Hadoop完全高可用集群安装

架构图（HA模型没有SNN节点）

 用vm规划了8台机器，用到了7台，SNN节点没用

	NN	DN	SN	ZKFC	ZK	JNN	RM	NM
node1	*			*
node2	*			*
node3
node4				*			*
node5				*			*
node6		*			*	*		*
node7		*			*	*		*
node8		*			*	*		*

集群搭建前准备工作：

*搭建集群之前需要关闭所有服务器的selinux和防火墙

　　1.更改所有服务器的主机名和hosts文件对应关系

[root@localhost ~]# hostnamectl set-hostname node1

[root@localhost ~]# cat /etc/hosts
127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
192.168.159.129 node1
192.168.159.130 node2
192.168.159.132 node3
192.168.159.133 node4
192.168.159.136 node5
192.168.159.137 node6
192.168.159.138 node7
192.168.159.139 node8

　　2.两个NameNode节点做对所有主机的免密登陆,包括自己的节点；两个resourcemanager节点互相做免密登陆，包括自己的节点

[root@localhost ~]# ssh-keygen
Generating public/private rsa key pair.
Enter file in which to save the key (/root/.ssh/id_rsa):
Created directory '/root/.ssh'.
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in /root/.ssh/id_rsa.
Your public key has been saved in /root/.ssh/id_rsa.pub.
The key fingerprint is:
SHA256:lIvGygyJHycNTZJ0KeuE/BM0BWGGq/UTgMUQNo7Qm2M root@node1
The key's randomart image is:
+---[RSA 2048]----+
|+@=**o           |
|*.XB.    .       |
|oo+*o   o        |
|.+E=.. o .       |
|o=*o+.+ S        |
|...Xoo           |
|  . =.           |
|                 |
|                 |
+----[SHA256]-----+

[root@localhost ~]# for i in `seq 1 8`;do ssh-copy-id root@node$i;done

　　3.同步所有服务器时间

[root@node1 ~]# ansible all -m shell -o -a 'ntpdate ntp1.aliyun.com'
node4 | CHANGED | rc=0 | (stdout) 20 Feb 16:08:37 ntpdate[2477]: adjust time server 120.25.115.20 offset 0.001546 sec
node6 | CHANGED | rc=0 | (stdout) 20 Feb 16:08:37 ntpdate[2470]: adjust time server 120.25.115.20 offset 0.000220 sec
node2 | CHANGED | rc=0 | (stdout) 20 Feb 16:08:37 ntpdate[2406]: adjust time server 120.25.115.20 offset -0.002414 sec
node3 | CHANGED | rc=0 | (stdout) 20 Feb 16:08:37 ntpdate[2465]: adjust time server 120.25.115.20 offset -0.001185 sec
node5 | CHANGED | rc=0 | (stdout) 20 Feb 16:08:37 ntpdate[2466]: adjust time server 120.25.115.20 offset 0.005768 sec
node7 | CHANGED | rc=0 | (stdout) 20 Feb 16:08:43 ntpdate[2503]: adjust time server 120.25.115.20 offset 0.000703 sec
node8 | CHANGED | rc=0 | (stdout) 20 Feb 16:08:43 ntpdate[2426]: adjust time server 120.25.115.20 offset -0.001338 sec

　　4.所有服务器安装JDK环境并配置好环境变量

[root@node1 ~]# tar -xf jdk-8u144-linux-x64.gz -C /usr/
[root@node1 ~]# ln -sv /usr/jdk1.8.0_144/ /usr/java
"/usr/java" -> "/usr/jdk1.8.0_144/"

[root@node1 ~]# cat /etc/profile.d/java.sh
export JAVA_HOME=/usr/java
export PATH=$PATH:$JAVA_HOME/bin

[root@node1 ~]# source /etc/profile.d/java.sh
[root@node1 ~]# java -version                 
java version "1.8.0_144"
Java(TM) SE Runtime Environment (build 1.8.0_144-b01)
Java HotSpot(TM) 64-Bit Server VM (build 25.144-b01, mixed mode)　　 

zookeeper集群搭建

在规划好的6、7、8节点上安装zookeeper（JDK环境要准备好）

#解压zookeeper程序到/usr目录下
[root@node6 ~]# tar xf zookeeper-3.4.6.tar.gz -C /usr/
#创建zookeeper存放数据目录
[root@node6 ~]# mkdir /usr/data/zookeeper
#将zookeeper的conf目录下sample配置文件更改成cfg文件
[root@node6 ~]# cp /usr/zookeeper-3.4.6/conf/zoo_sample.cfg  /usr/zookeeper-3.4.6/conf/zoo.cfg
#编辑配置文件，更改数据存放目录，并添加zookeeper集群配置信息
[root@node6 ~]# vim /usr/zookeeper-3.4.6/conf/zoo.cfg
dataDir=/usr/data/zookeeper   #修改
server.1=node6:2888:3888      #添加
server.2=node7:2888:3888      #添加
server.3=node8:2888:3888      #添加

#把配置好的zookeeper程序文件分发至其余的两个节点
[root@node6 ~]# scp -r /usr/zookeeper-3.4.6/ node7:/usr/zookeeper-3.4.6/
[root@node6 ~]# scp -r /usr/zookeeper-3.4.6/ node8:/usr/zookeeper-3.4.6/

#在刚刚创建的目录下当前zookeeper节点信息，必须为数字，且三个节点不能相同
[root@node6 ~]# echo 1 > /usr/data/zookeeper/myid

#在剩下的两个节点上也要创建数据存放目录和节点配置文件
[root@node7 ~]# mkdir /usr/data/zookeeper
[root@node7 ~]# echo 2 > /usr/data/zookeeper/myid
[root@node8 ~]# mkdir /usr/data/zookeeper
[root@node8 ~]# echo 3 > /usr/data/zookeeper/myid
#配置完成后启动zookeeper集群
[root@node6 ~]# /usr/zookeeper-3.4.6/bin/zkServer.sh start
[root@node7 ~]# /usr/zookeeper-3.4.6/bin/zkServer.sh start
[root@node8 ~]# /usr/zookeeper-3.4.6/bin/zkServer.sh start

#查看集群启动情况（先启动的会成为leader，同时启动数字大的会成为leader）
[root@node6 ~]# /usr/zookeeper-3.4.6/bin/zkServer.sh status
JMX enabled by default
Using config: /usr/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode: follower

[root@node7 ~]# /usr/zookeeper-3.4.6/bin/zkServer.sh status
JMX enabled by default
Using config: /usr/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode: follower

[root@node8 ~]# /usr/zookeeper-3.4.6/bin/zkServer.sh status
JMX enabled by default
Using config: /usr/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode: leader

[root@node8 ~]# netstat -tnlp | grep java   #只有主节点有2888
tcp6       0      0 :::2181                 :::*                    LISTEN      33766/java          
tcp6       0      0 192.168.159.139:2888    :::*                    LISTEN      33766/java          
tcp6       0      0 192.168.159.139:3888    :::*                    LISTEN      33766/java          
tcp6       0      0 :::43793                :::*                    LISTEN      33766/java         

Hadoop集群搭建　　

　　1.先添加hadoop的环境变量

[root@node1 ~]# cat /etc/profile.d/hadoop.sh
export HADOOP_HOME=/usr/hadoop-2.9.2
export PATH=$PATH:$HADOOP_HOME/bin:$HADOOP_HOME/sbin

　　2.解压hadoop程序包到/usr目录下

[root@node1 ~]# tar xf hadoop-2.9.2.tar.gz -C /usr
[root@node1 ~]# ln -sv /usr/hadoop-2.9.2/ /usr/hadoop
"/usr/hadoop" -> "/usr/hadoop-2.9.2/"

　　3.更改hadoop程序包内 hadoop-env.sh，mapred-env.sh，yarn-env.sh中的JAVA_HOME环境变量

[root@node1 ~]# grep 'export JAVA_HOME' /usr/hadoop/etc/hadoop/{hadoop-env.sh,mapred-env.sh,yarn-env.sh}
/usr/hadoop/etc/hadoop/hadoop-env.sh:export JAVA_HOME=/usr/java
/usr/hadoop/etc/hadoop/mapred-env.sh:export JAVA_HOME=/usr/java
/usr/hadoop/etc/hadoop/yarn-env.sh:export JAVA_HOME=/usr/java

　　4.修改core-site.xml文件（NameNode配置文件）

[root@node1 ~]# vim /usr/hadoop/etc/hadoop/core-site.xml
<configuration>
    <property>  
        <name>fs.defaultFS</name>  
        <value>hdfs://hadoop</value>
        <!--HA部署下，NameNode访问hdfs-site.xml中的dfs.nameservices值 -->  
    </property>  
    <property>  
        <name>hadoop.tmp.dir</name>  
        <value>/usr/data/hadoop</value>
        <!--Hadoop的文件存放目录 -->  
    </property>  
    <property>  
        <name>ha.zookeeper.quorum</name>  
        <value>node6:2181,node7:2181,node8:2181</value>
        <!--zookeeper集群地址 -->  
    </property>
</configuration>

　　5.在所有hadoop节点创建/usr/data/hadoop目录

　　6.修改hdfs-site.xml文件

[root@node1 ~]# vim /usr/hadoop/etc/hadoop/hdfs-site.xml
<configuration>
    <property>  
        <name>dfs.replication</name>  
        <value>3</value>  
        <!--数据文件副本数量-->
    </property> 
    <property>  
        <name>dfs.blocksize</name>  
        <value>134217728</value>
        <!--数据块大小，文件超过这个大小就会切开，128M -->
    </property>  
    <property>  
        <name>dfs.permissions.enabled</name>  
        <value>false</value>
        <!--  ****   -->  
    </property>  
    <property>  
        <name>dfs.nameservices</name>  
        <value>hadoop</value>
        <!--这个值就是core-site.xml中hdfs集群入口 -->  
    </property>  
    <property>  
        <name>dfs.ha.namenodes.hadoop</name>  
        <value>nn1,nn2</value>
        <!--集群中一共有两个namenode -->  
    </property>  
    <property>  
        <name>dfs.namenode.rpc-address.hadoop.nn1</name>  
        <value>node1:9000</value>
        <!--nn1的rpc通信地址 -->
    </property>  
    <property>  
        <name>dfs.namenode.http-address.hadoop.nn1</name>  
        <value>node1:50070</value>
        <!--nn1的http通信地址 -->  
    </property>  
    <property>  
        <name>dfs.namenode.rpc-address.hadoop.nn2</name>  
        <value>node2:9000</value>
        <!--nn2的rpc通信地址 -->
    </property>  
    <property>  
        <name>dfs.namenode.http-address.hadoop.nn2</name>  
        <value>node2:50070</value>
        <!--nn2的http通信地址 -->
    </property>  
    <property>  
        <name>dfs.namenode.shared.edits.dir</name>  
        <value>qjournal://node6:8485;node7:8485;node8:8485/hadoop</value> 
        <!-- 指定NameNode的元数据在JournalNode日志上的存放位置(一般和zookeeper部署在一起)-->
        <!-- 存储路径可以随便起,如果有多个集群,不一样就行-->   
    </property> 
    <property>  
        <name>dfs.ha.automatic-failover.enabled</name>  
        <value>true</value>
        <!--是否开启故障自动隔离-->  
    </property> 
    <property>  
        <name>dfs.journalnode.edits.dir</name>  
        <value>/usr/data/journalnode</value> 
        <!-- 指定JournalNode在本地磁盘存放数据的位置,这个需要指定,默认是放在tmp目录下 -->   
    </property>  
    <property>  
        <name>dfs.client.failover.proxy.provider.hadoop</name>  
        <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
        <!-- namenode故障转移实现的代理类,注意"name键"要改动-->
    </property>  
    <property>  
        <name>dfs.ha.fencing.methods</name>  
        <value>sshfence</value>
        <!--故障自动转移的方法，这里选用ssh远程登陆方法-->
    </property>  
    <property>  
        <name>dfs.ha.fencing.ssh.private-key-files</name>  
        <value>/root/.ssh/id_rsa</value>
        <!--选用了ssh远程登陆就需要ssh密钥,两台namenode需要互相做密钥认证-->
    </property>  
    <property>  
        <name>dfs.ha.fencing.ssh.connect-timeout</name>  
        <value>30000</value>
        <!--配置ssh超时时间-->  
    </property>  
</configuration>

　　7.在journalnode节点创建/usr/data/journalnode目录

　　8.修改mapred-site.xml（ 修改mapred-site.xml.template名称为mapred-site.xml）

[root@node1 ~]# vim /usr/hadoop/etc/hadoop/mapred-site.xml
<configuration>
    <property>  
        <name>mapreduce.framework.name</name>  
        <value>yarn</value>  
    </property>  
    <property>  
        <name>mapreduce.jobhistory.address</name>  
        <value>node3:10020</value>  
    </property>  
    <property>  
        <name>mapreduce.jobhistory.webapp.address</name>  
        <value>node3:19888</value>  
    </property>  
</configuration>

　　9.修改yarn-site.xml

[root@node1 ~]# vim /usr/hadoop/etc/hadoop/yarn-site.xml
<configuration>
    <property>  
        <name>yarn.nodemanager.aux-services</name>  
        <value>mapreduce_shuffle</value>  
    </property>
    <property>  
       <name>yarn.resourcemanager.ha.enabled</name>  
       <value>true</value> 
       <!--是否开启rm的高可用--> 
    </property>  
    <property>  
       <name>yarn.resourcemanager.cluster-id</name>  
       <value>rmcluster</value>
       <!--生成rm集群的唯一标识,name键不需要改动 -->
    </property>  
    <property>  
       <name>yarn.resourcemanager.ha.rm-ids</name>  
       <value>rm1,rm2</value>
       <!--rm集群的两台机器名称 -->  
    </property>  
    <property>  
       <name>yarn.resourcemanager.hostname.rm1</name>  
       <value>node4</value>
       <!--rm1的机器地址 -->
    </property>
    <property>
       <name>yarn.resourcemanager.webapp.address.rm1</name>
       <value>node4:8088</value>
       <!--rm1的网页访问地址 --> 
    </property>
    <property>  
       <name>yarn.resourcemanager.hostname.rm2</name>  
       <value>node5</value>
       <!--rm2的机器地址 -->
    </property>  
    <property>
       <name>yarn.resourcemanager.webapp.address.rm2</name>
       <value>node5:8088</value>
       <!--rm2的网页访问地址 -->
    </property>
    <property>  
       <name>yarn.resourcemanager.zk-address</name>  
       <value>node6:2181,node7:2181,node8:2181</value> 
       <!--指定zookeeper集群的地址--> 
    </property>  
    <property>  
       <name>yarn.resourcemanager.recovery.enabled</name>  
       <value>true</value>  
       <!--启用自动恢复,默认是false-->   
    </property>  
    <property>  
       <name>yarn.resourcemanager.store.class</name>  
       <value>org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore</value> 
       <!--指定resourcemanager的状态信息存储在zookeeper集群，默认是存放在FileSystem里--> 
    </property>
</configuration>

　　10.编辑datanode配置文件（也是nodemanager的启动位置）

[root@node1 ~]# vim /usr/hadoop/etc/hadoop/slaves
node6
node7
node8

仅首次初始化时需要的步骤如下：

　　1.首先启动三台journalnode集群

[root@node6 ~]# hadoop-daemon.sh start journalnode
starting journalnode, logging to /usr/hadoop-2.9.2/logs/hadoop-root-journalnode-node6.out
[root@node6 ~]# jps
2965 Jps
2904 JournalNode
2779 QuorumPeerMain

[root@node7 ~]# hadoop-daemon.sh start journalnode
starting journalnode, logging to /usr/hadoop-2.9.2/logs/hadoop-root-journalnode-node7.out
[root@node7 ~]# jps
2119 QuorumPeerMain
2220 JournalNode
2318 Jps

[root@node8 ~]# hadoop-daemon.sh start journalnode
starting journalnode, logging to /usr/hadoop-2.9.2/logs/hadoop-root-journalnode-node8.out
[root@node8 ~]# jps
2229 Jps
2025 QuorumPeerMain
2153 JournalNode

　　2.格式化NameNode主节点

[root@node1 ~]# hadoop namenode -format

 　　3.启动NameNode主节点

[root@node1 ~]# hadoop-daemon.sh start namenode
starting namenode, logging to /usr/hadoop-2.9.2/logs/hadoop-root-namenode-node1.out
[root@node1 ~]# jps
7302 Jps
7225 NameNode

　　4.格式化NameNode从节点

[root@node2 ~]# hadoop namenode -bootstrapStandby

 　　5.NameNode主节点向zookeeper提交初始化节点信息

[root@node1 ~]# hdfs zkfc -formatZK

 　　5.1可以在zookeeper节点上使用zkCli.sh命令查看hdfs信息

[root@node6 ~]# /usr/zookeeper-3.4.6/bin/zkCli.sh
Connecting to localhost:2181
......
......
[zk: localhost:2181(CONNECTED) 0] ls /
[zookeeper]     #namenode还没提交信息的时候
[zk: localhost:2181(CONNECTED) 1] ls /
[zookeeper, hadoop-ha]     #执行了上面那个提交命令
[zk: localhost:2181(CONNECTED) 2] ls /hadoop-ha/hadoop
[]

　　6.启动HDFS集群

[root@node1 ~]# start-dfs.sh
Starting namenodes on [node1 node2]
node2: starting namenode, logging to /usr/hadoop-2.9.2/logs/hadoop-root-namenode-node2.out
node1: namenode running as process 7225. Stop it first.
node8: starting datanode, logging to /usr/hadoop-2.9.2/logs/hadoop-root-datanode-node8.out
node6: starting datanode, logging to /usr/hadoop-2.9.2/logs/hadoop-root-datanode-node6.out
node7: starting datanode, logging to /usr/hadoop-2.9.2/logs/hadoop-root-datanode-node7.out
Starting journal nodes [node6 node7 node8]
node6: journalnode running as process 2904. Stop it first.
node7: journalnode running as process 2220. Stop it first.
node8: journalnode running as process 2153. Stop it first.
Starting ZK Failover Controllers on NN hosts [node1 node2]
node2: starting zkfc, logging to /usr/hadoop-2.9.2/logs/hadoop-root-zkfc-node2.out
node1: starting zkfc, logging to /usr/hadoop-2.9.2/logs/hadoop-root-zkfc-node1.out

[root@node1 ~]# jps
7857 DFSZKFailoverController
7924 Jps
7225 NameNode

[root@node2 ~]# jps
2788 Jps
2633 NameNode
2732 DFSZKFailoverController

[root@node6 ~]# jps
3235 Jps
3125 DataNode
2904 JournalNode
2779 QuorumPeerMain

[root@node7 ~]# jps
2119 QuorumPeerMain
2220 JournalNode
2572 Jps
2462 DataNode

[root@node8 ~]# jps
2483 Jps
2373 DataNode
2025 QuorumPeerMain
2153 JournalNode

　　7.此时zookeeper上就会有namenode的信息了，只存储主节点信息

以上HDFS高可用集群初始化完成，下面启动yarn集群

　　1.在namenode主节点上开启yarn集群，start-yarn.sh命令仅可以启动nodemanager，resourcemanager需要在对应节点上手动启动

[root@node1 ~]# start-yarn.sh
starting yarn daemons
starting resourcemanager, logging to /usr/hadoop-2.9.2/logs/yarn-root-resourcemanager-node1.out
node7: starting nodemanager, logging to /usr/hadoop-2.9.2/logs/yarn-root-nodemanager-node7.out
node8: starting nodemanager, logging to /usr/hadoop-2.9.2/logs/yarn-root-nodemanager-node8.out
node6: starting nodemanager, logging to /usr/hadoop-2.9.2/logs/yarn-root-nodemanager-node6.out

[root@node6 ~]# jps
3125 DataNode
3397 NodeManager
3509 Jps
2904 JournalNode
2779 QuorumPeerMain

[root@node7 ~]# jps
2736 NodeManager
2848 Jps
2119 QuorumPeerMain
2220 JournalNode
2462 DataNode

[root@node8 ~]# jps
2373 DataNode
2646 NodeManager
2758 Jps
2025 QuorumPeerMain
2153 JournalNode

　　2.在resourcemanager节点手动启动rm

[root@node4 ~]# yarn-daemon.sh start resourcemanager
starting resourcemanager, logging to /usr/hadoop-2.9.2/logs/yarn-root-resourcemanager-node4.out
[root@node4 ~]# jps
2840 ResourceManager
3103 Jps

[root@node5 ~]# yarn-daemon.sh start resourcemanager
starting resourcemanager, logging to /usr/hadoop-2.9.2/logs/yarn-root-resourcemanager-node5.out
[root@node5 ~]# jps
2994 Jps
2955 ResourceManager

以后在启动集群只需要

在namenode节点

start-dfs.sh
start-yarn.sh

在resourcemanager节点

yarn-daemon.sh start resourcemanager