gcp-managed-airflow-recommendations

Run the workload_restarts.py script with your project ID, location, environment name, and query duration to retrieve restart counts per component. High scheduler or worker restart counts typically indicate memory issues or unoptimized DAGs blocking the event loop.

Use the environment_health.py script, which queries Cloud Monitoring for environment, database, and webserver health percentages over a specified timeframe. It requires a minimum query duration of one hour and outputs results as JSON.

Check CPU usage, memory consumption, and disk usage per workload component using the provided monitoring scripts before resizing. Recommendations should be evidence-based rather than blindly increasing machine sizes without confirming actual resource saturation.

Yes, but recommendations must distinguish between generations since their architectures differ. The skill instructs the agent to verify the environment generation via gcloud configuration inspection before suggesting scaling or configuration changes.

High parsing times usually come from heavy top-level code in DAG files, such as database connections or API calls outside task definitions. The dag_parsing_stats.py script measures total parse time, parse errors, and processor timeouts to confirm the bottleneck.

The scripts authenticate via gcloud application-default credentials and call the Cloud Monitoring PromQL API. Your account needs read access to Cloud Monitoring metrics and permission to describe the Composer environment in the target project.