spark搭建部署

基础环境准备

• 安装JDK1.8+,并设置环境变量
• 搭建zookeeper集群
• 搭建Hadoop集群

Spark local模式

上传编译完成的spark安装程序到服务器上，并解压到指定目录

[root@hadoop01 soft]# tar zxvf spark-2.2.0-bin-2.6.0-cdh5.14.0.tgz -C /usr/local/
[root@hadoop01 soft]# cd /usr/local/
[root@hadoop01 local]# mv spark-2.2.0-bin-2.6.0-cdh5.14.0/ spark
[root@hadoop01 local]# cd spark/conf/

修啊改配置文件spark-env.sh.template为spark-env.sh

[root@hadoop01 conf]# mv spark-env.sh.template spark-env.sh

启动验证spark程序

[root@hadoop01 conf]# cd ../
[root@hadoop01 spark]# ./bin/spark-shell
[root@hadoop01 conf]# cd ../
[root@hadoop01 spark]# ./bin/spark-shell

退出spark命令行

scala> :quit

执行jar计算圆周率

bin/spark-submit 
--class org.apache.spark.examples.SparkPi 
--master local[2] 
--executor-memory 1G 
--total-executor-cores 2 
/usr/local/spark/examples/jars/spark-examples_2.11-2.2.0.jar 
100

spark的standAlone模式

修改配置文件spark-env.sh,添加下列内容

[root@hadoop01 conf]# vim spark-env.sh

export JAVA_HOME=/usr/local/java/jdk1.8.0_201
export SPARK_MASTER_HOST=node01
export SPARK_MASTER_PORT=7077
export SPARK_HISTORY_OPTS="-Dspark.history.ui.port=4000 -Dspark.history.retainedApplications=3 -Dspark.history.fs.logDirectory=hdfs://node01:9000/spark_log"

修改slaves文件，添加下列内容

[root@hadoop01 conf]# mv slaves.template slaves
[root@hadoop01 conf]# vim slaves

node01
node02
node03

修改配置文件spark-default.conf,添加下列内容

[root@hadoop01 conf]# mv spark-defaults.conf.template spark-defaults.conf
[root@hadoop01 conf]# vim spark-defaults.conf

spark.eventLog.enabled true
spark.eventLog.dir hdfs://node01:9000/spark_log
spark.eventLog.compress true

并在Hadoop的hdfs中创建spark日志目录spark_log

[root@hadoop01 conf]# hdfs dfs -ls /

分发配置好的spark程序到其它两台服务器上

[root@hadoop01 conf]# cd ../../
[root@hadoop01 conf]# scp -r spark root@node02:$PWD
[root@hadoop01 conf]# scp -r spark root@node03:$PWD

启动spark程序

[root@hadoop01 conf]# cd spark
[root@hadoop01 conf]# sbin/start-all.sh
[root@hadoop01 conf]# sbin/start-history-server.sh

spark的standAlone模式验证

[root@hadoop01 spark]# bin/spark-submit 
--class org.apache.spark.examples.SparkPi 
--master spark://node01:7077 
--executor-memory 1G 
--total-executor-cores 1 
/usr/local/spark/examples/jars/spark-examples_2.11-2.2.0.jar 
100

spark的HA高可用模式

停止启动的spark程序

[root@hadoop01 spark]# sbin/stop-all.sh
[root@hadoop01 spark]# sbin/stop-history-server.sh

解压spark程序到指定目录，并重命名为spark-HA

[root@hadoop01 soft]# tar zxvf spark-2.2.0-bin-2.6.0-cdh5.14.0.tgz -C /usr/local/
[root@hadoop01 soft]# cd /usr/local/
[root@hadoop01 local]# mv spark-2.2.0-bin-2.6.0-cdh5.14.0/ spark-HA

修改spark配置文件spark-env.sh

[root@hadoop01 local]# cd spark-HA/conf/
[root@hadoop01 conf]# mv spark-env.sh.template spark-env.sh
[root@hadoop01 conf]# vim spark-env.sh

export JAVA_HOME=/usr/local/java/jdk1.8.0_201
export SPARK_MASTER_PORT=7077
export SPARK_HISTORY_OPTS="-Dspark.history.ui.port=4000 -Dspark.history.retainedApplications=3 -Dspark.history.fs.logDirectory=hdfs://node01:9000/spark_log"
export SPARK_DAEMON_JAVA_OPTS="-Dspark.deploy.recoveryMode=ZOOKEEPER -Dspark.deploy.zookeeper.url=node01:2181,node02:2181,node03:2181 -Dspark.deploy.zookeeper.dir=/spark"

修改配置文件slave

[root@hadoop01 conf]# mv slaves.template slaves
[root@hadoop01 conf]# vim slaves

node01
node02
node03

修改配置文件spark-defaults.conf

[root@hadoop01 conf]# mv spark-defaults.conf.template spark-defaults.conf
[root@hadoop01 conf]# vim spark-defaults.conf

spark.eventLog.enabled true
spark.eventLog.dir hdfs://node01:9000/spark_log
spark.eventLog.compress true

在hadoop的hdfs上常见spark日志目录

[root@hadoop01 conf]# hdfs dfs -mkdir -p /spark_log

分发spark-HA程序到其他两台服务器上

启动spark高可用集群

[root@hadoop01 conf]# cd /usr/local/spark-HA/
[root@hadoop01 spark-HA]# sbin/start-all.sh
[root@hadoop01 spark-HA]# sbin/start-history-server.sh

node02服务器启动master节点

[root@hadoop01 spark-HA]# cd /usr/local/spark-HA/
[root@hadoop01 spark-HA]# sbin/start-master.sh

验证spark高可用集群

spark的HA模式下的spark的命令行

[root@hadoop01 spark-HA]# bin/spark-shell --master spark://node01:7077,node02:7077

运行jar包进行验证测试

[root@hadoop01 spark-HA]# bin/spark-submit 
--class org.apache.spark.examples.SparkPi 
--master spark://node01:7077,node02:7077 
--executor-memory 1G 
--total-executor-cores 1 
/usr/local/spark-HA/examples/jars/spark-examples_2.11-2.2.0.jar 
100

spark的on yarn模式

小提示：如果yarn集群资源不够，我们可以在yarn-site.xml当中添加以下两个配置，然后重启yarn集群，跳过yarn集群资源的检查

[root@hadoop01 conf]# scp -r spark-HA root@node02:$PWD
[root@hadoop01 conf]# scp -r spark-HA root@node03:$PWD

<property>

<name> yarn.nodemanager.pmem-check-enabled</name

<value>false</value>

</property>

<property>

<name> yarn.nodemanager.vmem-check-enabled</name

<value>false</value>

</property>

修改配置文件spark-env.sh

HADOOP_CONF_DIR=/usr/local/hadoop-HA/etc/hadoop
YARN_CONF_DIR=/usr/local/hadoop-HA/etc/hadoop

提交任务到yarn集群上进行验证

[root@hadoop01 spark-HA]# bin/spark-submit 
--class org.apache.spark.examples.SparkPi 
--master yarn 
--deploy-mode client 
/usr/local/spark-HA/examples/jars/spark-examples_2.11-2.2.0.jar 
100