1、下载spark压缩包,在虚拟机上解压安装
2、配置
在conf中复制spark-env.sh.template并重命名为spark-env.sh
mv spark-env.sh.template spark-env.sh
文件中添加如下配置:
export JAVA_HOME=/root/apps/jdk1.8.0_201
#export SPARK_MASTER_HOST=hdp-1
#export SPARK_MASTER_PORT=7077
export SPARK_DAEMON_JAVA_OPTS="-Dspark.deploy.recoveryMode=ZOOKEEPER -Dspark.deploy.zookeeper.url=hdp-1,hdp-2,hdp-3,hdp-4 -Dspark.deploy.zookeeper.dir=/spark"
export SPARK_WORKER_CORES=8
export SPARK_WORKER_MEMORY=6g
export SPARK_LOCAL_IP="192.168.150.151"
配置ha时需把上面两行注掉,不然其他机器启动master会失败
修改slaves在文件最后添加其他主机名:
hdp-2
hdp-3
hdp-4
启动spark(启动之前必须开启hdfs、yarn、zookeeper):
start-all.sh
./zkmanager.sh start