Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/17533#discussion_r110434821
--- Diff:
core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala ---
@@ -512,6 +521,51 @@ private[spark] class TaskSetManager(
}
}
+ private[this] def sortPendingTasks(): Unit = {
+ val taskIndexs = (0 until numTasks).toArray
+ def ordFunc(x: Int, y: Int): Boolean = {
+ getTaskInputSizeFromShuffledRDD(tasks(x)) <
getTaskInputSizeFromShuffledRDD(tasks(y))
+ }
+ if (tasks.nonEmpty) {
+ // Sort the tasks based on their input size from ShuffledRDD.
+ pendingTasksForExecutor.foreach {
+ case (k, v) => pendingTasksForExecutor(k) = v.sortWith(ordFunc)
+ }
+ pendingTasksForHost.foreach {
+ case (k, v) => pendingTasksForHost(k) = v.sortWith(ordFunc)
+ }
+ pendingTasksForRack.foreach {
+ case (k, v) => pendingTasksForRack(k) = v.sortWith(ordFunc)
+ }
+ pendingTasksWithNoPrefs = pendingTasksWithNoPrefs.sortWith(ordFunc)
+ allPendingTasks = allPendingTasks.sortWith(ordFunc)
+ }
+ }
+
+ // Visible for testing
+ private[spark] def setTaskInputSizeFromShuffledRDD(inputSize:
Map[Task[_], Long]) = {
+ taskInputSizeFromShuffledRDD.clear()
+ taskInputSizeFromShuffledRDD ++= inputSize
+ }
+
+ private[this] def getTaskInputSizeFromShuffledRDD(task: Task[_]): Long =
{
+ taskInputSizeFromShuffledRDD.get(task) match {
+ case Some(size) => size
+ case None =>
+ val size =
+ sched.dagScheduler.parentSplitsInShuffledRDD(task.stageId,
task.partitionId).map {
--- End diff --
I think the major complaint is this call, we don't want the TSM requesting
info from the DAGSCheduler. But you could change that -- instead the
DAGScheduler could *push* this info into the TSM after it has the input sizes.
This actually might not be that bad, since in any case the DAGScheduler has to
know this info when it calls `submitMissingTasks`. I don't think this info
should change at all after the taskset has been submitted, right, so you'd have
it all available at construction time?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
