Use Sparkling Water via Spark Packages¶
Sparkling Water is also published as a Spark package. You can use it directly from your Spark distribution. The name of the published package is ai.h2o:sparkling-water-package, and it references all published Sparkling Water modules. Moreover, each module can be used as a Spark package if necessary.
If you would like to use Sparkling Water version 3.46.0.5-1-2.4 on Spark 2.4 and launch example
CraigslistJobTitlesStreamingApp
, then you can use the following
command:
$SPARK_HOME/bin/spark-submit --packages ai.h2o:sparkling-water-package_2.11:3.46.0.5-1-2.4 --class ai.h2o.sparkling.examples.CraigslistJobTitlesStreamingApp /dev/null
The Spark option --packages
points to the Duke package and the published Sparkling Water packages in the Maven repository.
A similar command works for spark-shell
:
$SPARK_HOME/bin/spark-shell --packages ai.h2o:sparkling-water-package_2.11:3.46.0.5-1-2.4
Note: When you are using Spark packages, you do not need to download the Sparkling Water distribution. The Spark installation is sufficient.