Yarn ResourceManager高可用集群搭建

配置在Zookeeper之上，进行故障转移

单节点低可用配置：

进入mapred-site.xml添加

<property>

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

        <value>yarn</value>

</property>

进入 yarn-site.xml添加

    <property>

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

        <value>mapreduce_shuffle</value>

    </property>

高可用配置：

进入yarn-site.xml添加

<property>

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

   <value>true</value>

</property>

<property>

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

   <value>cluster1</value>

</property>

<property>

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

   <value>rm1,rm2</value>

</property>

<property>

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

   <value>bigdata03</value>

</property>

<property>

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

   <value>bigdata04</value>

</property>

<property>

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

   <value>bigdata02:2181,bigdata03:2181,bigdata04:2181</value>

</property>

分发给其他的2、3、4机带有zookeeper的机中

scp mapred-site.xml yarn-site.xml bigdata02:`pwd`

3、4台机中进行免密钥登陆

ssh-keygen -t dsa -P '' -f ~/.ssh/id_dsa

cat ~/.ssh/id_dsa.pub >> ~/.ssh/authorized_keys

3发送密钥到4机中，并追加到authorized_keys中

scp id_dsa.pub bigdata04:`pwd`/bigdata03.pub

cat ~/.ssh/bigdata03.pub >> ~/.ssh/authorized_keys

重复上述操作，让4可以直接访问3，从而实现两者之间的免密钥登陆

先要启动zookeeper

zkServer.sh start

启动hdfs

start-dfs.sh

启动yarn

start-yarn.sh

在3、4机上启动ResourceManager

yarn-daemon.sh start resourcemanager

 

测试wordcount

cd /opt/modules/hadoop-2.6.5/share/hadoop/mapreduce

hadoop jar hadoop-mapreduce-examples-2.6.5.jar wordcount /user/root/test.txt /wordcount

注意：要保持输出路径不存在

linux中查看输出结果文件

hdfs dfs -tail /wordcount/part-r-00000

停止顺序

yarn-daemon.sh stop resourcemanager （停止resourcemanager）

stop-dfs.sh

stop-yarn.sh (停止nodemanager)

zkServer.sh stop

 

踩坑经历：shell脚本语言对格式要求真的 很严格，tab键绝不能用空格来补全