spark集群安装

一、Scala

1、scala安装：http://www.scala-lang.org/download/2.10.6.html

2、解压： tar -zxvf /home/tools/scala-2.10.6.tgz -C /home/softwares/

3、配置环境变量：vi /etc/profile

export SCALA_HOME=/home/softwares/scala-2.10.6

export PATH=$PATH:$JAVA_HOME/bin:$SCALA_HOME/bin

Source /etc/profile

输入scala 测试是否安装成功

二、SPARK

1、下载spark-1.6.2

http://mirror.bit.edu.cn/apache/spark/spark-1.6.0/

解压:tar -zxvf /home/tools/spark-1.6.2-bin-hadoop2.6.tgz -C /home/softwares/

2、环境变量：

export SCALA_HOME=/home/softwares/scala-2.10.6

export SPARK_HOME=/home/softwares/spark-1.6.0-bin-hadoop2.6

export PATH=$PATH:$JAVA_HOME/bin:$SCALA_HOME/bin:${SPARK_HOME}/bin:${SPARK_HOME}/sbin:$PATH

3、conf/spark-env.sh:

export JAVA_HOME=/home/softwares/jdk1.7.0_79

export SCALA_HOME=/home/softwares/scala-2.10.6

export HADOOP_HOME=/home/softwares/hadoop-2.6.0

export HADOOP_CONF_DIR=${HADOOP_HOME}/etc/hadoop

export SPARK_MASTER_IP=59.67.152.31

export SPARK_WORKER_MEMORY=2g

export master=spark://59.67.152.31

export SPARK_EXECUTOR_MEMORY=2g

export SPARK_DRIVER_MEMORY=2g

export SPARK_WORKDER_CORES=2

4、conf/slaves(添加工作节点):

Slave1.Hadoop

Slave2.Hadoop

5、conf/spark-defaults.conf：

spark.master                    spark://59.67.152.31:7077

spark.executor.extraJavaOptions   -XX:+PrintGCDetails -Dkey=value -Dnumbers="one two three"

spark.eventLog.enabled           true

spark.eventLog.dir               hdfs://59.67.152.31:8020/historyserverforSpark

spark.yarn.historyServer.address   59.67.152.31:10020

spark.history.fs.logDirectory       hdfs://59.67.152.31:8020/historyserverforSpark

6、配置好之后将复制到其他机器中，进行相同操作（环境变量配置）

scp -r /home/softwares/spark-1.6.2-bin-hadoop2.6/ root@59.67.152.33:/home/softwares/

7、启动：

在sbin/start-all.sh或者bin/start-shell

 

 

三、出现问题：在安装过程中遇到问题如下：

[root@Master spark-1.6.0-bin-hadoop2.6]# sbin/start-all.sh  starting org.apache.spark.deploy.master.Master, logging to /home/softwares/spark-1.6.0-bin-hadoop2.6/logs/spark-root-org.apache.spark.deploy.master.Master-1-Master.Hadoop.out failed to launch org.apache.spark.deploy.master.Master: /bin/java: 没有那个文件或目录-bin-hadoop2.6/bin/spark-class: line 86: /home/softwares/jdk1.7.0_79 full log in /home/softwares/spark-1.6.0-bin-hadoop2.6/logs/spark-root-org.apache.spark.deploy.master.Master-1-Master.Hadoop.out Slave1.Hadoop: starting org.apache.spark.deploy.worker.Worker, logging to /home/softwares/spark-1.6.0-bin-hadoop2.6/logs/spark-root-org.apache.spark.deploy.worker.Worker-1-Slave1.Hadoop.out Slave2.Hadoop: starting org.apache.spark.deploy.worker.Worker, logging to /home/softwares/spark-1.6.0-bin-hadoop2.6/logs/spark-root-org.apache.spark.deploy.worker.Worker-1-Slave2.Hadoop.out Slave2.Hadoop: failed to launch org.apache.spark.deploy.worker.Worker: Slave1.Hadoop: failed to launch org.apache.spark.deploy.worker.Worker: /bin/java: 没有那个文件或目录res/spark-1.6.0-bin-hadoop2.6/bin/spark-class: line 86: /home/softwares/jdk1.7.0_79 /bin/java: 没有那个文件或目录res/spark-1.6.0-bin-hadoop2.6/bin/spark-class: line 86: /home/softwares/jdk1.7.0_79 Slave2.Hadoop: full log in /home/softwares/spark-1.6.0-bin-hadoop2.6/logs/spark-root-org.apache.spark.deploy.worker.Worker-1-Slave2.Hadoop.out Slave1.Hadoop: full log in /home/softwares/spark-1.6.0-bin-hadoop2.6/logs/spark-root-org.apache.spark.deploy.worker.Worker-1-Slave1.Hadoop.out [root@Master spark-1.6.0-bin-hadoop2.6]

 

    仔细检查配置文件conf/spark-env.sh 以及环境变量都正确，就是不知道什么原因，找了两天的错误，后来通过vi spark-env.sh进入配置文件，结果发现配置文件不正确，多了一些不知所以然的的字符，把这些字符删掉，重新启动正确。

   分析原因：可能实在网上找了了一些安装步骤，直接复制粘贴过去有一些格式出现问题，所以一定要细心，不要偷懒，一步一步仔细做，切勿急功近利。