Github user lianhuiwang commented on a diff in the pull request:
https://github.com/apache/spark/pull/4168#discussion_r23502082
--- Diff:
core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala ---
@@ -237,39 +257,18 @@ private[spark] class ExecutorAllocationManager(
return 0
}
- // The number of executors needed to satisfy all pending tasks is the
number of tasks pending
- // divided by the number of tasks each executor can fit, rounded up.
- val maxNumExecutorsPending =
- (listener.totalPendingTasks() + tasksPerExecutor - 1) /
tasksPerExecutor
- if (numExecutorsPending >= maxNumExecutorsPending) {
- logDebug(s"Not adding executors because there are already
$numExecutorsPending " +
- s"pending and pending tasks could only fill
$maxNumExecutorsPending")
- numExecutorsToAdd = 1
- return 0
- }
-
- // It's never useful to request more executors than could satisfy all
the pending tasks, so
- // cap request at that amount.
- // Also cap request with respect to the configured upper bound.
- val maxNumExecutorsToAdd = math.min(
- maxNumExecutorsPending - numExecutorsPending,
- maxNumExecutors - numExistingExecutors)
- assert(maxNumExecutorsToAdd > 0)
-
- val actualNumExecutorsToAdd = math.min(numExecutorsToAdd,
maxNumExecutorsToAdd)
-
- val newTotalExecutors = numExistingExecutors + actualNumExecutorsToAdd
- val addRequestAcknowledged = testing ||
client.requestExecutors(actualNumExecutorsToAdd)
+ val newTotalExecutors = math.min(numExistingExecutors +
numExecutorsToAdd,
--- End diff --
i think there we should make numExistingExecutors exclude
executorsPendingToRemove.size?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
