Skip to content

A reverse proxy server which allows secure connectivity to a Spark Connect server

License

Notifications You must be signed in to change notification settings

gizmodata/spark-connect-proxy

Repository files navigation

spark-connect-proxy

A reverse proxy server which allows secure connectivity to a Spark Connect server.

spark-connect-proxy-ci Supported Python Versions PyPI version PyPI Downloads

Why?

Because Spark Connect does NOT provide authentication and/or TLS encryption out of the box. This project provides a reverse proxy server which can be used to secure the connection to a Spark Connect server.

Setup (to run locally)

Install Python package

You can install spark-connect-proxy from PyPi or from source.

Option 1 - from PyPi

# Create the virtual environment
python3 -m venv .venv

# Activate the virtual environment
. .venv/bin/activate

pip install spark-connect-proxy[client]

Option 2 - from source - for development

git clone https://github.com/gizmodata/spark-connect-proxy

cd spark-connect-proxy

# Create the virtual environment
python3 -m venv .venv

# Activate the virtual environment
. .venv/bin/activate

# Upgrade pip, setuptools, and wheel
pip install --upgrade pip setuptools wheel

# Install Spark Connect Proxy - in editable mode with client and dev dependencies
pip install --editable .[client,dev]

Note

For the following commands - if you running from source and using --editable mode (for development purposes) - you will need to set the PYTHONPATH environment variable as follows:

export PYTHONPATH=$(pwd)/src

Usage

This repo contains scripts to let you provision an AWS EMR Spark cluster with a secure Spark Connect Proxy server to allow you to securely and remotely connect to it.

First - you'll need to open up a port for public access to the AWS EMR Spark Cluster - in addition to the ssh port: 22. Add port: 50051 as shown here:
Open port 50051

Note

Even though you are opening this port to the public, the Spark Connect Proxy will secure it with TLS and JWT Authentication.

The scripts use the AWS CLI to provision the EMR Spark cluster - so you will need to have the AWS CLI installed and configured with your AWS credentials.

You can create a file in your local copy of the scripts directory called .env with the following contents:

export AWS_ACCESS_KEY_ID="put value from AWS here"
export AWS_SECRET_ACCESS_KEY="put value from AWS here"
export AWS_SESSION_TOKEN="put value from AWS here"
export AWS_REGION="us-east-2"

To provision the EMR Spark cluster - run the following command from the root directory of this repo:

scripts/provision_emr_spark_cluster.sh

That will output several files (which will be git ignored for security reasons):

  • tls/ca.crt - the EMR Spark cluster generated TLS certificate - needed for your PySpark client to trust the Spark Connect Proxy server (b/c it is self-signed)
  • scripts/output/instance_details.txt - shows the ssh command for connecting to the master node of the EMR Spark cluster
  • scripts/output/spark_connect_proxy_details.log - shows how to run a PySpark Ibis client example - which connects securely from your local computer to the remote EMR Spark cluster. Example command:
spark-connect-proxy-ibis-client-example \
  --host ec2-01-01-01-01.us-east-2.compute.amazonaws.com \
  --port 50051 \
  --use-tls \
  --tls-roots tls/ca.crt \
  --token honey.badger.dontcare

Important

You must have installed the spark-connect-proxy package with the [client] extras onto the client computer to run the spark-connect-proxy-ibis-client-example command.

Handy development commands

Version management

Bump the version of the application - (you must have installed from source with the [dev] extras)
bumpver update --patch