workspace/data-engineering/05-airflow-pipelines

Automate Pipelines with Apache Airflow

Udacity Data Engineering Nanodegree — Project 5

Overview

Build a production-grade, orchestrated ETL pipeline using Apache Airflow. The pipeline runs on an hourly schedule, loading data from AWS S3 into AWS Redshift staging tables, transforming it into a star schema, and validating data quality — all using modular, reusable custom operators.

Architecture

S3 (raw JSON)
    │
    ▼
[StageToRedshiftOperator]    ← Stage events & songs
    │
    ▼
[LoadFactOperator]           ← Load songplays fact table
    │
    ▼
[LoadDimensionOperator] ×4   ← Load users, songs, artists, time
    │
    ▼
[DataQualityOperator]        ← Assert tables are non-empty

DAG

Property	Value
Schedule	Hourly (0 * * * *)
Start date	2019-01-12
Catchup	Disabled

Custom Operators

Operator	File	Purpose
StageToRedshiftOperator	plugins/operators/stage_redshift.py	COPY JSON from S3 to Redshift staging tables
LoadFactOperator	plugins/operators/load_fact.py	INSERT into fact table from staging
LoadDimensionOperator	plugins/operators/load_dimension.py	INSERT into dimension tables (supports truncate-insert or append)
DataQualityOperator	plugins/operators/data_quality.py	Assert row counts > 0 for all tables

Project Structure

05-airflow-pipelines/
├── create_tables.sql          # DDL for Redshift tables
├── dags/
│   └── udac_example_dag.py    # Main DAG definition
└── plugins/
    ├── helpers/
    │   └── sql_queries.py     # Shared SQL INSERT statements
    └── operators/
        ├── stage_redshift.py
        ├── load_fact.py
        ├── load_dimension.py
        └── data_quality.py

Key Concepts

• DAG design — directed acyclic graphs for workflow orchestration
• Custom operators — reusable, parameterized Airflow tasks
• Modular SQL helpers — shared query library via SqlQueries class
• Data quality checks — automated validation at end of every run
• AWS integration — S3 COPY + Redshift connections via Airflow Connections

How to Run

1. Set up an Airflow environment with AWS connections configured:

   • aws_credentials — IAM access key & secret
   • redshift — Redshift cluster connection string

2. Copy DAG and plugins into your Airflow home:

   cp -r dags/ $AIRFLOW_HOME/dags/
   cp -r plugins/ $AIRFLOW_HOME/plugins/
   

3. Run create_tables.sql against your Redshift cluster to create staging and DW tables.

4. Enable the DAG in the Airflow UI — it will trigger hourly.