Skip to content

[SPARK-46006][YARN] YarnAllocator miss clean targetNumExecutorsPerResourceProfileId after YarnSchedulerBackend call stop #43906

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
AngersZhuuuu wants to merge 2 commits into from
Closed

[SPARK-46006][YARN] YarnAllocator miss clean targetNumExecutorsPerResourceProfileId after YarnSchedulerBackend call stop #43906

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f3a0061
[SPARK-46006][YARN] YarnAllocator miss clean targetNumExecutorsPerRes…
AngersZhuuuu Nov 20, 2023
86f73b0
Update YarnAllocator.scala
AngersZhuuuu Nov 22, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 17 additions & 11 deletions resource-managers/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -384,19 +384,25 @@ private[yarn] class YarnAllocator(
this.numLocalityAwareTasksPerResourceProfileId = numLocalityAwareTasksPerResourceProfileId
this.hostToLocalTaskCountPerResourceProfileId = hostToLocalTaskCountPerResourceProfileId

val res = resourceProfileToTotalExecs.map { case (rp, numExecs) =>
createYarnResourceForResourceProfile(rp)
if (numExecs != getOrUpdateTargetNumExecutorsForRPId(rp.id)) {
logInfo(s"Driver requested a total number of $numExecs executor(s) " +
s"for resource profile id: ${rp.id}.")
targetNumExecutorsPerResourceProfileId(rp.id) = numExecs
allocatorNodeHealthTracker.setSchedulerExcludedNodes(excludedNodes)
true
} else {
false
if (resourceProfileToTotalExecs.isEmpty) {
targetNumExecutorsPerResourceProfileId.clear()
Copy link
Contributor

@zhouyifan279 zhouyifan279 Nov 27, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@AngersZhuuuu Will it be better if we set values in targetNumExecutorsPerResourceProfileId to 0 instead of clearing targetNumExecutorsPerResourceProfileId.
ContainerRequests can be cancelled in YarnAllocator#updateResourceRequests if we set values to 0.

Copy link
Contributor Author

@AngersZhuuuu AngersZhuuuu Nov 27, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@AngersZhuuuu Will it be better if we set values in targetNumExecutorsPerResourceProfileId to 0 instead of clearing targetNumExecutorsPerResourceProfileId. ContainerRequests can be cancelled in YarnAllocator#updateResourceRequests if we set values to 0.

Nice catch, will make a follower up pr later

allocatorNodeHealthTracker.setSchedulerExcludedNodes(excludedNodes)
true
} else {
val res = resourceProfileToTotalExecs.map { case (rp, numExecs) =>
createYarnResourceForResourceProfile(rp)
if (numExecs != getOrUpdateTargetNumExecutorsForRPId(rp.id)) {
logInfo(s"Driver requested a total number of $numExecs executor(s) " +
s"for resource profile id: ${rp.id}.")
targetNumExecutorsPerResourceProfileId(rp.id) = numExecs
allocatorNodeHealthTracker.setSchedulerExcludedNodes(excludedNodes)
true
} else {
false
}
}
res.exists(_ == true)
}
res.exists(_ == true)
}

/**
Expand Down