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

CapsuleHealth/airflow-exporter

 
 

Repository files navigation

Airflow prometheus exporter

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

Screenshots

Compatibility

Note: this version is compatible with Airflow 1.10.3+ only, see #46 for details

For compatibility with previous versions of Airflow use older version: v0.5.4

  • Airflow: airflow1.10.3+
  • Python: python2, python3
  • DataBase: postgresql, mysql

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
  • hostname

Value: number of tasks in specific status.

airflow_dag_status

Labels:

  • dag_id
  • owner
  • status

Value: number of dags in 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 completed.

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 84.0%
  • Shell 14.7%
  • Dockerfile 1.3%