xuanyuanking commented on a change in pull request #26040: [SPARK-9853][Core]
Optimize shuffle fetch of continuous partition IDs
URL: https://github.com/apache/spark/pull/26040#discussion_r332984232
##########
File path:
core/src/main/scala/org/apache/spark/shuffle/BlockStoreShuffleReader.scala
##########
@@ -41,6 +43,20 @@ private[spark] class BlockStoreShuffleReader[K, C](
private val dep = handle.dependency
+ private def fetchContinuousBlocksInBatch: Boolean = {
+ val conf = SparkEnv.get.conf
+ val compressed = conf.get(config.SHUFFLE_COMPRESS)
+ val featureEnabled =
conf.get(config.SHUFFLE_FETCH_CONTINUOUS_BLOCKS_IN_BATCH)
+ val serializerRelocatable =
dep.serializer.supportsRelocationOfSerializedObjects
+ // The batch fetching feature only works for reading consolidate file
written by
+ // SortShuffleWriter or UnsafeShuffleWriter.
+ val readConsolidateFile =
!handle.isInstanceOf[BypassMergeSortShuffleHandle[_, _]]
Review comment:
This flag added in 407b1e0 for fixing the Jenkins UT
`ReduceNumShufflePartitionsSuite` and `AdaptiveQueryExecSuite` failure. From my
first round debugging, I found the output written by
BypassMergeSortShuffleWriter can't read continuously as expected, let me check.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
