path: root/docs/running-on-yarn.md

---
layout: global
title: Launching Spark on YARN
---

Experimental support for running over a [YARN (Hadoop
NextGen)](http://hadoop.apache.org/docs/r2.0.2-alpha/hadoop-yarn/hadoop-yarn-site/YARN.html)
cluster was added to Spark in version 0.6.0.  This was merged into master as part of 0.7 effort.
To build spark core with YARN support, please use the hadoop2-yarn profile.
Ex:  mvn -Phadoop2-yarn clean install

# Building spark core consolidated jar.

Currently, only sbt can buid a consolidated jar which contains the entire spark code - which is required for launching jars on yarn.
To do this via sbt - though (right now) is a manual process of enabling it in project/SparkBuild.scala.
Please comment out the
  HADOOP_VERSION, HADOOP_MAJOR_VERSION and HADOOP_YARN
variables before the line 'For Hadoop 2 YARN support'
Next, uncomment the subsequent 3 variable declaration lines (for these three variables) which enable hadoop yarn support.

Currnetly, it is a TODO to add support for maven assembly.


# Preparations

- Building spark core assembled jar (see above).
- Your application code must be packaged into a separate JAR file.

If you want to test out the YARN deployment mode, you can use the current Spark examples. A `spark-examples_{{site.SCALA_VERSION}}-{{site.SPARK_VERSION}}` file can be generated by running `sbt/sbt package`. NOTE: since the documentation you're reading is for Spark version {{site.SPARK_VERSION}}, we are assuming here that you have downloaded Spark {{site.SPARK_VERSION}} or checked it out of source control. If you are using a different version of Spark, the version numbers in the jar generated by the sbt package command will obviously be different.

# Launching Spark on YARN

The command to launch the YARN Client is as follows:

    SPARK_JAR=<SPARK_YAR_FILE> ./run spark.deploy.yarn.Client \
      --jar <YOUR_APP_JAR_FILE> \
      --class <APP_MAIN_CLASS> \
      --args <APP_MAIN_ARGUMENTS> \
      --num-workers <NUMBER_OF_WORKER_MACHINES> \
      --master-memory <MEMORY_FOR_MASTER> \
      --worker-memory <MEMORY_PER_WORKER> \
      --worker-cores <CORES_PER_WORKER> \
      --user <hadoop_user> \
      --queue <queue_name>

For example:

    SPARK_JAR=./core/target/spark-core-assembly-{{site.SPARK_VERSION}}.jar ./run spark.deploy.yarn.Client \
      --jar examples/target/scala-{{site.SCALA_VERSION}}/spark-examples_{{site.SCALA_VERSION}}-{{site.SPARK_VERSION}}.jar \
      --class spark.examples.SparkPi \
      --args standalone \
      --num-workers 3 \
      --master-memory 4g \
      --worker-memory 2g \
      --worker-cores 1

The above starts a YARN Client programs which periodically polls the Application Master for status updates and displays them in the console. The client will exit once your application has finished running.

# Important Notes

- When your application instantiates a Spark context it must use a special "standalone" master url. This starts the scheduler without forcing it to connect to a cluster. A good way to handle this is to pass "standalone" as an argument to your program, as shown in the example above.
- YARN does not support requesting container resources based on the number of cores. Thus the numbers of cores given via command line arguments cannot be guaranteed.
- Currently, we have not yet integrated with hadoop security. If --user is present, the hadoop_user specified will be used to run the tasks on the cluster. If unspecified, current user will be used (which should be valid in cluster).
  Once hadoop security support is added, and if hadoop cluster is enabled with security, additional restrictions would apply via delegation tokens passed.