Skip to content

Instantly share code, notes, and snippets.

Building Spark for PySpark use on top of YARN

Build Spark on local machine (only if using PySpark; otherwise, remote machine works) (http://spark.apache.org/docs/latest/building-with-maven.html)

export MAVEN_OPTS="-Xmx2g -XX:MaxPermSize=512M -XX:ReservedCodeCacheSize=512m"
mvn -Pyarn -Phadoop-2.4 -Dhadoop.version=2.4.0 -DskipTests clean package

Copy the assembly/target/scala-2.10/...jar to the corresponding directory on the cluster node and also into a location in HDFS.