r7raul1984 opened a new pull request, #13028:
URL: https://github.com/apache/gluten/pull/13028
## What changes are proposed in this pull request?
Fixes #13027
`CelebornUtils.GET_REDUCER_FILE_GROUP_RESPONSE_NAME` used `.` as the inner
class
separator instead of `$`, which is required by `Class.forName()` for inner
classes.
When
`spark.celeborn.client.spark.shuffle.getReducerFileGroup.broadcast.enabled=true`
(Celeborn ≥ 0.6.0), the wrong separator caused:
1. `Class.forName()` → `ClassNotFoundException` → fallback to
`Object.class`
2. `SparkUtils.getMethod("serializeGetReducerFileGroupResponse",
Integer.class, Object.class)` → `NoSuchMethodException`
3. Wrapped as `RuntimeException`, preventing file group broadcast
4. All executors fail with `FetchFailedException` → job aborted
One-character fix: `.GetReducerFileGroupResponse` →
`$GetReducerFileGroupResponse`
Jobs without this flag enabled are not affected (the code path is guarded
by
an outer `if` on `getReducerFileGroupBroadcastEnabled()`).
## How was this patch tested?
Verified the correct class name resolves via `Class.forName()` against
`celeborn-client-spark-3-shaded` 0.6.1 and 0.6.3.
Confirmed the method signature of
`SparkUtils.serializeGetReducerFileGroupResponse`
takes `ControlMessages$GetReducerFileGroupResponse` as the second
parameter (not `Object`).
## Was this patch authored or co-authored using generative AI tooling?
Generated-by: Claude Sonnet 4.6
--
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]
