erikerlandson/openshift-jupyter-scala

A Dockerfile to build a jupyter-scala image runnable on OpenShift

openshift-jupyter-scala

Build

% cd /path/to/openshift-jupyter-scala
% docker build -t registry/openshift-jupyter-scala:latest .
% docker push registry/openshift-jupyter-scala:latest

Run in OpenShift

oc run jupyter-scala --image=registry/openshift-jupyter-scala:latest --expose --port=8888

Image may take a couple minutes for OpenShift to pull down.

Next, create a route for the jupyter-scala service. You should be able to open this route in your browser and get the Jupyter home-page, including a Scala kernel option.

This image defaults the jupyter password to just 'jupyter'. If you want to set another password you can configure the JUPYTER_NOTEBOOK_PASSWORD environment variable on your pod.

The image is based on the radanalytics.io base notebook image base-notebook

The latest rev of this image is based on the jupyter-scala project, which is now being packaged as almond on the develop branch. It is built on top of the ammonite Scala REPL.

In order to take advantage of the spark release pre-installed on this image, you need to get spark's jar files onto the ammonite classpath. This can be done by putting the following code on the first cell of your notebook:

// put the spark install from the base notebook image onto Ammonite's classpath
java.nio.file.Files.list(java.nio.file.Paths.get("/opt/spark/jars")).toArray.map(_.toString).foreach { fname =>
  val path = java.nio.file.FileSystems.getDefault().getPath(fname)
  val x = ammonite.ops.Path(path)
  interp.load.cp(x)
}
// Load the ammonite-spark package to get AmmoniteSparkSession
import $ivy.`sh.almond::ammonite-spark:0.1.1`

An example of connecting to a spark cluster using the AmmoniteSparkSession wrapper:

import org.apache.spark.sql._
val spark = {
    AmmoniteSparkSession.builder()
      .master("local[*]")
      .getOrCreate()
  }

Other examples of how to use this kernel can be found here.