Airflow

Airflow Components

Airflow architecture

Executors:

Database

Pipeline design:

Local development:

Infrastructure

Kubernetes

ci/cd pipeline

Logging

Monitoring and alerting

Resource allocation

Deployment

To avoid the restart of the scheduler and workers every time some one make a change in a DAG, every DAG ends up in a bucket

Configuration

Worker concurrency configuration

For custom jobs that require more resources we can use KubernetesPodOperator that gives runtime isolation this way we can isolate dependencies from airflow image

Problems

References