Monitor health with Apache Airflow and branch to stop or retrigger a DAG.
HighAvailabilityOperator runs a Python check returning (Result, Action).
Health and recovery are separate: a failed result can retrigger the DAG while
its current run follows a failure branch. Time limits stop monitoring, and
maxretrigger bounds recovery attempts while preserving the health result.
- Tutorial: run a DAG that retriggers itself builds a countdown with three runs.
- How-to guides cover service monitoring, alerts, Python and airflow-config, and stop boundaries.
- API reference defines results, branches, limits, callbacks, task models, and run configuration.
- Why health results and recovery actions are separate explains retries, retriggers, and workload ownership.
Published documentation is available at airflow-laminar.github.io/airflow-ha.
airflow-supervisor and airflow-nomad use this sensor inside external workload lifecycles and add log forwarding and diagnostics. airflow-pydantic provides declarative task models; airflow-config loads those models from YAML.
Apache 2.0. See LICENSE.