Pipeline issue finder uses three Orbit graph queries to trace CI/CD failure propagation across GitLab projects. An aggregation query finds failure counts per project in a sliding time window. A traversal query then identifies which failed job names are shared across multiple projects simultaneously — the blast-radius candidates. A path-finding query traces upstream trigger chains to find the root pipeline that started the cascade. A lightweight Python correlator service polls these three queries every 5 minutes, scores candidates by affected project count × failure rate, and fires a Slack alert with the top hit. The output is a cross-project heat map showing failure rate by project and day, plus a ranked blast-radius list of shared templates causing the most damage — something GitLab's native per-project pipeline view can't show without the graph layer Orbit provides.

Built With

Share this project:

Updates