I've just started developing my big data applications with Spark and the thing that I wanted most was a development environment for Eclipse (I'm not an IntelliJ user) that allowed me to build and run Spark/Scala applications locally. Fortunately all of the components have been created to make this happen and being a long time Java/Maven/Eclipse user getting them working together was relatively straightforward. The approach is already described in a number of articles (Google for Spark and Eclipse) but I wanted to make it work with the Jar versions that ship with BigInsights v4.0.0.1. Here's what I did.
Dependencies
The major dependencies I'm working with are as follows:
Windows 7
IBM JDK 7 (SR3 build 2.7)
Scala 2.10.4
Spark 1.2.1
Eclipse Juno SR 2
Maven 3.0.5
Summary Of The Steps Involved
5 - Import the project into Eclipse as a maven project
I describe steps 3 and 5 in more detail below.
Creating A Maven Project For Spark Development (Step 3 Detail)
In some directory create a new project directory with the
following structure:
SomeDirectory/
MySparkProject/
src/
main/
scala/
pom.xml
There are probably Maven archetypes for creating these directories but I didn't check as for my purposes I only needed this very simple structure. The pom.xml file is obviously the interesting thing here. I started with the following contents:
<?xml version="1.0"
encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/maven-v4_0_0.xsd">
<modelVersion>4.0.0</modelVersion>
<groupId>com.ibm.ets</groupId>
<artifactId>MyProject</artifactId>
<version>0.1-SNAPSHOT</version>
<name>MyProject</name>
<repositories>
<repository>
<id>scala-tools.org</id>
<name>Scala-Tools Maven2 Repository</name>
<url>http://scala-tools.org/repo-releases</url>
</repository>
</repositories>
<pluginRepositories>
<pluginRepository>
<id>scala-tools.org</id>
<name>Scala-Tools Maven2
Repository</name>
<url>http://scala-tools.org/repo-releases</url>
</pluginRepository>
</pluginRepositories>
<dependencies>
<dependency>
<groupId>org.scala-lang</groupId>
<artifactId>scala-library</artifactId>
<version>2.10.4</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-core_2.10</artifactId>
<version>1.2.1</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-sql_2.10</artifactId>
<version>1.2.1</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-hive_2.10</artifactId>
<version>1.2.1</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-streaming_2.10</artifactId>
<version>1.2.1</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-mllib_2.10</artifactId>
<version>1.2.1</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-graphx_2.10</artifactId>
<version>1.2.1</version>
</dependency>
<dependency>
<groupId>org.apache.hadoop</groupId>
<artifactId>hadoop-client</artifactId>
<version>2.6.0</version>
</dependency>
</dependencies>
<build>
<pluginManagement>
<plugins>
<plugin>
<groupId>org.scala-tools</groupId>
<artifactId>maven-scala-plugin</artifactId>
<version>2.15.2</version>
</plugin>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-compiler-plugin</artifactId>
<version>3.1</version>
<configuration>
<source>1.7</source>
<target>1.7</target>
</configuration>
</plugin>
</plugins>
</pluginManagement>
<plugins>
<plugin>
<groupId>org.scala-tools</groupId>
<artifactId>maven-scala-plugin</artifactId>
<executions>
<execution>
<id>scala-compile-first</id>
<phase>process-resources</phase>
<goals>
<goal>add-source</goal>
<goal>compile</goal>
</goals>
</execution>
</executions>
</plugin>
</plugins>
</build>
</project>
I've selected Scala, Spark and Hadoop versions that match with those shipped in BigInsights 4.
You can now build Scala/Spark (or Java/Spark) programs here, using you favorite text editor, e.g. notepad or vi, and compile them using Maven to produce a jar for use with
spark-submit. However I wanted to do my development in Eclipse.
Installing winutils So This All Works on Windows (Step 4a Detail)
This is of course for windows users only.
I initially got the following error when running a Spark program from Eclipse
Could not locate executable null\bin\winutils.exe
I copied the winutils.exe to a directory and then set HADOOP_HOME variable as follows
HADOOP_HOME=C:\simon\svn\bigdata2\Spark\winutil
Then remember to restart Eclipse!
Importing The Project Into Eclipse (Step 5 Detail)
You will have installed the "maven 2 eclipse" plugin at step 4 (see the summary of steps above) so you can now import your Spark maven project using the following option:
To run Spark programs from Eclipse locally you need to configure Spark to run locally (rather than on a cluster). As a convenience I pass a parameter into my code to indicate whether I'm running in Eclipse. I modified the sample Spark code as follows to process the parameter.
import
org.apache.spark.SparkContext;
import
org.apache.spark.SparkContext._;
import
org.apache.spark.SparkConf;
import
org.apache.spark.rdd.RDD;
object SparkTutorial {
def main(args:
Array[String]) {
var conf:SparkConf = null;
var taxi:RDD[String] = null;
if (args.length > 0 && args(0) == "local") {
conf = new
SparkConf().setAppName("Spark Tutorial").setMaster("local[*]");
} else {
conf = new
SparkConf().setAppName("Spark Tutorial").setMaster("yarn-cluster");
}
val sc = new SparkContext(conf);
if (args.length > 0 && args(0) == "local") {
taxi = sc.textFile("nyctaxisub.csv", 2);
} else {
taxi = sc.textFile("/user/spark/simonstuff/sparkdata/nyctaxisub/*");
}
val taxiSplitColumns = taxi.map(line=>line.split(','));
val taxiMedCountsTuples = taxiSplitColumns.map(vals=>(vals(6),1));
val taxiMedCountsOneLine = taxiMedCountsTuples.reduceByKey((x,y)
=> x + y);
for (pair<-taxiMedCountsOneLine.map(_.swap).top(10)){
println("Taxi Medallion %s had %s Trips".format(pair._2,pair._1));
}
}
}
Note that I check twice whether a single parameter value of "local" is passed into the program. If so the command setMaster("local[*]") instructs Spark to run locally, i.e. doesn't require a separate cluster installation. It also reads the input data from a local directory path rather than from an HDFS directory path.
I pass the parameter into the program via an Eclipse run configuration in the following way:
Running On a Cluster
Having developed your Spark application in Eclipse you probably want to run it on a cluster. I export the project as a Java jar file from Eclipse. You can also generate the Jar file by going back to your Maven project and building it there, for example, by running mvn package.
Once you have the Jar you can copy it to the head node of your cluster and run it using Spark, for example,
Spark-submit –class “SparkTutorial” –master yarn-cluster
/SomeDirectory/MyProject/target/MyProject.jar
