kafka是什么

一言以蔽之：分布式、高吞吐量 的 订阅、发布 消息系统

kafka的四方：producer     broker      consumer                  由 zookeeper协调转发

安装搭建：  

本机环境描述

本机ip：10.10.113.120

操作系统：Linux bogon 2.6.32-358.23.2.el6.i686 #1 SMP Wed Oct 16 17:21:31 UTC 2013 i686 i686 i386 GNU/Linux，CentOS release 6.4 (Final)

节点情况：zookeeper-1个 ，broker-3个

测试点：

• 添加 topic 
• producer 产生消息， 消费者 消费的
• 当非leader broker挂掉时 生产消费情况
• 一个leader broker 挂掉时 生产消费情况

关键步骤

1、下载编译好的    我试过该包提供的 window下bat处理文件，总是提示 classnotfound异常。于是转到centos下。

2、jdk安装什么的自行解决

3、解压后的目录结构

drwxr-xr-x. 3 xxx xxx  4096 Jun 13 10:01 bindrwxr-xr-x. 2 xxx xxx  4096 Jun 12 21:30 configdrwxr-xr-x. 2 xxx xxx  4096 Apr 23 03:26 libs-rw-rw-r--. 1 xxx xxx 11358 Apr 23 02:37 LICENSE -rw-rw-r--. 1 xxx xxx   162 Apr 23 02:37 NOTICE

4、kafka会用到hostname，所以 需要修改操作系统的hostname，否者后面执行kafka的shell命令时会报 unknownhostname的异常 centos的hostname修改方式：

#1 root权限hostname bogon

#2 修改/etc/hosts127.0.0.1       localhost.localdomain localhost::1             localhost6.localdomain6 localhost610.10.113.120 bogon bogon

#3 修改/etc/sysconfig/networkNETWORKING=yesNETWORKING_IPV6=noHOSTNAME=bogon

5、 jvm参数中-XX标识的是实验性参数，kafka用了很多用来优化运行的jvm参数，而你安装的 jdk所带的jvm不一定支持这些参数，比如： -XX:+UseCompressedOops

如果你遇到

Unrecognized VM option '+UseCompressedOops'

的错误，请 在bin/ kafka-run-class.sh中移除相关参数

上述两个步骤比较重要。

6、配置broker server的参数 config/server1.properties

# 关键的四个 配置broker.id=1port=9092host.name=10.10.113.120zookeeper.connect=10.10.113.120:2181

server2.properties 

# 关键的四个 配置broker.id=2port=9093host.name=10.10.113.120zookeeper.connect=10.10.113.120:2181

server3.properties 

# 关键的四个 配置broker.id=3port=9094host.name=10.10.113.120zookeeper.connect=10.10.113.120:2181

测试实验

在解压包的bin同级目录，编写启动zookeeper 和 broker server 的startall.sh，赋予执行权限

#!/bin/bash#start zookeeper servernohup bin/zookeeper-server-start.sh config/zookeeper.properties &#start 3 broker servers JMX_PORT=9997 bin/kafka-server-start.sh config/server1.properties &JMX_PORT=9998 bin/kafka-server-start.sh config/server2.properties &JMX_PORT=9999 bin/kafka-server-start.sh config/server3.properties &

运行 startall.sh，进行测试实验，执行脚本的命令行的使用方法和用 --help显示，每次更新命令行参数都有些改变

1、添加3个topic

bin/kafka-topics.sh --create --topic funckXXX1 --partitions 1 --replication-factor 3  --zookeeper 10.10.113.120:2181bin/kafka-topics.sh --create --topic funckXXX2 --partitions 1 --replication-factor 3  --zookeeper 10.10.113.120:2181bin/kafka-topics.sh --create --topic funckXXX3 --partitions 1 --replication-factor 3  --zookeeper 10.10.113.120:2181

查看已添加的topic

bin/kafka-topics.sh --create --describe --zookeeper 10.10.113.120:2181

如，主broker 为 Leader 2

Topic:funckXXX1 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX1        Partition: 0    Leader: 2       Replicas: 2,3,1 Isr: 2,3,1Topic: funckXXX2 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX2        Partition: 0    Leader: 2       Replicas: 2,3,1 Isr: 2,3,1Topic: funckXXX3 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX3        Partition: 0    Leader: 2       Replicas: 2,3,1 Isr: 2,3,1

说明：

 partiton： partion id，由于此处只有一个partition，因此partition id 为0 leader：

当前负责读写的lead broker id relicas：

当前partition的所有replication broker list isr：

relicas的子集，只包含出于活动状态的broker

2、启动消费者控制台

看看某个topic的消费情况

bin/kafka-console-consumer.sh --topic fuckXXX1 --zookeeper 10.10.113.120:2181 --from-beginning

3、添加消息到 top观察 消费者的消费情况

启动生产者控制台

bin/kafka-console-producer.sh --topic fuckXXX1 --broker-list 10.10.113.120:9092,10.10.113.120:9093,10.10.113.120:9094

在其中输入一些消息 回车，如：

[bogon@bogon kafaka0811]$ bin/kafka-console-producer.sh --topic fuckXXX1 --broker-list 10.10.113.120:9092,110.113.120:9093,10.10.113.120:9094SLF4J: Failed to load class "org.slf4j.impl.StaticLoggerBinder".SLF4J: Defaulting to no-operation (NOP) logger implementationSLF4J: See http://www.slf4j.org/codes.html#StaticLoggerBinder for further details.message1message2message3

那么在消费者端应该有动作：

[bogon@bogon kafaka0811]$ bin/kafka-console-consumer.sh --zookeeper 10.10.113.120:2181 --from-beginning --topic fuckXXX1SLF4J: Failed to load class "org.slf4j.impl.StaticLoggerBinder".SLF4J: Defaulting to no-operation (NOP) logger implementationSLF4J: See http://www.slf4j.org/codes.html#StaticLoggerBinder for further details.[2014-06-13 15:55:16,203] INFO Closing socket connection to /10.10.113.120. (kafka.network.Processor)[2014-06-13 15:56:12,871] INFO Topic creation {"version":1,"partitions":{"1":[3],"0":[2]}} (kafka.admin.AdminUtils$)[2014-06-13 15:56:12,879] INFO [KafkaApi-3] Auto creation of topic fuckXXX1 with 2 partitions and replication factor 1 is successful! (kafka.server.KafkaApis)[2014-06-13 15:56:12,954] INFO [ReplicaFetcherManager on broker 2] Removed fetcher for partitions [fuckXXX1,0] (kafka.server.ReplicaFetcherManager)[2014-06-13 15:56:12,958] INFO [ReplicaFetcherManager on broker 3] Removed fetcher for partitions [fuckXXX1,1] (kafka.server.ReplicaFetcherManager)[2014-06-13 15:56:12,973] INFO Completed load of log fuckXXX1-1 with log end offset 0 (kafka.log.Log)[2014-06-13 15:56:12,985] INFO Created log for partition [fuckXXX1,1] in /tmp/kafka-broke3-logs with properties {segment.index.bytes -> 10485760, file.delete.delay.ms -> 60000, segment.bytes -> 536870912, flush.ms -> 9223372036854775807, delete.retention.ms -> 86400000, index.interval.bytes -> 4096, retention.bytes -> -1, cleanup.policy -> delete, segment.ms -> 604800000, max.message.bytes -> 1000012, flush.messages -> 9223372036854775807, min.cleanable.dirty.ratio -> 0.5, retention.ms -> 604800000}. (kafka.log.LogManager)[2014-06-13 15:56:12,986] WARN Partition [fuckXXX1,1] on broker 3: No checkpointed highwatermark is found for partition [fuckXXX1,1] (kafka.cluster.Partition)[2014-06-13 15:56:12,997] INFO Completed load of log fuckXXX1-0 with log end offset 0 (kafka.log.Log)[2014-06-13 15:56:13,017] INFO Created log for partition [fuckXXX1,0] in /tmp/kafka-broke2-logs with properties {segment.index.bytes -> 10485760, file.delete.delay.ms -> 60000, segment.bytes -> 536870912, flush.ms -> 9223372036854775807, delete.retention.ms -> 86400000, index.interval.bytes -> 4096, retention.bytes -> -1, cleanup.policy -> delete, segment.ms -> 604800000, max.message.bytes -> 1000012, flush.messages -> 9223372036854775807, min.cleanable.dirty.ratio -> 0.5, retention.ms -> 604800000}. (kafka.log.LogManager)[2014-06-13 15:56:13,021] WARN Partition [fuckXXX1,0] on broker 2: No checkpointed highwatermark is found for partition [fuckXXX1,0] (kafka.cluster.Partition)[2014-06-13 15:56:13,059] INFO Closing socket connection to /10.10.113.120. (kafka.network.Processor)[2014-06-13 15:56:13,288] INFO Closing socket connection to /10.10.113.120. (kafka.network.Processor)[2014-06-13 15:56:13,914] INFO Closing socket connection to /10.10.113.120. (kafka.network.Processor)message1message2message3

4、杀掉非leaderbroker 观察

查看所有java进程   : ps axf | grep java

然后kill掉除Leader 标识之外的任一  broker的进程

pkill -9 -f server1.properties

此时topic信息为：

[bogon@bogon kafaka0811]$ bin/kafka-topics.sh --describe  --zookeeper 10.10.113.120:2181Topic:fuckXXX1  PartitionCount:2        ReplicationFactor:1     Configs:Topic: fuckXXX1 Partition: 0    Leader: 2       Replicas: 2     Isr: 2Topic: fuckXXX1 Partition: 1    Leader: 3       Replicas: 3     Isr: 3Topic:funckXXX1 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX1        Partition: 0    Leader: 2       Replicas: 2,3,1 Isr: 2,3Topic:funckXXX2 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX2        Partition: 0    Leader: 2       Replicas: 2,3,1 Isr: 2,3Topic:funckXXX3 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX3        Partition: 0    Leader: 2       Replicas: 2,3,1 Isr: 2,3

此时生产和消费消息依旧正常，请自己发布消息

5、杀掉leader broker 观察

杀掉主broker的进程，然后连续观察 topic详情

pkill -9 -f server2.properties

topic为，可看出 Leader 已经改变为3了

[bogon@bogon kafaka0811]$ bin/kafka-topics.sh --describe  --zookeeper 10.10.113.120:2181Topic:fuckXXX1  PartitionCount:2        ReplicationFactor:1     Configs:Topic: fuckXXX1 Partition: 0    Leader: -1      Replicas: 2     Isr: Topic: fuckXXX1 Partition: 1    Leader: 3       Replicas: 3     Isr: 3Topic:funckXXX1 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX1        Partition: 0    Leader: 3       Replicas: 2,3,1 Isr: 3Topic:funckXXX2 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX2        Partition: 0    Leader: 3       Replicas: 2,3,1 Isr: 3Topic:funckXXX3 PartitionCount:1        ReplicationFactor:3     Configs:Topic: funckXXX3        Partition: 0    Leader: 3       Replicas: 2,3,1 Isr: 3

此时生产和消费消息依旧正常，请自己发布消息

TODO：生产者消费者客户端功能编写