vanzin commented on a change in pull request #24554: [SPARK-27622][Core]
Avoiding the network when block manager fetches disk persisted RDD blocks from
the same host
URL: https://github.com/apache/spark/pull/24554#discussion_r288346457
##########
File path:
core/src/main/scala/org/apache/spark/storage/BlockManagerMasterEndpoint.scala
##########
@@ -522,7 +525,12 @@ class BlockManagerMasterEndpoint(
}
if (locations.nonEmpty && status.isDefined) {
- Some(BlockLocationsAndStatus(locations, status.get))
+ val localDirs = if (status.get.storageLevel.useDisk) {
+ locations.find(_.host ==
requesterHost).map(blockManagerInfo(_).localDirs)
Review comment:
Can it be that the RDD is only on disk in a different host (e.g. for RDDs
that have replication > 1)? Can that be detected here instead of triggering the
failure handling on the reader and falling back to the network path?
----------------------------------------------------------------
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]
