Skip to content
This repository has been archived by the owner on Apr 11, 2024. It is now read-only.

Latest commit

 

History

History
259 lines (163 loc) · 13.9 KB

File metadata and controls

259 lines (163 loc) · 13.9 KB

Universal Transfer Operator

transfers made easy

CI

The UniversalTransferOperator simplifies how users transfer data from a source to a destination using Apache Airflow. It offers a consistent agnostic interface, improving the users' experience so they do not need to use explicitly specific providers or operators.

At the moment, it supports transferring data between file locations and databases (in both directions) and cross-database transfers.

This project is maintained by Astronomer.

Prerequisites

  • Apache Airflow >= 2.4.0.

Install

The apache-airflow-provider-transfers is available at PyPI. Use the standard Python installation tools.

To install a cloud-agnostic version of the apache-airflow-provider-transfers, run:

pip install apache-airflow-provider-transfers

You can also install dependencies for using the UniversalTransferOperator with popular cloud providers:

pip install apache-airflow-provider-transfers[amazon,google,snowflake]

Quickstart

Users can get started quickly with following two approaches:

  1. Spinning up a local Airflow infrastructure using the open-source Astro CLI and Docker
  2. Using vanilla Airflow and Python

Run UniversalTransferOperator using astro

Run UniversalTransferOperator using vanilla airflow and python

  • Install airflow and setup project following this documentation.

  • Ensure that your Airflow environment is set up correctly by running the following commands:

    export AIRFLOW_HOME=`pwd`
    airflow db init
  • Add the following in requirements.txt

    apache-airflow-provider-transfers[all]
    
  • Copy file named example_transfer_and_return_files.py and example_snowflake_transfers.py and add it to the dags directory of your Airflow project:

https://github.com/astronomer/apache-airflow-provider-transfers/blob/04b53d780790eaa3b424458742bc89d6fbec2ccd/example_dags/example_transfer_and_return_files.py#L1-L45

https://github.com/astronomer/apache-airflow-provider-transfers/blob/a80dc84b7f33bb86ae244f79411b240f4f4c7e22/example_dags/example_snowflake_transfers.py#L1-L46

Alternatively, you can download example_transfer_and_return_files.py and example_snowflake_transfers.py.

 curl -O https://github.com/astronomer/apache-airflow-provider-transfers/blob/main/example_dags/example_transfer_and_return_files.py

 curl -O https://github.com/astronomer/apache-airflow-provider-transfers/blob/main/example_dags/example_snowflake_transfers.py

Example DAGs

Checkout the example_dags folder for examples of how the UniversalTransferOperator can be used.

How Universal Transfer Operator Works

Approach

With UniversalTransferOperator, users can perform data transfers using the following transfer modes:

  1. Non-native
  2. Native
  3. Third-party

Non-native transfer

Non-native transfers rely on transferring the data through the Airflow worker node. Chunking is applied where possible. This method can be suitable for datasets smaller than 2GB, depending on the source and target. The performance of this method is highly dependent upon the worker's memory, disk, processor and network configuration.

Internally, the steps involved are:

  • Retrieve the dataset data in chunks from dataset storage to the worker node.
  • Send data to the cloud dataset from the worker node.

Following is an example of non-native transfers between Google cloud storage and Sqlite:

https://github.com/astronomer/apache-airflow-provider-transfers/blob/a80dc84b7f33bb86ae244f79411b240f4f4c7e22/example_dags/example_universal_transfer_operator.py#L68-L74

Improving bottlenecks by using native transfer

An alternative to using the Non-native transfer method is the native method. The native transfers rely on mechanisms and tools offered by the data source or data target providers. In the case of moving from object storage to a Snowflake database, for instance, a native transfer consists in using the built-in COPY INTO command. When loading data from S3 to BigQuery, the Universal Transfer Operator uses the GCP Storage Transfer Service.

The benefit of native transfers is that they will likely perform better for larger datasets (2 GB) and do not rely on the Airflow worker node hardware configuration. With this approach, the Airflow worker nodes are used as orchestrators and do not perform the transfer. The speed depends exclusively on the service being used and the bandwidth between the source and destination.

Steps:

  • Request destination dataset to ingest data from the source dataset.
  • Destination dataset requests source dataset for data.

NOTE: The Native method implementation is in progress and will be available in future releases.

Transfer using a third-party tool

The UniversalTransferOperator can also offer an interface to generic third-party services that transfer data, similar to Fivetran.

Here is an example of how to use Fivetran for transfers:

https://github.com/astronomer/apache-airflow-provider-transfers/blob/7d5188c4af214d5cdaeb714654e9bdf5b48cb3fb/example_dags/example_dag_fivetran.py#L35-L56

Supported technologies

Documentation

The documentation is a work in progress -- we aim to follow the Diátaxis system.

Changelog

The Universal Transfer Operator follows semantic versioning for releases. Check the changelog for the latest changes.

Release management

See Managing Releases to learn more about our release philosophy and steps.

Contribution guidelines

All contributions, bug reports, bug fixes, documentation improvements, enhancements, and ideas are welcome.

Read the Contribution Guideline for a detailed overview of how to contribute.

Contributors and maintainers should abide by the Contributor Code of Conduct.

License

Apache Licence 2.0