LakeHouse Sharing

Lakehouse is getting more and more popular nowadays, and many Oragnization seeing massive value in building and maintaining the lakehouse instead of maintaining the Warehouse technologies for various reasons. There are lot of great articles covering this topic in the internet or ask ChatGPT to know more about this area.

This repo in particular belongs to one such part of lakehouse ecosystem, where the concern is how to share data securely, within the org and outside of the organization. Heavily inspired, infact the complete idea is from Delta-sharing protocol. delta-sharing solves the data-sharing problem for the people using Delta Table format and Databricks (company behind Delta-lake) provides excellent self-service tools on top of open source delta-sharig

Motivation of this Repo:

Delta-sharing Framework is a protocol which can be adopted to other Table formats (Hudi and iceberg) apart from Delta
- A Quick POC (Super alpha stage) to prove that protocol can be implemented easily and adopted to other table formats as well
Original Delta-sharing server was written in Scala, This repo inspired that server implementation and written in Python language which can be good starting point for python developers, and can be adopted by the organizations where python deployment stack (infra stack) is already available.
This Repo provides Alpha implementation of Sharing protocol for Iceberg table format using Pyiceberg. This can be enhanced/improved based on the interest.
This Repo rewrites the sharing protocol for Delta table format as well using delta-rs (actually this is super to implement this protocol on top of this python package)

Difference between normal querying and querying via delta-sharing

Lot of Engines like Spark,Trino,Presto,Dremio queries Table format efficiently why we need delta-sharing.
In Delta sharing, you can logically group required few tables and send them as shares to the other teams or other org.
you need to share the token to the user, they can use that token to authenticate with delta-sharing server, once authenticated they can read the data from lakehouse without worrying about s3 authentication IAM cross account configurations etc..

Installation

Run this commands in the root folder of this project

help:

Usage:
  make <target>

Targets:
  venv                           create a virtual environment for development
  start_backend_server           starts prefect server
  start_frontend_server          starts prefect agent
  help                           Show help

install requirements

make venv

To share iceberg table format install following extra package and setup catalog like AWS Glue or Hive, refer PyIceberg documentation

for iceberg

# install iceberg
pip install pyiceberg

To share delta-lake table format, install delta-lake package and delta-lake doesn't need any catalog it will directly fetch the metadata from table formats metadata in cloud storage files.

for delta-lake

pip install deltalake

start backend server

make start_backend_server

start Frontend server

In another termianl start frontend streamlit APP.

make start_frontend_server

Use docker setup

use docker setup to quickly setup the app

docker-compose up

Set few of the Environment variables before starting the docker-compose up refer .env.example file for setting the variables

APP urls:

Once docker-compose was up and running successfully, we can expect following urls

RDS Admin : http://localhost:8081/
Postgres : host: localhost, port:5433
Backend (FastAPI): http://localhost:8001/docs
Frontend (streamlit): http://localhost:8501
Set the following env variables accordingly and Run sqls/prepopulate_data.py for creating superuser and few test tables
- for local setup set export env=local
- for docker setup set export env=docker

Frontend:

Login Username : admin
Login password : admin@123