Ngone51 commented on a change in pull request #27773: [SPARK-29154][CORE]
Update Spark scheduler for stage level scheduling
URL: https://github.com/apache/spark/pull/27773#discussion_r390064076
##########
File path:
core/src/main/scala/org/apache/spark/scheduler/cluster/CoarseGrainedSchedulerBackend.scala
##########
@@ -606,10 +608,10 @@ class CoarseGrainedSchedulerBackend(scheduler:
TaskSchedulerImpl, val rpcEnv: Rp
}
- override def maxNumConcurrentTasks(): Int = synchronized {
- executorDataMap.values.map { executor =>
- executor.totalCores / scheduler.CPUS_PER_TASK
- }.sum
+ override def maxNumConcurrentTasks(rp: ResourceProfile): Int = synchronized {
+ val cpusPerTask = rp.getTaskCpus.getOrElse(scheduler.CPUS_PER_TASK)
+ val executorsWithResourceProfile =
executorDataMap.values.filter(_.resourceProfileId == rp.id)
Review comment:
Let's say user has a barrier job like:
```
rdd.withResources(rp).barrirer().mapPartition { part =>
// do some barrier stuff
}.collect()
```
So, at the time we're
`createShuffleMapStage`/`mergeResourceProfilesForStage`/`checkBarrierStageWithNumSlots`,
it's possible that some executors haven't launched?
----------------------------------------------------------------
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]
