Use the Livy interpreter to access Spark

This section describes how to use the Livy interpreter to access Apache Spark.

The Livy interpreter offers several advantages over the default Spark interpreter (%spark):

  • Sharing of Spark context across multiple Zeppelin instances.

  • Reduced resource use, by recycling resources after 60 minutes of activity (by default). The default Spark interpreter runs jobs--and retains job resources--indefinitely.

  • User impersonation. When the Zeppelin server runs with authentication enabled, the Livy interpreter propagates user identity to the Spark job so that the job runs as the originating user. This is especially useful when multiple users are expected to connect to the same set of data repositories within an enterprise. (The default Spark interpreter runs jobs as the default Zeppelin user.)

  • The ability to run Spark in yarn-cluster mode.

Prerequisites:

  • Before using SparkR through Livy, R must be installed on all nodes of your cluster. For more information, see "SparkR Prerequisites" in the HDP Apache Spark guide.

  • Before using Livy in a note, check the Interpreter page to ensure that the Livy interpreter is configured properly for your cluster.

Note: The Interpreter page is subject to access control settings. If the Interpreters page does not list access settings, check with your system administrator for more information.

To access PySpark using Livy, specify the corresponding interpreter directive before the code that accesses Spark; for example:

%livy.pyspark
print "1"

1

Similarly, to access SparkR using Livy, specify the corresponding interpreter directive:

%livy.sparkr
hello <- function( name ) {
    sprintf( "Hello, %s", name );
}

hello("livy")

Livy sessions are recycled after a specified period of session inactivity. The default is one hour.

For more information about using Livy with Spark, see "Submitting Spark Applications Through Livy" in the HDP Apache Spark guide.

Importing External Packages

To import an external package for use in a note that runs with Livy:

  1. Navigate to the interpreter settings.

  2. If you are running the Livy interpreter in local mode (as specified by livy.spark.master), add jar files to the /usr/hdp/<version>/livy/repl-jars directory.

  3. If you are running the Livy interepreter in yarn-cluster mode, either complete step 2 or edit the Livy configuration on the Interpreters page as follows:

    1. Add a new key, livy.spark.jars.packages.

    2. Set its value to <group>:<id>:<version>.

      Here is an example for the spray-json library, which implements JSON in Scala:

      io.spray:spray-json_2.10:1.3.1