Skip to content

[SPARK-20564][Deploy] Reduce massive executor failures when executor count is large (>2000) #17854

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
wants to merge 1 commit into from
Closed

[SPARK-20564][Deploy] Reduce massive executor failures when executor count is large (>2000) #17854

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
38 changes: 31 additions & 7 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 @@ -257,7 +257,9 @@ private[yarn] class YarnAllocator(
* This must be synchronized because variables read in this method are mutated by other methods.
*/
def allocateResources(): Unit = synchronized {
updateResourceRequests()
val launchContainerCount =
sparkConf.getInt("spark.yarn.launchContainer.count.simultaneously", -1)
updateResourceRequests(launchContainerCount)

val progressIndicator = 0.1f
// Poll the ResourceManager. This doubles as a heartbeat if there are no pending container
Expand Down Expand Up @@ -292,9 +294,16 @@ private[yarn] class YarnAllocator(
* Visible for testing.
*/
def updateResourceRequests(): Unit = {
updateResourceRequests(-1)
}

def updateResourceRequests(maxCount : Int): Unit = {
val pendingAllocate = getPendingAllocate
val numPendingAllocate = pendingAllocate.size
val missing = targetNumExecutors - numPendingAllocate - numExecutorsRunning
var missing = targetNumExecutors - numPendingAllocate - numExecutorsRunning
if (maxCount > 0) {
missing = math.min(missing, maxCount)
}

if (missing > 0) {
logInfo(s"Will request $missing executor container(s), each with " +
Expand Down Expand Up @@ -431,18 +440,33 @@ private[yarn] class YarnAllocator(
remainingAfterOffRackMatches)
}

if (!remainingAfterOffRackMatches.isEmpty) {
logDebug(s"Releasing ${remainingAfterOffRackMatches.size} unneeded containers that were " +
var launchContainerCount =
sparkConf.getInt("spark.yarn.launchContainer.count.simultaneously", -1)
val containerNumToLaunch = {
if (launchContainerCount <= 0) {
containersToUse.size
} else {
math.min(launchContainerCount, containersToUse.size)
}
}
val filteredContainersToUse = containersToUse.drop(containersToUse.size - containerNumToLaunch)
val remainingAfterFilter
= containersToUse -- filteredContainersToUse ++ remainingAfterOffRackMatches

if (!remainingAfterFilter.isEmpty) {
logDebug(s"Releasing ${remainingAfterFilter.size} unneeded containers that were " +
s"allocated to us")
for (container <- remainingAfterOffRackMatches) {
for (container <- remainingAfterFilter) {
internalReleaseContainer(container)
}
}

runAllocatedContainers(containersToUse)
if (filteredContainersToUse.length > 0) {
runAllocatedContainers(filteredContainersToUse)
}

logInfo("Received %d containers from YARN, launching executors on %d of them."
.format(allocatedContainers.size, containersToUse.size))
.format(allocatedContainers.size, filteredContainersToUse.size))
}

/**
Expand Down