Skip to content

Airflow plugin to export dag and task based metrics to Prometheus.

License

Notifications You must be signed in to change notification settings

boycs007/airflow-exporter

 
 

Repository files navigation

Airflow prometheus exporter

Exposes dag and task based metrics from Airflow to a Prometheus compatible endpoint.

Compatibility with Airflow versions

>=2.0

Current version is compatible with Airflow 2.0+

<=1.10.14, >=1.10.3

Version v1.3.2 is compatible

Note: Airflow 1.10.14 with Python 3.8 users

You should install importlib-metadata package in order for plugin to be loaded. See #85 for details.

<1.10.3

Version v0.5.4 is compatible

Install

pip install airflow-exporter

That's it. You're done.

Exporting extra labels to Prometheus

It is possible to add extra labels to DAG-related metrics by providing labels dict to DAG params.

Example

dag = DAG(
    'dummy_dag',
    schedule_interval=timedelta(hours=5),
    default_args=default_args,
    catchup=False,
    params={
        'labels': {
            'env': 'test'
        }
    }
)

Label env with value test will be added to all metrics related to dummy_dag:

airflow_dag_status{dag_id="dummy_dag",env="test",owner="owner",status="running"} 12.0

Metrics

Metrics will be available at

http://<your_airflow_host_and_port>/admin/metrics/

airflow_task_status

Labels:

  • dag_id
  • task_id
  • owner
  • status

Value: number of tasks in a specific status.

airflow_dag_status

Labels:

  • dag_id
  • owner
  • status

Value: number of dags in a specific status.

airflow_dag_run_duration

Labels:

  • dag_id: unique identifier for a given DAG

Value: duration in seconds of the longest DAG Run for given DAG. This metric is not available for DAGs that have already finished.

airflow_dag_last_status

Labels:

  • dag_id
  • owner
  • status

Value: 0 or 1 depending on wherever the current state of each dag_id is status.

License

Distributed under the BSD license. See LICENSE for more information.

About

Airflow plugin to export dag and task based metrics to Prometheus.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 89.0%
  • Shell 10.1%
  • Dockerfile 0.9%