Ngone51 commented on a change in pull request #23871:
[SPARK-23433][SPARK-25250] [CORE] Later created TaskSet should learn about the
finished partitions
URL: https://github.com/apache/spark/pull/23871#discussion_r259582843
##########
File path:
core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala
##########
@@ -208,7 +211,7 @@ private[spark] class TaskSchedulerImpl(
taskSetsByStageIdAndAttempt.getOrElseUpdate(stage, new HashMap[Int,
TaskSetManager])
stageTaskSets(taskSet.stageAttemptId) = manager
val conflictingTaskSet = stageTaskSets.exists { case (_, ts) =>
- ts.taskSet != taskSet && !ts.isZombie
+ ts.taskSet != manager.taskSet && !ts.isZombie
Review comment:
Previously, I filter the already completed tasks in `createTaskSetManager()`
and creates a new TaskSet by filtered tasks. So, the submitted TaskSet form
DAGScheduler is not the same with TaskSet in TaskSetManager. But it failed
plenty of unit tests since those tests would keep the original submitted
TaskSet for some operations.
And now I have moved the `filter` logic into the TaskSetManager.
----------------------------------------------------------------
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]
