Power BI Monitoring for Apache Airflow
Apache Airflow orchestrates the pipelines that feed your BI layer. When a DAG fails or runs behind schedule, the downstream effect is a stale Power BI report, a missing Tableau Cloud extract, or a dbt model that never ran. MetricSign monitors every DAG run via the Airflow REST API v2 and surfaces three incident types: dag_failed (a DAG run returned a failed state), dag_slow (a run exceeded its historical duration baseline using the same MAD algorithm as Databricks), and dag_missing_run (a scheduled DAG didn't start within its expected window). All four managed Airflow variants — self-hosted, MWAA, Astronomer, and Cloud Composer — use the same REST API, so a single connector type covers your entire Airflow fleet. Task-level error detail is available on-demand in the incident view without requiring any pipeline instrumentation.
What you can monitor
- Detect when a DAG run fails and immediately see which downstream Power BI datasets or Tableau Cloud extracts are affected via cross-stack lineage
- Alert when a DAG runs significantly slower than its historical baseline before the SLA window closes
- Catch silent schedule drift — DAGs that quietly stop running without returning a failed state
- Monitor MWAA, Astronomer, and Cloud Composer alongside self-hosted Airflow in a single incident feed