holdenk commented on a change in pull request #28708:
URL: https://github.com/apache/spark/pull/28708#discussion_r447247000
##########
File path:
core/src/main/scala/org/apache/spark/shuffle/IndexShuffleBlockResolver.scala
##########
@@ -55,6 +58,25 @@ private[spark] class IndexShuffleBlockResolver(
def getDataFile(shuffleId: Int, mapId: Long): File = getDataFile(shuffleId,
mapId, None)
+ /**
+ * Get the shuffle files that are stored locally. Used for block migrations.
+ */
+ override def getStoredShuffles(): Set[ShuffleBlockInfo] = {
+ // Matches ShuffleIndexBlockId name
+ val pattern = "shuffle_(\\d+)_(\\d+)_.+\\.index".r
+ val rootDirs = blockManager.diskBlockManager.localDirs
+ // ExecutorDiskUtil puts things inside one level hashed sub directories
+ val searchDirs = rootDirs.flatMap(_.listFiles()).filter(_.isDirectory())
++ rootDirs
+ val filenames = searchDirs.flatMap(_.list())
+ logDebug(s"Got block files ${filenames.toList}")
+ filenames.flatMap { fname =>
+ pattern.findAllIn(fname).matchData.map {
+ matched => ShuffleBlockInfo(matched.group(1).toInt,
matched.group(2).toLong)
+ }
+ }.toSet
Review comment:
There shouldn't be any duplicates, but if there are we only need to
transfer one anyways.
----------------------------------------------------------------
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
