JingsongLi commented on PR #9207: URL: https://github.com/apache/paimon/pull/9207#issuecomment-5748062718
I reviewed the current head with the end-to-end requirement in mind. **Requirement fit: SUPPORTED, but the current delivery needs to PIVOT on scope.** The underlying problem is real: managed BLOB packs are shared, so ordinary snapshot/file expiration cannot safely reclaim them, and a reachability-based orphan cleanup has user-visible storage value. The current PR is still roughly 6.5k added lines across core, Flink, Spark, procedures, and tests. The earlier request to split it has only been acknowledged; the code has not yet been split. That makes it impractical to validate the safety boundary (live-reference collection, grace-period handling, deletion, and distributed execution) as one review unit. Please split this into independently reviewable changes, for example: 1. core reachability model plus local cleanup and focused failure/recovery tests; 2. distributed execution/integration for Flink and Spark; 3. procedures, docs, and engine-facing wiring. Each part should remain buildable and should demonstrate that a blob referenced by any retained snapshot/tag/branch is never deleted, while an actually unreachable blob is eventually removed. I would not merge the current monolithic form, but I would continue reviewing the split series because the end-to-end value is clear. -- 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]
