Thrift和spark-sql客户端部署

更多精彩，请点击上方蓝字关注我们！

1 Spark Thrift部署

• 下载地址：

http://mirrors.hust.edu.cn/apache/spark/spark-2.1.0/spark-2.1.0-bin-hadoop2.6.tgz

• 解压

[root@cdh03 ~]# tar -zxvf spark-2.1.0-bin-hadoop2.6.tgz

• 拷贝至/opt/cloudera/parcels/SPARK2/lib/spark2/jars目录

• 在集群所有节点上执行

[root@cdh03 ~]# cd root/spark-2.1.0-bin-hadoop2.6/jars/
[root@cdh03 jars]# ll *hive*.jar
[root@cdh03 jars]# cp hive-cli-1.2.1.spark2.jar spark-hive-thriftserver_2.11-2.1.0.jar opt/cloudera/parcels/SPARK2/lib/spark2/jars/
[root@cdh03 jars]# ll opt/cloudera/parcels/SPARK2/lib/spark2/jars/*hive*.jar

• 把/opt/cloudera/parcels/SPARK2/lib/spark2/jars目录下的所有jar上传至HDFS

[root@cdh03 jars]# kinit spark/admin
Password for spark/admin@FAYSON.COM: 
[root@cdh03 jars]# cd opt/cloudera/parcels/SPARK2/lib/spark2/jars/
[root@cdh03 jars]# hadoop fs -mkdir -p user/spark/share/spark2-jars
[root@cdh03 jars]# hadoop fs -put *.jar user/spark/share/spark2-jars
[root@cdh03 jars]# hadoop fs -ls user/spark/share/spark2-jars

• 修改Spark的配置

spark.yarn.jars=hdfs://nameservice1/user/spark/share/spark2-jars/*.jar

• Spark Thrift启动和停止脚本

[root@cdh03 jars]# cd root/spark-2.2.0-bin-hadoop2.6/sbin/
[root@cdh03 sbin]# ll *thrift*.sh
[root@cdh03 sbin]# cp *thrift*.sh opt/cloudera/parcels/SPARK2/lib/spark2/sbin/
[root@cdh03 sbin]# ll opt/cloudera/parcels/SPARK2/lib/spark2/sbin/*thriftserver*

• 修改load-spark-env.sh脚本

[root@cdh03 sbin]# cd opt/cloudera/parcels/SPARK2/lib/spark2/bin
[root@cdh03 bin]# vim load-spark-env.sh 
# 该脚本是启动Spark相关服务加载依赖环境

• 内容注释

exec "$SPARK_HOME/bin/$SCRIPT" "$@"

• 部署Spark SQL客户端

[root@cdh03 ~]# cd spark-2.1.0-bin-hadoop2.6/bin/
[root@cdh03 bin]# pwd
[root@cdh03 bin]# cp root/spark-2.1.0-bin-hadoop2.6/bin/spark-sql opt/cloudera/parcels/SPARK2/lib/spark2/bin
[root@cdh03 bin]# ll opt/cloudera/parcels/SPARK2/lib/spark2/bin/spark-sql 

2 Spark Thrift的启动与停止

• 创建一个Kerberos账号，导出hive.keytab文件

[root@cdh01 ~]# kadmin.local 
kadmin.local:  addprinc -randkey hive/cdh03.fayson.com@FAYSON.COM
kadmin.local:  xst -norandkey -k hive-cdh03.keytab hive/cdh03.fayson.com@FAYSON.COM

• 查看

[root@cdh01 ~]# klist -e -kt hive-cdh03.keytab 

• 文件拷贝hive-cdh03.keytab至Spark2.1 ThriftServer服务所在服务器

[root@cdh01 ~]# scp hive-cdh03.keytab cdh03.fayson.com:/opt/cloudera/parcels/SPARK2/lib/spark2/sbin/

• 启动Thrift Server

cd opt/cloudera/parcels/SPARK2/lib/spark2/sbin
./start-thriftserver.sh --hiveconf hive.server2.authentication.kerberos.principal=hive/cdh03.fayson.com@FAYSON.COM \
  --hiveconf hive.server2.authentication.kerberos.keytab=hive-cdh03.keytab \
  --hiveconf hive.server2.thrift.port=10002 \
  --hiveconf hive.server2.thrift.bind.host=0.0.0.0 \
  --principal hive/cdh03.fayson.com@FAYSON.COM --keytab hive-cdh03.keytab

• 检查监听

[root@cdh03 sbin]# netstat -apn |grep 10002

• Spark ThriftServer服务停止

[root@cdh03 sbin]# ./stop-thriftserver.sh

3 验证

• beeline测试

[root@cdh01 ~]# kinit fayson
[root@cdh01 ~]# klist
[root@cdh01 ~]# beeline 
beeline> !connect jdbc:hive2://cdh03.fayson.com:10002/;principal=hive/cdh03.fayson.com@FAYSON.COM

• SQL测试

0: jdbc:hive2://cdh03.fayson.com:10001/> show tables;
0: jdbc:hive2://cdh03.fayson.com:10001/> select * from test;
0: jdbc:hive2://cdh03.fayson.com:10001/> select count(*) from test;
0: jdbc:hive2://cdh03.fayson.com:10001/> 

• 查看SQL操作是否是通过Spark执行

  
  

  
  

• spark-sql验证

[root@cdh03 ~]# kinit fayson
[root@cdh03  ~]# /opt/cloudera/parcels/CDH/lib/spark/bin/spark-sql

关注公众号：领取精彩视频课程&海量免费语音课程