Hadoop HA 集群搭建

集群部署节点角色的规划（3节点）

----------------------------------------------------------------------------------------------------------------

Server01 192.168.2.11:hadoop01:node01 namenode resourcemanager zkfc nodemanager datanode zookeeper journalnode

Server02 192.168.2.12:hadoop02:node02 namenode resourcemanager zkfc nodemanager datanode zookeeper journalnode

Server03 192.168.2.13:hadoop03:node03 datanode nodemanager zookeeper journalnode

----------------------------------------------------------------------------------------------------------------

环境准备

1. 修改Linux主机名
2. 修改主机名
3. 关闭防火墙
4. ssh免登录
5. 安装JDK
6. 配置时间同步
7. 安装zookeeper

安装配置Hadoop集群

将Hadoop解压到指定目录，并添加配置指定jdk目录

[root@node01 hadoop-2.7.7]# tar zxvf /root/soft/hadoop-2.7.7.tar.gz -C /apps/

[root@node01 hadoop-2.7.7]# cd /apps/hadoop-2.7.7/etc/hadoop/

[root@node01 hadoop]# vim hadoop-env.sh

export JAVA_HOME=/apps/jdk1.8.0_151

   

配置core-site.xml文件

[root@node01 hadoop]# vim core-site.xml

    <!-- 集群名称在这里指定！该值来自于hdfs-site.xml中的配置 -->

    <property>

        <name>fs.defaultFS</name>

        <value>hdfs://cluster1</value>

    </property>

    <!-- 这里的路径默认是NameNode、DataNode、JournalNode等存放数据的公共目录 -->

    <property>

        <name>hadoop.tmp.dir</name>

        <value>/data/hadoop/tmp</value>

    </property>

   

    <!-- ZooKeeper集群的地址和端口。注意，数量一定是奇数，且不少于三个节点-->

    <property>

        <name>ha.zookeeper.quorum</name>

        <value>node01:2181,node02:2181,node03:2181</value>

    </property>

   

配置hdfs-site.xml文件

[root@node01 hadoop]# vim hdfs-site.xml

    <!--指定hdfs的nameservice为cluster1，需要和core-site.xml中的保持一致 -->

    <property>

        <name>dfs.nameservices</name>

        <value>cluster1</value>

    </property>

    <!-- cluster1下面有两个NameNode，分别是nn1，nn2 -->

    <property>

        <name>dfs.ha.namenodes.cluster1</name>

        <value>nn1,nn2</value>

    </property>

    <!-- nn1的RPC通信地址 -->

    <property>

        <name>dfs.namenode.rpc-address.cluster1.nn1</name>

        <value>node01:9000</value>

    </property>

    <!-- nn1的http通信地址 -->

    <property>

        <name>dfs.namenode.http-address.cluster1.nn1</name>

        <value>node01:50070</value>

    </property>

    <!-- nn2的RPC通信地址 -->

    <property>

        <name>dfs.namenode.rpc-address.cluster1.nn2</name>

        <value>node02:9000</value>

    </property>

    <!-- nn2的http通信地址 -->

    <property>

        <name>dfs.namenode.http-address.cluster1.nn2</name>

        <value>node02:50070</value>

    </property>

    <!-- 指定NameNode的edits元数据在JournalNode上的存放位置 -->

    <property>

        <name>dfs.namenode.shared.edits.dir</name>

        <value>qjournal://node01:8485;node02:8485;node03:8485/cluster1</value>

    </property>

    <!-- 指定JournalNode在本地磁盘存放数据的位置 -->

    <property>

        <name>dfs.journalnode.edits.dir</name>

        <value>/data/hadoop/journaldata</value>

    </property>

    <!-- 开启NameNode失败自动切换 -->

    <property>

        <name>dfs.ha.automatic-failover.enabled</name>

        <value>true</value>

    </property>

    <!-- 指定该集群出故障时，哪个实现类负责执行故障切换 -->

    <property>

        <name>dfs.client.failover.proxy.provider.cluster1</name>

        <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>

    </property>

    <!-- 配置隔离机制方法，多个机制用换行分割，即每个机制暂用一行-->

    <property>

        <name>dfs.ha.fencing.methods</name>

        <value>sshfence</value>

    </property>

    <!-- 使用sshfence隔离机制时需要ssh免登陆 -->

    <property>

        <name>dfs.ha.fencing.ssh.private-key-files</name>

        <value>/root/.ssh/id_rsa</value>

    </property>

    <!-- 配置sshfence隔离机制超时时间 -->

    <property>

        <name>dfs.ha.fencing.ssh.connect-timeout</name>

        <value>30000</value>

    </property>

   

配置mapred-site.xml文件

[root@node01 hadoop]# cp mapred-site.xml.template mapred-site.xml

[root@node01 hadoop]# vim mapred-site.xml

    <!-- 指定mr框架为yarn方式 -->

    <property>

        <name>mapreduce.framework.name</name>

        <value>yarn</value>

    </property>

   

配置yarn-site.xml文件

[root@node01 hadoop]# vim yarn-site.xml

    <!-- 开启RM高可用 -->

    <property>

        <name>yarn.resourcemanager.ha.enabled</name>

        <value>true</value>

    </property>

    <!-- 指定RM的cluster id -->

    <property>

        <name>yarn.resourcemanager.cluster-id</name>

        <value>yrc</value>

    </property>

    <!-- 指定RM的名字 -->

    <property>

        <name>yarn.resourcemanager.ha.rm-ids</name>

        <value>rm1,rm2</value>

    </property>

    <!-- 分别指定RM的地址 -->

    <property>

        <name>yarn.resourcemanager.hostname.rm1</name>

        <value>node01</value>

    </property>

    <property>

        <name>yarn.resourcemanager.hostname.rm2</name>

        <value>node02</value>

    </property>

    <!-- 指定zk集群地址 -->

    <property>

        <name>yarn.resourcemanager.zk-address</name>

        <value>node01:2181,node02:2181,node03:2181</value>

    </property>

    <property>

        <name>yarn.nodemanager.aux-services</name>

        <value>mapreduce_shuffle</value>

    </property>

   

在文件slaves中配置集群节点

[root@node01 hadoop]# vim slaves

node01

node02

node03

[root@node01 apps]# scp -r hadoop-2.7.7 node02:$PWD

[root@node01 apps]# scp -r hadoop-2.7.7 node03:$PWD

   

配置Hadoop环境变量

[root@node02 sbin]# vim /etc/profile

export HADOOP_HOME=/apps/hadoop-2.7.7

export PATH=$PATH:$HADOOP_HOME/bin:$HADOOP_HOME/sbin

手动启动node01、node01、node03三个节点上的journalnode

[root@node01 sbin]# ./hadoop-daemon.sh start journalnode

格式化namenode

[root@node01 bin]# hdfs namenode -format

[root@node01 data]# scp -r hadoop node02:$PWD

[root@node01 data]# scp -r hadoop node03:$PWD

   

格式化ZKFC（在active上执行）

[root@node01 apps]# hdfs zkfc -formatZK

   

启动HDFS

[root@node01 apps]# start-dfs.sh

   

启动Yarn，并且手动启动standby节点

[root@node01 apps]# start-yarn.sh

[root@node02 sbin]# yarn-daemon.sh start resourcemanager