Github user vanzin commented on a diff in the pull request:
https://github.com/apache/spark/pull/20604#discussion_r169420544
--- Diff:
core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala ---
@@ -455,7 +461,12 @@ private[spark] class ExecutorAllocationManager(
val executorsRemoved = if (testing) {
executorIdsToBeRemoved
} else {
- client.killExecutors(executorIdsToBeRemoved)
+ // We don't want to change our target number of executors, because
we already did that
+ // when the task backlog decreased. Normally there wouldn't be any
tasks running on these
+ // executors, but maybe the scheduler *just* decided to run a task
there -- in that case,
--- End diff --
Also don't follow this part. If `force = false`, doesn't that mean the
executor won't be killed if tasks are running on it? So wouldn't
`countFailures` be meaningless in that context?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
