mchades commented on PR #13771: URL: https://github.com/apache/gravitino/pull/13771#issuecomment-6139619591
Create-only writes and conflict reload are a reasonable safeguard against identity reassignment. The minimum remaining work for this PR is: - Preserve valid Hive-backed Iceberg case aliases and existing registration IDs. Restore the uppercase SQL success test; changing `%S` to `%s` only removes the regression coverage. - Define a safe recovery path for shared sources. Removing `gravitino.identifier` from a shared HMS changes other catalogs’ identity checks. Unregistering Gravitino metadata must be distinct from dropping the physical object. - Cover aliases, HA imports and re-registration with relational-store tests, verifying that privileges/tags remain attached to the correct registration. Document that topics remain outside this fix. These failures are engine-visible: [Spark](https://github.com/apache/gravitino/blob/03272ce21e12532074aa5e5b06f968547d2d72fe/spark-connector/spark-common/src/main/java/org/apache/gravitino/spark/connector/catalog/BaseCatalog.java#L272) loads Gravitino metadata before reads/writes, [Trino’s `getTableHandle`](https://github.com/apache/gravitino/blob/03272ce21e12532074aa5e5b06f968547d2d72fe/trino-connector/trino-connector/src/main/java/org/apache/gravitino/trino/connector/GravitinoMetadata.java#L145) calls `tableExists`, and [Flink’s `getTable/tableExists`](https://github.com/apache/gravitino/blob/03272ce21e12532074aa5e5b06f968547d2d72fe/flink-connector/flink-common/src/main/java/org/apache/gravitino/flink/connector/catalog/BaseCatalog.java#L276) propagate failures. Valid operations can therefore fail during metadata resolution or existence checks, even when listing succeeds. Keep identity conflicts explicit, not `exists=false`, and add connector regression coverage, including Sp ark `CREATE TABLE IF NOT EXISTS`. Separating external object identity from catalog-scoped registration identity can remain a follow-up design; it need not expand this bounded fix. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
