4、安装部署Spark(Spark on Yarn模式)

4.1使用下面的命令，解压Spark安装包到用户根目录：

[[email protected] ~]$ cd /home/zkpk/tgz/spark/
[[email protected] spark]$ tar -xzvf spark-2.1.1-bin-hadoop2.7.tgz -C /home/zkpk/
[[email protected] spark]$ cd
[[email protected] ~]$ cd spark-2.1.1-bin-hadoop2.7/
[[email protected] spark-2.1.1-bin-hadoop2.7]$ ls -l

执行ls -l命令会看到下面的图片所示内容，这些内容是Spark包含的文件：

4.2配置Hadoop环境变量

4.2.1在Yarn上运行Spark需要配置HADOOP_CONF_DIR、YARN_CONF_DIR和HDFS_CONF_DIR环境变量

4.2.1.1命令：

[[email protected] ~]$ cd
[[email protected] ~]$ gedit ~/.bash_profile

4.2.1.2在文件末尾添加如下内容；保存、退出

#SPARK ON YARN
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
export HDFS_CONF_DIR=$HADOOP_HOME/etc/hadoop
export YARN_CONF_DIR=$HADOOP_HOME/etc/hadoop

4.2.1.3重新编译文件，使环境变量生效

[[email protected] ~]$ source ~/.bash_profile

4.3验证Spark安装

4.3.1修改${HADOOP_HOME}/etc/Hadoop/yarn-site.xml；

说明：在master和slave01、slave02节点都要如此修改此文件

4.3.2添加两个property

[[email protected] ~]$ vim ~/hadoop-2.7.3/etc/hadoop/yarn-site.xml

    <property>
        <name>yarn.nodemanager.pmem-check-enabled</name>
        <value>false</value>
    </property>
    <property>
        <name>yarn.nodemanager.vmem-check-enabled</name>
        <value>false</value>
    </property>

![在这里插入图片描述](https://img-blog.csdnimg.cn/30b25836994545c191442ff18f227621.png)

4.4重启hadoop集群（使配置生效）

[[email protected] ~]$ stop-all.sh
[[email protected] ~]$ start-all.sh

4.5进入Spark安装主目录

[[email protected] ~]$ cd ~/spark-2.1.1-bin-hadoop2.7

4.5.1执行下面的命令（注意这是1行代码）：

[[email protected] spark-2.1.1-bin-hadoop2.7]$　 ./bin/spark-submit --class org.apache.spark.examples.SparkPi --master yarn --num-executors 3 --driver-memory 1g --executor-memory 1g --executor-cores 1 examples/jars/spark-examples*.jar 10

4.5.2执行命令后会出现如下界面：

4.5.3Web UI验证

4.5.3.1进入spark-shell交互终端，命令如下：

[[email protected] spark-2.1.1-bin-hadoop2.7]$　./bin/spark-shell

4.5.3.2打开浏览器，输入下面地址，查看运行界面（地址：http://master:4040/）

4.5.3.3退出交互终端，按ctrl+d组合键

scala> :quit

4.6安装部署Spark-SQL

4.6.1将hadoop安装目录下的hdfs-site.xml文件复制到spark安装目录下的conf目录下

[[email protected] spark-2.1.1-bin-hadoop2.7]$ cd
[[email protected] ~]$ cd hadoop-2.7.3/etc/hadoop/ 
[[email protected] hadoop]$ cp hdfs-site.xml /home/zkpk/spark-2.1.1-bin-hadoop2.7/conf

4.6.2将Hive安装目录conf子目录下的hive-site.xml文件，拷贝到spark的配置子目录

[[email protected] hadoop]$ cd
[[email protected] ~]$ cd apache-hive-2.1.1-bin/conf/
[[email protected] conf]$ cp hive-site.xml /home/zkpk/spark-2.1.1-bin-hadoop2.7/conf/

4.6.3修改spark配置目录中的hive-site.xml文件

[[email protected] conf]$ cd
[[email protected] ~]$ cd spark-2.1.1-bin-hadoop2.7/conf/
[[email protected] conf]$ vim hive-site.xml

4.6.3.1添加如下属性

<property>
          <name>hive.metastore.warehouse.dir</name>
          <value>/user/spark/warehouse</value>
 </property>

4.6.4将mysql连接的驱动包拷贝到spark目录的jars子目录

[[email protected] conf]$ cd
[[email protected] ~]$ cd apache-hive-2.1.1-bin/lib/
[[email protected] lib]$ cp mysql-connector-java-5.1.28.jar /home/zkpk/spark-2.1.1-bin-hadoop2.7/jars/

4.6.5重启Hadoop集群并验证spark-sql；下图，进入spark shell客户端，说明spark sql配置成功

[[email protected] lib]$ cd
[[email protected] ~]$ stop-all.sh
[[email protected] ~]$ start-all.sh
[[email protected] ~]$ cd ~/spark-2.1.1-bin-hadoop2.7
[[email protected] spark-2.1.1-bin-hadoop2.7]$　./bin/spark-sql --master yarn

4.6.6按ctrl+d组合键，退出spark shell

4.6.7若hadoop集群不再使用，请关闭集群

[[email protected] spark-2.1.1-bin-hadoop2.7]$ cd
[[email protected] ~]$ stop-all.sh