airflow-docker 0.4.3

An opinionated implementation of exclusively using airflow DockerOperators for all Operators

airflow-docker

Description

An opinionated implementation of exclusively using airflow DockerOperators for all Operators.

Default Operator

from airflow_docker.operator import Operator

task = Operator(
    image='some-image:latest',
    ...
)

Default Sensor

from airflow_docker.operator import Sensor

sensor = Sensor(
    image='some-image:latest',
    ...
)

Task Code

from airflow_docker_helper import client

client.sensor(True)

Branch Operator

Dag Task

from airflow_docker.operator import BranchOperator

branching_task = BranchOperator(
    image='some-image:latest',
    ...
)

Task Code

from airflow_docker_helper import client

client.branch_to_tasks(['task1', 'task2'])

Short Circuit Operator

Dag Task

from airflow_docker.operator import ShortCircuitOperator

short_circuit = ShortCircuitOperator(
    image='some-image:latest',
    ...
)

Task Code

from airflow_docker_helper import client

client.short_circuit()  # This task will short circuit if this function gets called

Context Usage

Dag Task

from airflow_docker.operator import Operator

task = Operator(
    image='some-image:latest',
    provide_context=True,
    ...
)

Task Code

from airflow_docker_helper import client

context = client.context()

Configuration

The following operator defaults can be set under the airflowdocker namespace:

• force_pull (boolean true/false)
• auto_remove (boolean true/false)
• network_mode

For example, to set force_pull to False by default set the following environment variable like so:

export AIRFLOW__AIRFLOWDOCKER__FORCE_PULL=false

Plugin

This package works as an airflow plugin as well. When installed and running airflow, dags can import like so

from airflow.{type, like "operators", "sensors"}.{name specificed inside the plugin class} import *

i.e.

from airflow.operators.airflow_docker import Operator