mirror of
https://github.com/open-metadata/OpenMetadata
synced 2026-05-24 09:39:11 +00:00
- AssetDAO.update / AssetRepository.update: switch the UPDATE target from fqnHash to id. Two assets can share the same fullyQualifiedName (e.g. successive revisions of the same context file), so the old SQL could silently update sibling rows. - ContextFileExtractionService: run the extraction pipeline on a dedicated fixed thread pool instead of AsyncService.getExecutorService. process() blocks on assetService.read(...).join(), and S3/Azure reads are themselves scheduled on AsyncService — sharing the same bounded pool risks starving those reads (and deadlocking) once every thread is busy running extractions. - postgres/schemaChanges.sql: wrap the generated deleted column in COALESCE((json ->> 'deleted')::boolean, false) (and the asset_entity CAST variant) so an absent 'deleted' key is stored as FALSE, not NULL. Otherwise "non-deleted" filters based on the boolean index drop rows silently. Matches the MySQL IFNULL(..., FALSE) side of the migration. - ContextFileUploadSupport.sanitizeEntityName: treat null/blank input as 'file' instead of NPE-ing on replaceAll. Multipart uploads can arrive without filename metadata; the upload should still succeed with a stable generated name. |
||
|---|---|---|
| .. | ||
| migrations | ||
| schema | ||