cloud-fan commented on code in PR #36238:
URL: https://github.com/apache/spark/pull/36238#discussion_r853768026
##########
core/src/main/scala/org/apache/spark/executor/Executor.scala:
##########
@@ -264,16 +290,26 @@ private[spark] class Executor(
decommissioned = true
}
+ private[executor] def createTaskRunner(context: ExecutorBackend,
+ taskDescription: TaskDescription) = new TaskRunner(context,
taskDescription, plugins)
+
def launchTask(context: ExecutorBackend, taskDescription: TaskDescription):
Unit = {
- val tr = new TaskRunner(context, taskDescription, plugins)
- runningTasks.put(taskDescription.taskId, tr)
+ val taskId = taskDescription.taskId
+ val tr = createTaskRunner(context, taskDescription)
+ runningTasks.put(taskId, tr)
+ val killMark = killMarks.get(taskId)
+ if (killMark != null) {
+ tr.kill(killMark._1, killMark._2)
+ killMarks.remove(taskId)
+ }
threadPool.execute(tr)
Review Comment:
Ideally we should kill the task as early as possible, but in this case, I
think it's safer to reuse the existing code path and launch an already killed
task, instead of adding a new code path to skip task launching, which may miss
some bookkeeping work or event triggering.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
