14（spark环境搭建）

1，安装java8

cd /software    上传文件

tar -zxvf ./jdk-8u181-linux-x64.tar.gz

vi /etc/profile     四台机器

export JAVA_HOME=/software/jdk1.8.0_181
export PATH=$JAVA_HOME/bin:$PATH

. /etc/profile

java -version

scp -r jdk1.8.0_181/ node0002:`pwd`

scp -r jdk1.8.0_181/ node0003:`pwd`

scp -r jdk1.8.0_181/ node0004:`pwd`

. /etc/profile

java -version   查看四台机器是否都是1.8

2，修改jdk指向

cd /usr/java

ll     (指向1.7)

ln -sf /software/jdk1.8.0_181/bin/java  /usr/bin/java      （四台机器）

cd /usr/bin

ll    (java指向1.8)

3，修改hadoop的jdk配置

cd /opt/sxt/hadoop-2.6.5/etc/hadoop

vi hadoop-env.sh     (四台机器)

export JAVA_HOME=/software/jdk1.8.0_181/bin/java

4，安装spark

cd /software

tar -zxvf spark-2.3.1-bin-hadoop2.6.tgz

mv spark-2.3.1-bin-hadoop2.6 spark-2.3.1

cd spark-2.3.1/conf/

cp slaves.template slaves

vi slaves          （配置worker节点）

node0002

node0003

cp spark-env.sh.template spark-env.sh

vi spark-env.sh      (配置master节点，添加)

export SPARK_MASTER_HOST=node0001
export SPARK_MASTER_PORT=7077
export SPARK_WORKER_CORES=2
export SPARK_WORKER_MEMORY=3g

5，分发到node0002,node0003

cd /software

scp -r spark-2.3.1 node0002:`pwd`

scp -r spark-2.3.1 node0003:`pwd`

6,启动spark

cd spark-2.3.1/sbin/

./start-all.sh     (一定要加./，否则会启动hdfs集群)

页面访问：node0001:8080

7，基于yarn任务提交

node0002:

scp -r spark-2.3.1/ node0004:`pwd`

node0004:

cd /software/spark-2.3.1/conf/

 rm -rf ./slaves

vi spark-env.sh      (注释掉export的几个)

8，配置yarn

vi /opt/sxt/hadoop-2.6.5/etc/hadoop/yarn-site.xml    （添加）

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

cd /opt/sxt/hadoop-2.6.5/etc/hadoop

scp ./yarn-site.xml node0002:`pwd`

scp ./yarn-site.xml node0002:`pwd`

scp ./yarn-site.xml node0002:`pwd`

vi /opt/sxt/hadoop-2.6.5/etc/hadoop/hadoop-env.sh

改为：export JAVA_HOME=/software/jdk1.8.0_181

scp ./hadoop-env.sh node0002:`pwd`

scp ./hadoop-env.sh node0002:`pwd`

scp ./hadoop-env.sh node0002:`pwd`