vanzin commented on a change in pull request #25299: [SPARK-27651][Core] Avoid
the network when shuffle blocks are fetched from the same host
URL: https://github.com/apache/spark/pull/25299#discussion_r308939960
##########
File path:
core/src/main/scala/org/apache/spark/storage/BlockManagerMasterEndpoint.scala
##########
@@ -514,15 +527,16 @@ class BlockManagerMasterEndpoint(
if (locations.nonEmpty && status.isDefined) {
val localDirs = locations.find { loc =>
- if (loc.port != externalShuffleServicePort && loc.host ==
requesterHost) {
+ // When the external shuffle service running on the same host is found
among the block
+ // locations then the block must be persisted on the disk. In this
case the executorId
+ // can be used to access this block even when the original executor is
already stopped.
+ loc.host == requesterHost &&
+ (loc.port == externalShuffleServicePort ||
blockManagerInfo
.get(loc)
.flatMap(_.getStatus(blockId).map(_.storageLevel.useDisk))
- .getOrElse(false)
- } else {
- false
- }
- }.map(blockManagerInfo(_).localDirs)
+ .getOrElse(false))
+ }.map(bmId => executorIdLocalDirs(bmId.executorId))
Review comment:
`.map { bmId => ... }`
----------------------------------------------------------------
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]
