Skip to content

[SPARK-19537] Move pendingPartitions to ShuffleMapStage. #16876

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
kayousterhout wants to merge 2 commits into from
Closed

[SPARK-19537] Move pendingPartitions to ShuffleMapStage. #16876

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3e9af42
[SPARK-19537] Move pendingPartitions to ShuffleMapStage.
kayousterhout Feb 9, 2017
8fdecda
Updated DAGScheduler partition log message
kayousterhout Feb 10, 2017
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
20 changes: 13 additions & 7 deletions core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -931,8 +931,6 @@ class DAGScheduler(
/** Called when stage's parents are available and we can now do its task. */
private def submitMissingTasks(stage: Stage, jobId: Int) {
logDebug("submitMissingTasks(" + stage + ")")
// Get our pending tasks and remember them in our pendingTasks entry
stage.pendingPartitions.clear()

// First figure out the indexes of partition ids to compute.
val partitionsToCompute: Seq[Int] = stage.findMissingPartitions()
Expand Down Expand Up @@ -1012,9 +1010,11 @@ class DAGScheduler(
val serializedTaskMetrics = closureSerializer.serialize(stage.latestInfo.taskMetrics).array()
stage match {
case stage: ShuffleMapStage =>
stage.pendingPartitions.clear()
partitionsToCompute.map { id =>
val locs = taskIdToLocations(id)
val part = stage.rdd.partitions(id)
stage.pendingPartitions += id
new ShuffleMapTask(stage.id, stage.latestInfo.attemptId,
taskBinary, part, locs, properties, serializedTaskMetrics, Option(jobId),
Option(sc.applicationId), sc.applicationAttemptId)
Expand All @@ -1038,9 +1038,8 @@ class DAGScheduler(
}

if (tasks.size > 0) {
logInfo("Submitting " + tasks.size + " missing tasks from " + stage + " (" + stage.rdd + ")")
stage.pendingPartitions ++= tasks.map(_.partitionId)
logDebug("New pending partitions: " + stage.pendingPartitions)
logInfo(s"Submitting ${tasks.size} missing tasks (for partitions " +
s"${tasks.map(_.partitionId)}) from $stage (${stage.rdd})")
Copy link
Contributor

@squito squito Feb 10, 2017 •
edited
Loading

Choose a reason for hiding this comment

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

I think including all the partitions could make this log line a little too long for an info message (what if there are 1k or 10k tasks). I understand not wanting to stick it in a debug message (you often dont' realize you want debug logging on until after something has failed), but I think this will be too much for regular runs.

Would it still be useful if you truncate to the first 25 tasks? tasks.take(25).map(...)

Copy link
Contributor Author

@kayousterhout kayousterhout Feb 10, 2017

Choose a reason for hiding this comment

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

Oh good point. 25 seemed like maybe too many for an info-level message so I split the difference with your original comment and did 15. I could also just leave the original logging if you think that's better.

taskScheduler.submitTasks(new TaskSet(
tasks.toArray, stage.id, stage.latestInfo.attemptId, jobId, properties))
stage.latestInfo.submissionTime = Some(clock.getTimeMillis())
Expand Down Expand Up @@ -1146,7 +1145,6 @@ class DAGScheduler(
val stage = stageIdToStage(task.stageId)
event.reason match {
case Success =>
stage.pendingPartitions -= task.partitionId
task match {
case rt: ResultTask[_, _] =>
// Cast to ResultStage here because it's part of the ResultTask
Expand Down Expand Up @@ -1182,6 +1180,7 @@ class DAGScheduler(

case smt: ShuffleMapTask =>
val shuffleStage = stage.asInstanceOf[ShuffleMapStage]
shuffleStage.pendingPartitions -= task.partitionId
updateAccumulators(event)
val status = event.result.asInstanceOf[MapStatus]
val execId = status.location.executorId
Expand Down Expand Up @@ -1234,7 +1233,14 @@ class DAGScheduler(

case Resubmitted =>
logInfo("Resubmitted " + task + ", so marking it as still running")
stage.pendingPartitions += task.partitionId
stage match {
case sms: ShuffleMapStage =>
sms.pendingPartitions += task.partitionId

case _ =>
assert(false, "TaskSetManagers should only send Resubmitted task statuses for " +
"tasks in ShuffleMapStages.")
}

case FetchFailed(bmAddress, shuffleId, mapId, reduceId, failureMessage) =>
val failedStage = stageIdToStage(task.stageId)
Expand Down
13 changes: 13 additions & 0 deletions core/src/main/scala/org/apache/spark/scheduler/ShuffleMapStage.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,8 @@

package org.apache.spark.scheduler

import scala.collection.mutable.HashSet

import org.apache.spark.ShuffleDependency
import org.apache.spark.rdd.RDD
import org.apache.spark.storage.BlockManagerId
Expand Down Expand Up @@ -47,6 +49,17 @@ private[spark] class ShuffleMapStage(

private[this] var _numAvailableOutputs: Int = 0

/**
* Partitions that either haven't yet been computed, or that were computed on an executor
* that has since been lost, so should be re-computed. This variable is used by the
* DAGScheduler to determine when a stage has completed. Task successes in both the active
* attempt for the stage or in earlier attempts for this stage can cause paritition ids to get
* removed from pendingPartitions. As a result, this variable may be inconsistent with the pending
* tasks in the TaskSetManager for the active attempt for the stage (the partitions stored here
* will always be a subset of the partitions that the TaskSetManager thinks are pending).
*/
val pendingPartitions = new HashSet[Int]

/**
* List of [[MapStatus]] for each partition. The index of the array is the map partition id,
* and each value in the array is the list of possible [[MapStatus]] for a partition
Expand Down
2 changes: 0 additions & 2 deletions core/src/main/scala/org/apache/spark/scheduler/Stage.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,8 +67,6 @@ private[scheduler] abstract class Stage(
/** Set of jobs that this stage belongs to. */
val jobIds = new HashSet[Int]

val pendingPartitions = new HashSet[Int]

/** The ID to use for the next new attempt for this stage. */
private var nextAttemptId: Int = 0

Expand Down