Joy-2000 opened a new issue, #19793: URL: https://github.com/apache/hudi/issues/19793
### Bug Description **What happened:** Under Non-Blocking Concurrency Control (NBCC), performing a `delete_partition` on a partition and then re-inserting rows into that same partition results in **silent data loss** — the re-inserted rows are written to storage but become invisible to all readers, with no error raised. This is the same underlying invariant violation as [#19791](https://github.com/apache/hudi/pull/19791) (`insert overwrite` under NBCC), but reached through a different operation pair. #19791 deliberately scoped its fix to `insert overwrite` only; this issue tracks the broader root cause. **What you expected:** When a bucket's fixed file group is retired by a replacecommit, subsequent NBCC writes to that bucket should land on a **fresh** file id (e.g. bump the trailing `-0` generation in `newBucketFileIdForNBCC` / the bucket-id-getter) so the new file group id no longer matches the retired one. This fixes both the `insert overwrite` and `delete_partition`+re-insert cases at the source. Needs careful thought on compaction, log-file grouping, and reader-side bucket resolution. **Steps to reproduce:** 1. Create a MOR + simple-bucket table with `hoodie.write.concurrency.mode = NON_BLOCKING_CONCURRENCY_CONTROL`. 2. Insert some rows into partition `p`. 3. Run `delete_partition` on `p`. This produces a **replacecommit** that records bucket `p`'s fixed file id (`…-0`) as replaced. 4. Re-insert rows into partition `p`. NBCC routes them to the **same** fixed file id. 5. Query the table. **Expected:** the rows from step 4 are visible. **Actual:** the rows from step 4 are silently invisible — the file system view still treats that file group id as replaced by the step-3 replacecommit, so it filters the freshly written slice out. No error is raised. ### Environment **Hudi version:** **Query engine:** (Spark/Flink/Trino etc) **Relevant configs:** ### Logs and Stack Trace _No response_ -- 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]
