redsanket commented on a change in pull request #23677: [SPARK-26755] :
Optimize Spark Scheduler to dequeue speculative tasks…
URL: https://github.com/apache/spark/pull/23677#discussion_r251872181
##########
File path: core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala
##########
@@ -245,6 +257,28 @@ private[spark] class TaskSetManager(
allPendingTasks += index // No point scanning this whole list to find the
old task there
}
+ private[spark] def addPendingSpeculativeTask(index: Int) {
+ for (loc <- tasks(index).preferredLocations) {
+ loc match {
+ case e: ExecutorCacheTaskLocation =>
+ pendingSpeculatableTasksForExecutor.getOrElseUpdate(
+ e.executorId, new ArrayBuffer) += index
+ case _ =>
+ }
+ pendingSpeculatableTasksForHost.getOrElseUpdate(loc.host, new
ArrayBuffer) += index
+ for (rack <- sched.getRackForHost(loc.host)) {
+ pendingSpeculatableTasksForRack.getOrElseUpdate(rack, new ArrayBuffer)
+= index
+ }
+ }
+
+ if (tasks(index).preferredLocations == Nil) {
+ pendingSpeculatableTasksWithNoPrefs += index
+ }
+
+ // No point scanning this whole list to find the old task there
+ allPendingSpeculatableTasks += index
Review comment:
Also in your tests would be interesting to note what is locality affinity...
do you see more process local tasks vs node local etc., that might be related
to scheduling logic, just curious to know about it.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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]
