前置条件说明

Hive on Spark是Hive跑在Spark上，用的是Spark执行引擎，而不是MapReduce，和Hive on Tez的道理一样。
从Hive 1.1版本开始，Hive on Spark已经成为Hive代码的一部分了，并且在spark分支上面，可以看这里https://github.com/apache/hive/tree/spark，并会定期的移到master分支上面去。
关于Hive on Spark的讨论和进度，可以看这里https://issues.apache.org/jira/browse/HIVE-7292。
hive on spark文档:https://issues.apache.org/jira/secure/attachment/12652517/Hive-on-Spark.pdf

源码下载

git clone https://github.com/apache/hive.git hive_on_spark

编译

 cd hive_on_spark/
 git branch -r
  origin/HEAD -> origin/master
  origin/HIVE-4115
  origin/HIVE-8065
  origin/beeline-cli
  origin/branch-0.10
  origin/branch-0.11
  origin/branch-0.12
  origin/branch-0.13
  origin/branch-0.14
  origin/branch-0.2
  origin/branch-0.3
  origin/branch-0.4
  origin/branch-0.5
  origin/branch-0.6
  origin/branch-0.7
  origin/branch-0.8
  origin/branch-0.8-r2
  origin/branch-0.9
  origin/branch-1
  origin/branch-1.0
  origin/branch-1.0.1
  origin/branch-1.1
  origin/branch-1.1.1
  origin/branch-1.2
  origin/cbo
  origin/hbase-metastore
  origin/llap
  origin/master
  origin/maven
  origin/next
  origin/parquet
  origin/ptf-windowing
  origin/release-1.1
  origin/spark
  origin/spark-new
  origin/spark2
  origin/tez
  origin/vectorization

 git checkout origin/spark
 git branch* （分离自 origin/spark）
  master123456789101112131415161718192021222324252627282930313233343536373839404142434445

修改$HIVE_ON_SPARK/pom.xml
spark版本改成spark1.4.1

 <spark.version>1.4.1</spark.version>1

hadoop版本改成2.3.0-cdh5.1.0

<hadoop-23.version>2.3.0-cdh5.1.0</hadoop-23.version>1

编译命令

export MAVEN_OPTS="-Xmx2g -XX:MaxPermSize=512M -XX:ReservedCodeCacheSize=512m"mvn clean package -Phadoop-2 -DskipTests12

添加Spark的依赖到Hive的方法

spark home:/home/cluster/apps/spark/spark-1.4.1
hive home:/home/cluster/apps/hive_on_spark

1.set the property ‘spark.home’ to point to the Spark installation:

hive> set spark.home=/home/cluster/apps/spark/spark-1.4.1;  1

1. Define the SPARK_HOME environment variable before starting Hive CLI/HiveServer2:

export SPARK_HOME=/home/cluster/apps/spark/spark-1.4.11

3.Set the spark-assembly jar on the Hive auxpath:

hive --auxpath /home/cluster/apps/spark/spark-1.4.1/lib/spark-assembly-*.jar1

1. Add the spark-assembly jar for the current user session:

hive> add jar /home/cluster/apps/spark/spark-1.4.1/lib/spark-assembly-*.jar;1

1. Link the spark-assembly jar to $HIVE_HOME/lib.

启动Hive过程中可能出现的错误：

[ERROR] Terminal initialization failed; falling back to unsupportedjava.lang.IncompatibleClassChangeError: Found class jline.Terminal, but interface was expected
        at jline.TerminalFactory.create(TerminalFactory.java:101)
        at jline.TerminalFactory.get(TerminalFactory.java:158)
        at jline.console.ConsoleReader.<init>(ConsoleReader.java:229)
        at jline.console.ConsoleReader.<init>(ConsoleReader.java:221)
        at jline.console.ConsoleReader.<init>(ConsoleReader.java:209)
        at org.apache.hadoop.hive.cli.CliDriver.getConsoleReader(CliDriver.java:773)
        at org.apache.hadoop.hive.cli.CliDriver.executeDriver(CliDriver.java:715)
        at org.apache.hadoop.hive.cli.CliDriver.run(CliDriver.java:675)
        at org.apache.hadoop.hive.cli.CliDriver.main(CliDriver.java:615)
        at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
        at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
        at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
        at java.lang.reflect.Method.invoke(Method.java:606)
        at org.apache.hadoop.util.RunJar.main(RunJar.java:212)

Exception in thread "main" java.lang.IncompatibleClassChangeError: Found class jline.Terminal, but interface was expected123456789101112131415161718

解决方法：export HADOOP_USER_CLASSPATH_FIRST=true

其他场景的错误解决方法参见：https://cwiki.apache.org/confluence/display/Hive/Hive+on+Spark%3A+Getting+Started

需要设置spark.eventLog.dir参数，比如：

set spark.eventLog.dir= hdfs://master:8020/directory
否则查询会报错，否则一直报错:/tmp/spark-event类似的文件夹不存在

启动hive后设置执行引擎为spark：

hive> set hive.execution.engine=spark;1

设置spark的运行模式：

hive> set spark.master=spark://master:70771

或者yarn：spark.master=yarn

Configure Spark-application configs for Hive

可以配置在spark-defaults.conf或者hive-site.xml

spark.master=<Spark Master URL>
spark.eventLog.enabled=true;            
spark.executor.memory=512m;             
spark.serializer=org.apache.spark.serializer.KryoSerializer;
spark.executor.memory=...  #Amount of memory to use per executor process.spark.executor.cores=...  #Number of cores per executor.spark.yarn.executor.memoryOverhead=...spark.executor.instances=...  #The number of executors assigned to each application.spark.driver.memory=...  #The amount of memory assigned to the Remote Spark Context (RSC). We recommend 4GB.spark.yarn.driver.memoryOverhead=...  #We recommend 400 (MB).12345678910

参数配置详见文档：https://cwiki.apache.org/confluence/display/Hive/Hive+on+Spark%3A+Getting+Started

执行sql语句后可以在监控页面查看job/stages等信息

hive (default)> select city_id, count(*) c from city_info group by city_id order by c desc limit 5;
Query ID = spark_20150309173838_444cb5b1-b72e-4fc3-87db-4162e364cb1e
Total jobs = 1Launching Job 1 out of 1In order to change the average load for a reducer (in bytes):  set hive.exec.reducers.bytes.per.reducer=<number>In order to limit the maximum number of reducers:  set hive.exec.reducers.max=<number>In order to set a constant number of reducers:  set mapreduce.job.reduces=<number>
state = SENT
state = STARTED
state = STARTED
state = STARTED
state = STARTED
Query Hive on Spark job[0] stages:1Status: Running (Hive on Spark job[0])
Job Progress Format
CurrentTime StageId_StageAttemptId: SucceededTasksCount(+RunningTasksCount-FailedTasksCount)/TotalTasksCount [StageCost]2015-03-09 17:38:11,822 Stage-0_0: 0(+1)/1      Stage-1_0: 0/1  Stage-2_0: 0/1state = STARTED
state = STARTED
state = STARTED2015-03-09 17:38:14,845 Stage-0_0: 0(+1)/1      Stage-1_0: 0/1  Stage-2_0: 0/1state = STARTED
state = STARTED2015-03-09 17:38:16,861 Stage-0_0: 1/1 Finished Stage-1_0: 0(+1)/1      Stage-2_0: 0/1state = SUCCEEDED2015-03-09 17:38:17,867 Stage-0_0: 1/1 Finished Stage-1_0: 1/1 Finished Stage-2_0: 1/1 Finished
Status: Finished successfully in 10.07 seconds
OK
city_id c
-1000   22826-10     17294-20     10608-1      6186
    4158Time taken: 18.417 seconds, Fetched: 5 row(s)