100000063

ctime = Tue Jan  :: CST

mZxid = 0x100000063

mtime = Tue Jan  :: CST

pZxid = 0x100000063

cversion =

dataVersion =

aclVersion =

ephemeralOwner = 0x20000d319500002

dataLength =

numChildren =

[zk: 127.0.0.1:(CONNECTED) ] 

#还有一个是查看partion

[zk: 127.0.0.1:(CONNECTED) ] get /brokers/topics/lenmom/partitions/

null

cZxid = 0x10000006b

ctime = Tue Jan  :: CST

mZxid = 0x10000006b

mtime = Tue Jan  :: CST

pZxid = 0x10000006c

cversion =

dataVersion =

aclVersion =

ephemeralOwner = 0x0

dataLength =

numChildren =