Spark Sandbox

Install sbt

• Download sbt-launch.jar, and put it into $HOME/bin.
• Create $HOME/bin/sbt, and change mode to 755. The content is:

SBT_OPTS="-Xms512M -Xmx1536M -Xss1M -XX:+CMSClassUnloadingEnabled -XX:MaxPermSize=256M -Dsbt.override.build.repos=true"
java $SBT_OPTS -jar `dirname $0`/sbt-launch.jar "$@"

• Create $HOME/.sbt/repositories, content is:

[repositories]
  local
  my-ivy-proxy-releases: http://10.20.8.31:8081/nexus/content/groups/ivy-releases/, [organization]/[module]/(scala_[scalaVersion]/)(sbt_[sbtVersion]/)[revision]/[type]s/[artifact](-[classifier]).[ext]
  my-maven-proxy-releases: http://10.20.8.31:8081/nexus/content/groups/public/

Install Spark

• Download Spark, choose the version corresponding to your HDFS.
• Extract the tar ball, say /path/to/spark
• Setup $SPARK_HOME=/path/to/spark
• Add $SPARK_HOME/bin to $PATH

Import Project

$ git clone [email protected]:jizhang/spark-sandbox
$ cd spark-sandbox
$ sbt eclipse

And import the project into Eclipse, provided ScalaIDE for Eclipse is installed.

Wordcount

• Run locally:

$ cd spark-sandbox
$ sbt "run-main Wordcount data/wordcount.txt"

• Submit to cluster:

$ sbt package
$ spark-submit --class Wordcount --master local target/scala-2.10/spark-sandbox_2.10-0.1.0.jar data/wordcount.txt

Logistic Regression

$ spark-submit --class LogisticRegression --master local target/scala-2.10/spark-sandbox_2.10-0.1.0.jar data/lr_data.txt 10 10

In-Memory Log Mining

$ spark-submit --class LogMining --master local target/scala-2.10/spark-sandbox_2.10-0.1.0.jar data/logs.txt

Streaming Wordcount

$ nc -lk 9999
$ spark-submit --class StreamingWordcount --master local[2] target/scala-2.10/spark-sandbox_2.10-0.1.0.jar

KMeans

$ spark-submit --class KMeans --master local target/scala-2.10/spark-sandbox_2.10-0.1.0.jar data/kmeans_data.txt 2 0.01

Recommendation

$ sbt "run-main recommendation.MainClass als"