wangyum opened a new pull request, #17309: URL: https://github.com/apache/iceberg/pull/17309
## What `MergingSnapshotProducer.mergeDVs()` writes a Puffin file each time duplicate DVs are merged. When the delete manifest cache is invalidated (by a subsequent `addDeletes` call) or the commit fails, only the manifest files were deleted — the Puffin files were left orphaned on disk. ## Why After #15006, `mergeDVs()` writes a Puffin file (e.g. `merged-dvs-<snapshotId>-<n>.puffin`) and caches the resulting delete manifests. Two cleanup paths missed the Puffin files: 1. **Cache invalidation** (`newDeleteFilesAsManifests`): deleted the cached manifests but not the Puffin files they referenced. 2. **Commit failure** (`cleanUncommittedAppends`): deleted uncommitted delete manifests but not their Puffin files. The orphaned files can only be removed by `remove_orphan_files`. No data corruption — read correctness is preserved. ## How Track merged DV file locations in a `cachedMergedDVLocations` field (paralleling `cachedNewDeleteManifests`) and delete them in both cleanup paths. The location is resolved through the `FileIO` so its scheme matches `DeleteFile.location()` (e.g. stripping a `file:` URI scheme). ## Tests - `testMergedDVPuffinFileCleanedUpOnCacheInvalidation` — verifies the Puffin file is deleted when a subsequent `addDeletes` invalidates the cache, and the committed snapshot has a correctly merged DV. - `testMergedDVPuffinFileCleanedUpOnCommitFailure` — verifies the Puffin file is deleted when the commit fails (via injected `failCommits`). Closes https://github.com/apache/iceberg/issues/17307 --- **AI Disclosure** - Model: GLM 5.2 - Platform/Tool: opencode - Human Oversight: fully reviewed - Prompt Summary: Fix orphaned merged DV Puffin file on cache invalidation and commit failure, with tests covering both paths. ``` -- 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]
