Skip to content

ALL-DATA0006 · unhealthy_data_dependency

Count source references exposed to explicitly unhealthy data dependencies.

This is a deterministic rule for all languages. Read its implementation.

For each resolved source reference, read the health the provider recorded for every asset upstream of the one this code reads, and count each pair of a source reference and an unhealthy upstream asset once. Code that reads a healthy table fed by a broken one is producing wrong answers with no error anywhere, which is the failure mode a schema check cannot see and a freshness alert on the wrong asset will not raise.

Unknown health stays unknown. An asset nobody measured is not evidence of a problem, and treating it as one would make the count grow with the size of the lineage graph rather than with the number of real failures.

Each finding names the source reference and the upstream asset whose retained quality evidence failed. The value is the number of source reference and unhealthy upstream pairs.

An upstream asset marked healthy or unknown is never counted, so a lineage graph with no quality evidence at all reports nothing rather than everything. A reference whose own asset the catalog does not hold is skipped, since ALL-DATA0001 owns it and its upstream lineage is unknowable. One unhealthy asset feeding four source references is reported four times on purpose, because each of those four places is a separate reader producing a wrong answer.

A source reference whose upstream map records one asset unhealthy and another unknown returns 1. Two source references reading that same asset return 1 each. A reference whose upstream assets are all healthy or all unknown returns 0.

  • Cites “OpenLineage specification”, lineage model
  • Cites “DAMA-DMBOK”, data quality dimensions