Skip to content

[SPARK-29562][sql] Speed up and slim down metric aggregation in SQL listener. #26218

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
vanzin wants to merge 4 commits into from
Closed

[SPARK-29562][sql] Speed up and slim down metric aggregation in SQL listener. #26218

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
067d63e
[SPARK-29562][sql] Speed up and slim down metric aggregation in SQL l…
Oct 18, 2019
051ff5f
Feedback.
Oct 23, 2019
9174f52
Add bench result files.
Oct 24, 2019
6377c70
Sigh.
Oct 25, 2019
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
Prev Previous commit
Next Next commit
Feedback.
  • Loading branch information
Marcelo Vanzin committed Oct 23, 2019
commit 051ff5f719b4de083cfe922e598aedf2a1ae13bd
13 changes: 9 additions & 4 deletions sql/core/src/main/scala/org/apache/spark/sql/execution/ui/SQLAppStatusListener.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,7 +146,8 @@ class SQLAppStatusListener(

override def onExecutorMetricsUpdate(event: SparkListenerExecutorMetricsUpdate): Unit = {
event.accumUpdates.foreach { case (taskId, stageId, attemptId, accumUpdates) =>
updateStageMetrics(stageId, attemptId, taskId, -1, accumUpdates, false)
updateStageMetrics(stageId, attemptId, taskId, SQLAppStatusListener.UNKNOWN_PARTITION,
accumUpdates, false)
}
}

Expand Down Expand Up @@ -245,7 +246,7 @@ class SQLAppStatusListener(
stageId: Int,
attemptId: Int,
taskId: Long,
partIdx: Int, // -1 if unknown from the event data.
partIdx: Int,
accumUpdates: Seq[AccumulableInfo],
succeeded: Boolean): Unit = {
Option(stageMetrics.get(stageId)).foreach { metrics =>
Expand Down Expand Up @@ -462,7 +463,7 @@ private class LiveStageMetrics(
eventPartIdx: Int,
finished: Boolean,
accumUpdates: Seq[AccumulableInfo]): Unit = {
val partIdx = if (eventPartIdx == -1) {
val partIdx = if (eventPartIdx == SQLAppStatusListener.UNKNOWN_PARTITION) {
if (!taskIndices.contains(taskId)) {
// We probably missed the start event for the task, just ignore it.
return
Expand All @@ -482,7 +483,7 @@ private class LiveStageMetrics(
.filter { acc => acc.update.isDefined && accumulatorIds.contains(acc.id) }
.foreach { acc =>
// In a live application, accumulators have Long values, but when reading from event
// logs, they have String values. For now, assume all accumulators are Long and covert
// logs, they have String values. For now, assume all accumulators are Long and convert
// accordingly.
val value = acc.update.get match {
case s: String => s.toLong
Expand All @@ -501,3 +502,7 @@ private class LiveStageMetrics(

def metricValues(): Seq[(Long, Array[Long])] = taskMetrics.asScala.toSeq
}

private object SQLAppStatusListener {
val UNKNOWN_PARTITION = -1
}
1 change: 0 additions & 1 deletion sql/core/src/test/scala/org/apache/spark/sql/execution/ui/MetricsAggregationBenchmark.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -207,5 +207,4 @@ object MetricsAggregationBenchmark extends BenchmarkBase {
case class Measurements(
taskEventsTimes: Seq[Long],
aggregationTime: Long)

}