CDM Jupyterhub dockerfiles (Prototype)

⚠️ DEPRECATED: This repository has been deprecated and is no longer maintained. Please use the new repository at https://github.com/BERDataLakehouse/spark_notebook

This prototype establishes a Docker container configuration for JupyterHub, designed to furnish a multi-user environment tailored for executing Spark jobs via Jupyter notebooks.

User How-To Guide

Accessing Spark Jupyter Notebook

MinIO Guide

Using docker-compose.yaml

To deploy the JupyterHub container and Spark nodes locally, execute the following command:

docker-compose up --build

Test Submitting a Spark Job Locally

Submitting a Spark Job via spark-test-node

docker exec -it spark-test-node \
    sh -c '
	    /opt/bitnami/spark/bin/spark-submit \
	    --master $SPARK_MASTER_URL \
	    examples/src/main/python/pi.py 10 \
	    2>/dev/null
    '

Submitting a Spark Job via Jupyter Notebook

After launching the Jupyter Notebook, establish a Spark context or session with the Spark master set to the environment variable SPARK_MASTER_URL and proceed to submit your job. Once the job is submitted, you can monitor the job status and logs in the Spark UI.

Sample code to calculate Pi using SparkContext:

from pyspark import SparkConf, SparkContext
import random
import os

spark_master_url = os.environ['SPARK_MASTER_URL']

conf = SparkConf().setMaster(spark_master_url).setAppName("Pi")
sc = SparkContext(conf=conf)

num_samples = 100000000
def inside(p):     
  x, y = random.random(), random.random()
  return x*x + y*y < 1
count = sc.parallelize(range(0, num_samples)).filter(inside).count()
pi = 4 * count / num_samples
print(pi)
sc.stop()

Development

Install requirements

In order to build the uv dependencies, including when adding new modules, graphviz-dev must be installed. For example:

macOS:

# Install Graphviz system library (required for pygraphviz \ependency)
brew install graphviz

# Optionally, set environment variables to help uv find Graphviz headers and libraries
# This fixes build errors where the compiler can't locate the Graphviz headers
export CPATH=$(brew --prefix graphviz)/include:$CPATH
export LIBRARY_PATH=$(brew --prefix graphviz)/lib:$LIBRARY_PATH

Linux (Debian/Ubuntu):

# Install Graphviz development libraries
sudo apt install graphviz-dev

Running tests

Python 3.12.10 must be installed on the system.

Install dependencies

# Install Python dependencies
uv sync --locked # only the first time or when uv.lock changes

Run tests

PYTHONPATH=src uv run pytest tests

Racher Deployment

Environment Variables

• SPARK_MASTER_URL: spark://spark-master:7077
• NOTEBOOK_PORT: 4041
• SPARK_DRIVER_HOST: notebook (the hostname of the Jupyter notebook container).

Spark Session/Context Configuration

When running Spark in the Jupyter notebook container, the default spark.driver.host configuration is set to the hostname (SPARK_DRIVER_HOST) of the container. In addition, the environment variable SPARK_MASTER_URL should also be configured.

Using Predefined SparkSession from spark.utils.get_spark_session method

from spark.utils import get_spark_session

spark = get_spark_session(app_name)

# To build spark session for Delta Lake operations, set the delta_lake parameter to True
spark = get_spark_session(app_name, delta_lake=True)

Manually Configuring SparkSession/SparkContext

If you want to configure the SparkSession manually, you can do so as follows:

Example SparkSession Configuration

spark = SparkSession.builder \
    .master(os.environ['SPARK_MASTER_URL']) \
    .appName("TestSparkJob") \
    .getOrCreate()

Example SparkContext Configuration

conf = SparkConf() \
    .setMaster(os.environ['SPARK_MASTER_URL']) \
    .setAppName("TestSparkJob")
sc = SparkContext(conf=conf)

Submitting a Job Using Terminal

/opt/bitnami/spark/bin/spark-submit \
    --master $SPARK_MASTER_URL \
    /opt/bitnami/spark/examples/src/main/python/pi.py 10 \
    2>/dev/null

Connecting to the Jupyter Notebook

Tunneling to the Jupyter Notebook

ssh -f -N -L localhost:44041:10.58.2.201:4041 <kbase_developer_username>@login1.berkeley.kbase.us

where kbase_developer_username is an Argonne account, typically starting with ac..

Accessing the Jupyter Notebook

Navigate to http://localhost:44041/ in your browser.

Enjoy your Spark journey!

For more information, please consult the User Guide.

Regenerate cdm-spark-cluster-manager-api-client with openapi-python-client

Prerequisites

1. Python 3.9+ installed on your system.
2. The openapi-python-client package installed. If not already installed, you can do so using pip:

   pip install openapi-python-client
   

3. Access to the OpenAPI specification for the cdm-kube-spark-manager, either via a URL or a local file.
   • TODO - Post the URL after deployment to Rancher2

Generate the Client

From a URL:

openapi-python-client generate --url https://api.example.com/openapi.json --output-path cdm-spark-cluster-manager-api-client

From a Local File:

openapi-python-client generate --path ./openapi.yaml --output-path cdm-spark-cluster-manager-api-client 

Copy the generated client files to cdm_spark_cluster_manager_api_client

cp -r path_of_openapi_ouptput_path/cdm-spark-cluster-manager-api-client/cdm_spark_cluster_manager_api_client path_of_cdm-jupyterhub/src/spark
cp path_of_openapi_ouptput_path/cdm-spark-cluster-manager-api-client/README.md path_of_cdm-jupyterhub/src/spark/cdm_spark_cluster_manager_api_client