otterc commented on code in PR #36305:
URL: https://github.com/apache/spark/pull/36305#discussion_r861264663
##########
core/src/main/scala/org/apache/spark/storage/ShuffleBlockFetcherIterator.scala:
##########
@@ -1166,6 +1166,8 @@ final class ShuffleBlockFetcherIterator(
case ShuffleBlockBatchId(shuffleId, mapId, startReduceId, _) =>
throw SparkCoreErrors.fetchFailedError(address, shuffleId, mapId,
mapIndex, startReduceId,
msg, e)
+ case ShuffleBlockChunkId(shuffleId, _, reduceId, _) =>
+ SparkCoreErrors.fetchFailedError(address, shuffleId, -1L, -1,
reduceId, msg, e)
Review Comment:
This will propagate the `FetchFailure` with mapId/mapIndex = -1. Currently
the driver will not unregister the mergeStatus because mapIndex = -1. Since the
mergeStatus is still registered, the next attempt will still try to fetch the
corrupted merged block. Eventually the application will fail.
So, this requires a change in the driver as well. Also once we make that
change, we should add a UT to `DAGScheduler` to verify that.
--
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]
