LuciferYang opened a new issue, #10233: URL: https://github.com/apache/paimon/issues/10233
### Search before asking - [X] I searched in the [issues](https://github.com/apache/paimon/issues) and found nothing similar. ### Paimon version master ### Compute Engine Flink (batch write) ### Minimal reproduce step `TagManager.deleteAllTagsOfOneSnapshot` deletes the tag files directly and never calls `TagCallback.notifyDeletion`. The sibling `deleteTag` used by the same expiry loop in `TagBatchCreation.expireTag` does notify callbacks. The loop routes every over-budget snapshot through `deleteAllTagsOfOneSnapshot` and only the boundary snapshot through `deleteTag`, so most tags expired by batch writes never reach the callbacks. Reproduce: a table with `tag.num-retained-max` set and one of the tag callbacks configured (`metadata.iceberg.storage`, `tag.create-success-file`, or `metastore.tag-to-partition`), written in Flink batch mode so `BatchWriteGeneratorTagOperator` drives `TagBatchCreation`. As tags accumulate past the retained max, the callback side effects for bulk-expired tags are never applied. ### What doesn't meet your expectations? Callbacks should fire for every deleted tag. With `metadata.iceberg.storage` enabled, `IcebergCommitCallback.notifyDeletion` removes the tag from the Iceberg metadata `refs`; skipping it leaves stale refs for tags that no longer exist in Paimon. The same skip orphans `{tagName}_SUCCESS` files and leaves stale metastore partitions when those features are on. ### Anything else? The batch-expiry path (`deleteAllTagsOfOneSnapshot`) is the common one, not a rare corner. Fix routes it through the same delete-and-notify helper the single-tag path uses. ### Are you willing to submit a PR? - [X] I'm willing to submit a PR! -- 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]
