Recognises undetected drift between database and model
Nothing that compares the deployed catalogue with the current generated DDL. Hotfix indexes, emergency nullability changes and a column someone added under incident pressure accumulate in production, and the model stops describing the thing people actually query — while continuing to look right, which is what makes it dangerous.
Can AI do this?
Not tested yet
No AI benchmark covers this knobit yet. New results are announced in the Can AI Do the Job? newsletter.
What it looks like in practice
Expected result: The missing comparison with the generated DDL
How it is assessed: Flags the deficient case and passes the acceptable one. Deficient: ix_fact_payment_status_code added in production during an incident eighteen months ago. It exists in no environment's DDL and in no model, and the next rebuild of the mart is slow for reasons nobody can reconstruct Acceptable: A scheduled comparison of the deployed catalogue (information_schema, or the engine's own schema-compare tooling) with the generated DDL, each difference closed deliberately — into the model, or out of the database — and the comparison's result part of the stage-8 evidence
Also known as: Drift between the database and the model never detected (generated DDL)
Related knobits
Builds on
- K-859 Explains testing DDL by execution and detecting drift
- K-914 Explains forward and reverse engineering of data models
All 33 knobits of Generate and review DDL.
Source: Review checklist 08-ddl, row 26; BCBS 239; Humble & Farley 2010