JingsongLi opened a new pull request, #8584: URL: https://github.com/apache/paimon/pull/8584
## What changed - Build primary-key vector ANN segments asynchronously on a dedicated executor and honor `prepareCommit(waitCompaction)`. - Derive LSM-like ANN levels from active segments and rebuild high-fanout or stale segments without persisting level metadata. - Search stale-source segments with active row filtering and apply exact-search fallback according to `GlobalIndexSearchMode`. - Add independent ANN compaction target-size, level-fanout, and stale-ratio options with validation and documentation. ## Why Synchronous per-compaction ANN construction increases commit latency, while accumulating many small or stale ANN segments increases query fan-out. The new maintenance path decouples ANN work from data compaction while preserving restart correctness and snapshot-safe search coverage. ## Validation - 43 focused primary-key vector tests passed in fast mode. - The same 43 tests passed with Spotless, Checkstyle, and Enforcer enabled. - Generated core configuration documentation was rebuilt successfully. - `git diff --check` passed. -- 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]
