Spark集群搭建简要

Spark集群搭建

1 Spark编译

1.1 下载源代码

git clone git://github.com/apache/spark.git -b branch-1.6

1.2 修改pom文件

增加cdh5.0.2相关profile,如下：
<profile>
  <id>cdh5.0.2</id>
  <properties>
	<hadoop.version>2.3.0-cdh5.0.2</hadoop.version>
	<hbase.version>0.96.1.1-cdh5.0.2</hbase.version>
	<flume.version>1.4.0-cdh5.0.2</flume.version>
	<zookeeper.version>3.4.5-cdh5.0.2</zookeeper.version>
  </properties>
</profile>

1.3 编译

build/mvn -Pyarn -Pcdh5.0.2 -Phive -Phive-thriftserver -Pnative -DskipTests package

上述命令，由于国外maven.twttr.com被墙，添加hosts，199.16.156.89 maven.twttr.com,再次执行。

2 Spark集群搭建[SPARK ON YARN]

2.1 修改配置文件

--spark-env.sh--
export SPARK_SSH_OPTS="-p9413"
export HADOOP_CONF_DIR=/opt/hadoop/hadoop-cluster/modules/hadoop-2.3.0-cdh5.0.2/etc/hadoop
export SPARK_EXECUTOR_INSTANCES=1
export SPARK_EXECUTOR_CORES=4
export SPARK_EXECUTOR_MEMORY=1G
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HADOOP_HOME/lib/native/
--slaves--
192.168.3.211 hadoop-dev-211
192.168.3.212 hadoop-dev-212
192.168.3.213 hadoop-dev-213
192.168.3.214 hadoop-dev-214

2.2 集群规划，启动集群

--集群规划--
hadoop-dev-211	Master、Woker
hadoop-dev-212  Woker
hadoop-dev-213	Woker
hadoop-dev-214	Woker
--启动Master--
sbin/start-master.sh
--启动Wokers--
sbin/start-slaves.sh

2.3 查看界面

3 集成hive

将hive-site.xml和hive-log4j.properties至spark中conf目录

4 Spark实例演示

4.1 读取mysql数据至hive

# 步骤1，启动spark-shell
bin/spark-shell --jars lib_managed/jars/hadoop-lzo-0.4.17.jar 
--driver-class-path /opt/hadoop/hadoop-cluster/modules/apache-hive-1.2.1-bin/lib/mysql-connector-java-5.6-bin.jar
# 步骤2，读取mysql数据
val jdbcDF = sqlContext.read.format("jdbc").options(Map("url" -> "jdbc:mysql://hadoop-dev-212:3306/hive","dbtable" -> "VERSION", "user" -> "hive", "password" -> "123456")).load();
# 步骤3，转成hive表
jdbcDF.saveAsTable("test");