wenzhenghu commented on issue #65779: URL: https://github.com/apache/doris/issues/65779#issuecomment-5057540787
Additional follow-up from the PR #65126 review: `MetaCacheEntry.invalidateIf()` currently calls `bumpAllGenerations()` before evaluating the selective invalidation predicate. A narrow invalidation for table A can therefore fence an overlapping in-flight load for unrelated table B, causing B's result to be returned without being cached and potentially forcing a later remote reload. This does not affect the correctness of the returned metadata, and it requires the targeted invalidation to overlap an unrelated in-flight load, so the practical impact is limited. Fixing it safely is non-trivial because selective invalidation must fence matching cached and in-flight keys without weakening stale-load publication protection. Track this together with the other external metadata cache performance follow-ups. The eventual fix should retain the global generation bump for `invalidateAll()` and add an A-invalidation/B-load concurrency test for `invalidateIf()`. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
