attilapiros 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_r348521651
##########
File path: core/src/main/scala/org/apache/spark/storage/BlockManager.scala
##########
@@ -113,6 +116,47 @@ private[spark] class ByteBufferBlockData(
}
+private[spark] class HostLocalDirManager(
+ futureExecutionContext: ExecutionContext,
+ cacheSize: Int,
+ externalBlockStoreClient: ExternalBlockStoreClient,
+ host: String,
+ externalShuffleServicePort: Int) extends Logging {
+
+ private val executorIdToLocalDirsCache =
+ CacheBuilder
+ .newBuilder()
+ .maximumSize(cacheSize)
+ .build[String, Array[String]]()
+
+ private[spark] def getCachedHostLocalDirs()
+ : scala.collection.Map[String, Array[String]] =
executorIdToLocalDirsCache.synchronized {
+ import scala.collection.JavaConverters._
+ return executorIdToLocalDirsCache.asMap().asScala
+ }
+
+ private[spark] def getHostLocalDirs(
+ executorIds: Array[String],
+ callback: FutureCallback[java.util.Map[String, Array[String]]]): Unit = {
Review comment:
I see. It is not used across modules but as you suggested I will change the
callback type to `Try[java.util.Map[String, Array[String]] => Unit`.
----------------------------------------------------------------
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]
