Skip to content
/ fuga Public

Python framework for building Google Cloud Composer workflows.

License

Notifications You must be signed in to change notification settings

reproio/fuga

Repository files navigation

(This project is in public-beta status)

Fuga ['fu:ga]

pypi CircleCI Documentation Status

Fuga is a toolset (and API wrappers) for Google Cloud Composer (Airflow), which allows Composer users to develop/deploy workflows easier and in more organized way.

Quickstart

Install Requirements

Note: pyenv always provides an entrypoint for both python2/3 and throw an error when one doesn't actually exists, which makes gcloud cli crashes at its runtime. To prevent this, use something like pyenv global 3.x.x 2.x.x to provide both python2/3 or alias python2=python to let your system choose where to route the command.

Install fuga (cli)

$ # (on your own machine)
$ pip install fuga

Install fuga templates

Fuga powers cookiecutter to offer various templates/boilerplates for fuga experiments. You need to install it to your environment before using fuga.

$ git clone git@github.com:reproio/fuga-cookiecutter-experiment-default.git \
  ~/.cookiecutters/fuga-cookiecutter-experiment-default
$ git clone git@github.com:reproio/fuga-cookiecutter-pod-operator-default.git \
  ~/.cookiecutters/fuga-cookiecutter-pod-operator-default

Create Cloud Composer Environment

Creating new Cloud Composer Environment is not supported by the tool at the moment.

Install fuga to Composer Environment

You need to install fuga to Compooser Environment if you want to use fuga airflow utilities.

Initialize Fuga environment

Fuga needs to know which GCP project and GCS bucket to use with it. fuga environment init command with let you choose one or create new for each.

$ fuga environment init
# follow instructions

Create your experiment

$ fuga experiment new my_experiment
...

Deploy your experiment

$ cd my_experiment
$ fuga experiment deploy
...

(optional) Create your implementation for KubernetesPodOperator

If you want to use an operator with external dependency which is not able to be resolved using just PyPI packages, (e.g. MeCab) you may need to use KuberenetesPodOperator.

Fuga supports generating a scaffold for KuberenetesPodOperator and deploy(and rollback(WIP)) it on Google Cloud Container Registry.

$ cd my_experiment
$ fuga pod-operator new my_pod_operator
...
$ fuga pod-operator deploy my_pod_operator
...