query-dag-failure-debugger

To investigate Airflow DAG failures, paste an Airflow task error or log snippet to initialize a triage. The Skill classifies the failure into categories like SPARK_RESOURCE or SQL_FAILURE using DAG code, task logs, and metadata to provide root cause evidence and next steps.

Airflow DAG failures on Spark and Trino are categorized into 9 buckets including SPARK_RESOURCE, SPARK_RUNTIME, SQL_FAILURE, and QUERY_ENGINE_TRANSIENT. The Skill evaluates task logs and DAG code to identify the specific root cause bucket with confidence levels.

Yes, you can investigate Airflow task logs using Coralogix and Grafana. The Skill pulls live evidence from Coralogix, Friday MCPs for AWS and Kubernetes, and Grafana to inform Phase 2 live investigation and root-cause reasoning.

To classify Airflow task errors, the Skill performs fast Phase 1 classification using DAG code and task logs to identify the root bucket. It outputs a bucket classification, confidence score, root cause reasoning, and concrete next steps for resolution.

Yes, Airflow DAG triage works with Spark EMR and Presto workloads. The Skill is specifically designed for Airflow-backed DAGs spanning Spark/EMR and Trino/Presto, evaluating failures across orchestration, Spark runtime, and SQL query engine backends.

Your Airflow DAG may return an UNKNOWN bucket classification when task logs and DAG code lack sufficient evidence to categorize the failure into defined buckets like SPARK_RESOURCE or PERMISSION_ACCESS. The Skill outputs this classification with low confidence and suggests escalation to Phase 2 live investigation.