Skip to content

README.md

Latest commit

 

History

History
99 lines (72 loc) · 4.82 KB

README.md

File metadata and controls

99 lines (72 loc) · 4.82 KB
Raw

Databricks-Connect Container

This container is designed for developing PySpark application in VS Code using Databricks-Connect. Databricks VSCode extension can be used alongside to sync your workspace. Technically it should work for Scala, Java & R - though I haven't tried.

It can be used as a Docker Container or as a cloud hosted container using CodeSpaces.

Requirements

Mac OS

Windows 10

Alternative

For a more advanced option you can use any OS and a Remote Docker Host.

This maybe useful if you cannot run local containers and do not want to use a cloud container.

Getting Started

  • Set these enviroment variables in a terminal before launching VS Code. Refer to this guide to locate the connection properties below.
export DATABRICKS_ADDRESS=<YOUR_DATABRICKS_URL>
export DATABRICKS_API_TOKEN=<YOUR_DATABRICKS_API_TOKEN>
export DATABRICKS_CLUSTER_ID=<YOUR_DATABRICKS_CLUSTER_ID>
export DATABRICKS_ORG_ID=<YOUR_DATABRICKS_ORG_ID>
export DATABRICKS_PORT=<YOUR_DATABRICKS_PORT>

IMPORTANT: Correct the Databricks Connect version in devcontainer.json to match Databricks Runtime your cluster is running. Changing any setting in the devcontainer.json after the container has been build requires you to rebuild the container for it take effect

To open using Docker locally:

  • Click on the Green icon in the bottom left of VSCode and select "Reopen in Container"

To open in a CodeSpace:

  • Commit your folder to a repo first
  • Open the Remote Explorer (left hand toolbar)
  • Ensure CodeSpaces is selected in the top drop down
  • Click + (Create new CodeSpace)
  • Follow the prompts

Set up and configure the Databricks VSCode Extension using instructions in the Setup and Configuration (VSCode Connection Manager) section.

Test it out

Test Databricks connection

First from command prompt check that you databricks connect install can connect:

databricks-connect test

Run some Python code

To run python code create a test.py file and paste this code:

from pyspark.sql import SparkSession
spark = SparkSession\
.builder\
.getOrCreate()

print("Testing simple count")

# The Spark code will execute on the Azure Databricks cluster.
print(spark.range(100).count())

Press F5 and select the Python debugger.

Run a Jupyter notebook

Navigate to the Databricks VSCode extension and navigate to the Workspace to find a notebook to open. Double-click the notebook you'd like to open in the editor.

NOTE: A notebook may display in JSON format. If this happens right click on the tab at the top with the name of the file, select Reopen Editor With... and click on Jupyter Notebook

Before running a cell, on the top right corner click on Select Kernel and select `Python 3.x.x 64-bit ('dbconnect': conda). Note the Python version maybe different depending on what Databricks Runtime is being used.

Run a cell to execute your code Databricks cluster. VSCode may ask you to install ipykernel - accept the prompt if it does.

Why?

Because setting up Databricks-Connect (particulary on Windows is a PIA). This allow:

  • A common setup between team members
  • Better development enviroment
  • Multiple side by side versions
  • Ability to reset your environment
  • Even run the whole thing from a browser!

More Information

About VSCode and Containers: https://code.visualstudio.com/docs/remote/containers