一、概要

    本文记录hbase高可用集群部署过程，在部署hbase之前需要事先部署好hadoop集群，因为hbase的数据需要存放在hdfs上，hadoop集群的部署后续会有一篇文章记录，本文假设hadoop集群已经部署好，分布式hbase集群需要依赖zk，并且zk可以是hbase自己托管的也可以是我们自己单独搭建的，这里我们使用自己单独搭建的zk集群，我们的hadoop集群是用的cdh的发行版，所以hbase也会使用cdh的源。

二、环境

1、软件版本

centos6

zookeeper-3.4.5+cdh6.9.0+98-1.cdh6.9.0.p0.30.el6.x86_64

hadoop-2.6.0+cdh6.9.0+1799-1.cdh6.9.0.p0.30.el6.x86_64

hbase-1.2.0+cdh6.9.0+205-1.cdh6.9.0.p0.30.el6.x86_64

2、角色

a、zk集群

10.10.20.64:2181
10.10.40.212:2181
10.10.102.207:2181

b、hbase

10.10.40.212 HMaster
10.10.20.64  HMaster
10.10.10.114 HRegionServer
10.10.40.169 HRegionServer
10.10.30.174 HRegionServer

三、部署

1、配置cdh的yum源

vim /etc/yum.repos.d/cloudera-cdh.repo
[cloudera-cdh6]
# Packages for Cloudera's Distribution for Hadoop, Version 5.4.4, on RedHat or CentOS 6 x86_64
name=Cloudera's Distribution for Hadoop, Version 5.4.8
baseurl=http://archive.cloudera.com/cdh6/redhat/6/x86_64/cdh/5.9.0/
gpgkey=http://archive.cloudera.com/cdh6/redhat/6/x86_64/cdh/RPM-GPG-KEY-cloudera    
gpgcheck=1

[cloudera-gplextras5b2]
# Packages for Cloudera's GPLExtras, Version 5.4.4, on RedHat or CentOS 6 x86_64
name=Cloudera's GPLExtras, Version 5.4.8
baseurl=http://archive.cloudera.com/gplextras5/redhat/6/x86_64/gplextras/5.9.0/
gpgkey=http://archive.cloudera.com/gplextras5/redhat/6/x86_64/gplextras/RPM-GPG-KEY-cloudera    
gpgcheck=1

2、安装zk集群（所有zk节点都操作）

1、安装

yum -y install zookeeper zookeeper-server

b、配置

vim /etc/zookeeper/conf/zoo.cfg 

tickTime=2000
initLimit=10
syncLimit=5
dataDir=/data/lib/zookeeper
clientPort=2181
maxClientCnxns=0
server.1=10.10.20.64:2888:3888
server.2=10.10.40.212:2888:3888
server.3=10.10.102.207:2888:3888
autopurge.snapRetainCount=3
autopurge.purgeInterval=1

mkdir -p /data/lib/zookeeper  #建zk的dir目录

echo 1 >/data/lib/zookeeper/myid  #10.10.20.64上操作
echo 2 >/data/lib/zookeeper/myid  #10.10.40.212上操作
echo 3 >/data/lib/zookeeper/myid  #10.10.102.207上操作

c、启动服务

 /etc/init.d/zookeeper-server start

3、安装配置hbase集群

a、安装

yum -y install hbase hbase-master        #HMaster节点操作    
yum -y install hbase hbase-regionserver  #HRegionServer节点操作

b、配置（所有base节点操作）

vim /etc/hbase/conf/hbase-site.xml 




    
        hbase.zookeeper.quorum
        10.10.20.64:2181,10.10.40.212:2181,10.10.102.207:2181
    
    
        hbase.zookeeper.property.clientPort
        2181
    
    
        hbase.zookeeper.property.dataDir
        /data/lib/zookeeper/
    
    
        hbase.rootdir
        hdfs://mycluster:8020/hbase
    
    
        hbase.cluster.distributed
        true
     集群的模式，分布式还是单机模式，如果设置成false的话，HBase进程和Zookeeper进程在同一个JVM进程
     
    

echo "export HBASE_MANAGES_ZK=false" >>/etc/hbase/conf/hbase-env.sh
#设置hbase使用独立的zk集群

vim /etc/hbase/conf/regionservers 
ip-10-10-30-174.ec2.internal
ip-10-10-10-114.ec2.internal
ip-10-10-40-169.ec2.internal
#添加HRegionServer的主机名到regionservers，我没有在/etc/hosts下做主机名的映射，直接用了ec2的默认主机名

c、启动服务

/etc/init.d/hbase-master start       #HMaster节点操作
/etc/init.d/hbase-regionserver start #HRegionServer节点操作

4、验证

a、验证基本功能

[root@ip-10-10-20-64 ~]# hbase  shell  
2017-05-10 16:31:20,225 INFO  [main] Configuration.deprecation: hadoop.native.lib is deprecated. Instead, use io.native.lib.available
HBase Shell; enter 'help' for list of supported commands.
Type "exit" to leave the HBase Shell
Version 1.2.0-cdh6.9.0, rUnknown, Fri Oct 21 01:19:47 PDT 2016

hbase(main):001:0>  status
1 active master, 1 backup masters, 3 servers, 0 dead, 1.3333 average load

hbase(main):002:0> list
TABLE                                                                                                                                                                                        
test                                                                                                                                                                                         
test1                                                                                                                                                                                        
2 row(s) in 0.0330 seconds

=> ["test", "test1"]
hbase(main):003:0> describe 'test'
Table test is ENABLED                                                                                                                                                                        
test                                                                                                                                                                                         
COLUMN FAMILIES DESCRIPTION                                                                                                                                                                  
{NAME => 'id', BLOOMFILTER => 'ROW', VERSIONS => '1', IN_MEMORY => 'false', KEEP_DELETED_CELLS => 'FALSE', DATA_BLOCK_ENCODING => 'NONE', TTL => 'FOREVER', COMPRESSION => 'NONE', MIN_VERSIO
NS => '0', BLOCKCACHE => 'true', BLOCKSIZE => '65536', REPLICATION_SCOPE => '0'}                                                                                                             
{NAME => 'name', BLOOMFILTER => 'ROW', VERSIONS => '1', IN_MEMORY => 'false', KEEP_DELETED_CELLS => 'FALSE', DATA_BLOCK_ENCODING => 'NONE', TTL => 'FOREVER', COMPRESSION => 'NONE', MIN_VERS
IONS => '0', BLOCKCACHE => 'true', BLOCKSIZE => '65536', REPLICATION_SCOPE => '0'}                                                                                                           
{NAME => 'text', BLOOMFILTER => 'ROW', VERSIONS => '1', IN_MEMORY => 'false', KEEP_DELETED_CELLS => 'FALSE', DATA_BLOCK_ENCODING => 'NONE', TTL => 'FOREVER', COMPRESSION => 'NONE', MIN_VERS
IONS => '0', BLOCKCACHE => 'true', BLOCKSIZE => '65536', REPLICATION_SCOPE => '0'}                                                                                                           
3 row(s) in 0.1150 seconds

hbase(main):004:0>

b、验证HA功能

1、hbase默认的web管理端口是60010，两个HMaster谁先启动谁就是主active节点，10.10.40.212先启动，10.10.20.64后启动，web截图如下：

2、停止10.10.40.212的HMaster进程，查看10.10.20.64是否会提升为master

/etc/init.d/hbase-master stop