codope opened a new pull request, #19464:
URL: https://github.com/apache/hudi/pull/19464
### Describe the issue this Pull Request addresses
Follow-up to #18988. `completeClustering` fires the write commit callback
with
`clusteringInstant.getAction()`, but that is the **inflight** instant,
captured before the
transition and immutable. On timeline v2 its action is `clustering`, while
`transitionClusterInflightToComplete` records the completed instant as
`replacecommit`. So on table
version 8+ the callback reports an action that does not exist on the
completed timeline. On the 0.x
layout the two coincide, which is why it went unnoticed.
### Summary and Changelog
Report the action the commit lands under, not the one it was scheduled under.
- `ClusteringUtils.transitionClusteringOrReplaceInflightToComplete` now
returns the completed
`HoodieInstant`. Both delegates already returned it; the util discarded it.
- `BaseHoodieTableServiceClient#completeClustering` and
`HoodieFlinkTableServiceClient#completeClustering` pass that instant's
action to the callback.
This keeps the completed action defined only in the timeline
implementation, so it cannot drift.
The sibling call sites are already correct: compaction fires
`COMMIT_ACTION`, log compaction
`DELTA_COMMIT_ACTION`.
-
`TestHoodieJavaClientOnMergeOnReadStorage#testWriteCommitCallbackFiresOnClustering`
asserted
`REPLACE_COMMIT_ACTION || CLUSTERING_ACTION`, which accepted the wrong
value and could not catch
this; tightened to `assertEquals`, mirroring the compaction test.
-
`TestHoodieFlinkTableServiceClient#testCompleteClusteringCommitsAndCleansMarkers`
statically mocks
`ClusteringUtils`, so the transition is stubbed to return a completed
instant, and the test now
also asserts the Flink path reports `replacecommit` — that fire previously
had no coverage.
### Impact
Consumers of `HoodieWriteCommitCallback` on table version 8+ see
`replacecommit` instead of
`clustering` for clustering commits — the value the timeline records, and
the value already reported
on older table versions.
`transitionClusteringOrReplaceInflightToComplete` changes from `void` to
returning `HoodieInstant`.
It is not annotated public API and has two callers, both in this PR, so this
is source-compatible
in-tree; a pre-compiled out-of-tree caller would need recompiling.
### Risk Level
low — no change to when or whether the callback fires, or to what is written
to the timeline.
### Documentation Update
none
### Contributor's checklist
- [x] Read through [contributor's
guide](https://hudi.apache.org/contribute/how-to-contribute)
- [x] Enough context is provided in the sections above
- [x] Adequate tests were added if applicable
--
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]
