spark搭建
下载spark-2.3.2
wget https://archive.apache.org/dist/spark/spark-2.3.2/spark-2.3.2-bin-hadoop2.7.tgz
需下载-hadoop-2.7版本的spark, 不然要自己加很多依赖进spark目录
修改配置
复制\$HADOOP_HOME/etc/hadoop/core-site.xml 至 \$SPARK_HOME/conf
复制\$HADOOP_HOME/etc/hadoop/hdfs-site.xml 至 \$SPARK_HOME/conf
复制\$HIVE_HOME/conf/hive-site.xml 至 \$SPARK_HOME/conf
修改$SPARK_HOME/conf/hive-site.xml, 将sparksql监听的端口改为10002, 免得与原hive的hiveserver2端口冲突
<property>
<name>hive.server2.thrift.port</name>
<value>10002</value>
<description>Port number of HiveServer2 Thrift interface when hive.server2.transport.mode is 'binary'.</description>
</property>
新建启动脚本并执行
在$SPARK_HOME目录创建文件startThriftServer.sh
vim startThriftServer.sh
添加以下内容
#!/bin/bash
./sbin/start-thriftserver.sh \
--master yarn
执行脚本
chmod +x ./startThriftServer.sh
./startThriftServer.sh
启动测试
执行beeline连接, 在$SPARK_HOME目录
[jevoncode@s1 spark-2.3.2-bin-hadoop2.7]$ ./bin/beeline
Beeline version 1.2.1.spark2 by Apache Hive
beeline> !connect jdbc:hive2://localhost:10002/hive_data
Connecting to jdbc:hive2://localhost:10002/hive_data
Enter username for jdbc:hive2://localhost:10002/hive_data: jevoncode
Enter password for jdbc:hive2://localhost:10002/hive_data: ***************
2018-10-14 11:15:24 INFO Utils:310 - Supplied authorities: localhost:10002
2018-10-14 11:15:24 INFO Utils:397 - Resolved authority: localhost:10002
2018-10-14 11:15:24 INFO HiveConnection:203 - Will try to open client transport with JDBC Uri: jdbc:hive2://localhost:10002/hive_data
Connected to: Spark SQL (version 2.3.2)
Driver: Hive JDBC (version 1.2.1.spark2)
Transaction isolation: TRANSACTION_REPEATABLE_READ
0: jdbc:hive2://localhost:10002/hive_data>
就可以执行sql语句了
spark动态资源配置
搭建完spark之后, 发现执行sql很慢, 从其webUI来看, 只有两个Executors执行, 而yarn集群有7台服务器, 从zabbix可以看到资源资源利用率低.
webUI在yarn界面, 点击Thrift JDBC/ODBC Server的ApplicationMaster即可进入
此问题的解决方法, 启动spark动态资源功能即可, 配置如下:
1.配置\$SPARK_HOME/conf/spark-defaults.conf
spark.dynamicAllocation.enabled true
spark.shuffle.service.enabled true
2.配置\$HADOOP_HOME/etc/hadoop/yarn-site.xml, 每个NodeManager都要配置
<property>\
<name>yarn.nodemanager.aux-services</name>\
<value>mapreduce_shuffle,spark_shuffle</value>
</property>
<property>
<name>yarn.nodemanager.aux-services.spark_shuffle.class</name>
<value>org.apache.spark.network.yarn.YarnShuffleService</value>
</property>
3.复制\$SPARK_HOME/yarn/spark-2.3.2-yarn-shuffle.jar至\$HADOOP_HOME/share/hadoop/yarn/
4.重启每个NodeManager
5.此时执行sql就可以看到有很多Executors在执行
TroubleShoot
1.shuffle的配置问题
2018-10-14 10:24:05 WARN YarnScheduler:66 - Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources
2018-10-14 10:24:20 WARN YarnScheduler:66 - Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources
2018-10-14 10:24:35 WARN YarnScheduler:66 - Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources
webUI什么错误信息也没, 状态也看不到
最后在yarn的application的日志里找到这个错误
2018-10-14 10:20:38 ERROR YarnAllocator:91 - Failed to launch executor 23 on container container_e69_1538148198468_17372_01_000024
org.apache.spark.SparkException: Exception while starting container container_e69_1538148198468_17372_01_000024 on host jevoncode.com
at org.apache.spark.deploy.yarn.ExecutorRunnable.startContainer(ExecutorRunnable.scala:125)
at org.apache.spark.deploy.yarn.ExecutorRunnable.run(ExecutorRunnable.scala:65)
at org.apache.spark.deploy.yarn.YarnAllocator$$anonfun$runAllocatedContainers$1$$anon$1.run(YarnAllocator.scala:534)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: org.apache.hadoop.yarn.exceptions.InvalidAuxServiceException: The auxService:spark_shuffle does not exist
at sun.reflect.GeneratedConstructorAccessor35.newInstance(Unknown Source)
at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
at java.lang.reflect.Constructor.newInstance(Constructor.java:423)
at org.apache.hadoop.yarn.api.records.impl.pb.SerializedExceptionPBImpl.instantiateException(SerializedExceptionPBImpl.java:168)
at org.apache.hadoop.yarn.api.records.impl.pb.SerializedExceptionPBImpl.deSerialize(SerializedExceptionPBImpl.java:106)
at org.apache.hadoop.yarn.client.api.impl.NMClientImpl.startContainer(NMClientImpl.java:205)
at org.apache.spark.deploy.yarn.ExecutorRunnable.startContainer(ExecutorRunnable.scala:122)
... 5 more
shuffle的配置问题, 上述错误是因为yarn-site.xml没有配置spark_shuffle和指定spark_shuffle.class
2.HADOOP_CONF_DIR配置问题
需在~/.bashrc增加配置
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
**粗体** _斜体_ [链接](http://example.com) `代码` - 列表 > 引用
。你还可以使用@
来通知其他用户。