agrawaldevesh commented on a change in pull request #29468:
URL: https://github.com/apache/spark/pull/29468#discussion_r472366419
##########
File path:
core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala
##########
@@ -1062,25 +1062,36 @@ private[spark] class TaskSchedulerImpl(
}
def getExecutorsAliveOnHost(host: String): Option[Set[String]] =
synchronized {
- hostToExecutors.get(host).map(_.toSet)
+
hostToExecutors.get(host).map(_.filterNot(isExecutorDecommissioned)).map(_.toSet)
}
def hasExecutorsAliveOnHost(host: String): Boolean = synchronized {
- hostToExecutors.contains(host)
+ hostToExecutors.get(host)
+ .exists(executors => executors.exists(e =>
!isExecutorDecommissioned(e)))
}
def hasHostAliveOnRack(rack: String): Boolean = synchronized {
- hostsByRack.contains(rack)
+ hostsByRack.get(rack)
Review comment:
I haven't looked at this code too deeply, but does it make sense it
eagerly remove the decommissioned executor from hostToExecutors and
recomputeLocality and thus rebuild hostsByRack etc ?
----------------------------------------------------------------
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
