A tag edit should not erase a Data Product’s glossary terms

An OpenMetadata issue generated from code analysis reports a risky edit path: the Explore summary panel filters glossary labels out before passing a Data Product's tags to the editor, then saves the reduced list as a replacement. If the report is correct for a deployed version, adding one classification tag could remove existing glossary links without a specific warning.

Test the exact editor path

Use a non-production Data Product with one classification tag and one directly assigned glossary term. Record the tags through the API, change only the classification in the Explore right panel, then read the full tag list again. Compare it with the normal Data Product edit route. This matters because the issue concerns one UI path, not every tag update.

Protect the governance contract

OpenMetadata distinguishes classifications from glossary meaning. The Data Product API exposes the entity for independent read-back. If glossary labels disappear in a reproduction, pause bulk tag edits through that panel, preserve before-and-after payloads, and open a version-specific vendor case. Do not restore labels from memory; use a known snapshot.

Check downstream effects

Glossary terms support discovery and ownership language. Reconcile which Data Products changed during the affected period and whether consumers lost term-based navigation or policy context. An audit event that merely says tags changed may be too coarse for this task; inspect the removed labels explicitly.

Adovo’s view

Governance metadata is business data. A small UI edit needs the same read-after-write discipline as a pipeline change when it can rewrite a full set of labels. Make preservation of untouched terms a test, not an assumption.