[SPARK-9104][SPARK-9105][SPARK-9106][SPARK-9107][CORE] Netty network layer memory usage on webUI #7753
refine the code according to Imran's comments and the design doc
- Loading branch information
commit 27b7da1b775b0c4101af3f1f6ee454668a49cdc1
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -20,7 +20,6 @@ package org.apache.spark.scheduler | |
| import java.io._ | ||
| import java.net.URI | ||
|
|
||
| import org.apache.spark.executor.TransportMetrics | ||
|
|
||
|
|
||
| import scala.collection.mutable | ||
|
|
@@ -97,8 +96,9 @@ private[spark] class EventLoggingListener( | |
| private[scheduler] val logPath = getLogPath( | ||
| logBaseDir, appId, appAttemptId, compressionCodecName) | ||
|
Contributor
There was a problem hiding this comment. this shouldn't need to change, right?
Contributor
Author
There was a problem hiding this comment. yes, my mistake, the width is 100, it's correct. I'll change it back later, thanks |
||
|
|
||
| private val executorIdToLatestMetrics = new HashMap[String, SparkListenerExecutorMetricsUpdate] | ||
| private val executorIdToModifiedMaxMetrics = new | ||
| HashMap[String, SparkListenerExecutorMetricsUpdate] | ||
|
|
||
| /** | ||
| * Creates the log file in the configured log directory. | ||
|
|
@@ -161,17 +161,23 @@ private[spark] class EventLoggingListener( | |
| } | ||
| } | ||
|
|
||
| // When a stage is submitted and completed, we updated our executor memory metrics for that stage, | ||
| // and then log the metrics. Anytime we receive more executor metrics, we update our running set of | ||
| // {{executorIdToLatestMetrics}} and {{executorIdToModifiedMaxMetrics}}. Since stages submit and | ||
| // complete time might be interleaved, we maintain the latest and max metrics for each time segment. | ||
| // So, for each stage start and stage complete, we replace each item in | ||
| // {{executorIdToModifiedMaxMetrics}} with that in {{executorIdToLatestMetrics}}. | ||
| private def updateAndLogExecutorMemoryMetrics() : Unit = { | ||
| executorIdToModifiedMaxMetrics.foreach { case(_, metrics) => logEvent(metrics) } | ||
| executorIdToLatestMetrics.foreach {case(_, metrics) => logEvent(metrics) } | ||
| executorIdToLatestMetrics.foreach { case (executorId, metrics) => | ||
| executorIdToModifiedMaxMetrics.update(executorId, metrics) | ||
| } | ||
| } | ||
|
Contributor
There was a problem hiding this comment. I'd rename this to
I don't understand the last two sentences of the comment -- can you expand on that? Finally you should use modifiedMetrics.foreach { case (_, metrics) => logEvent(metrics) }
latestMetrics.foreach { case (executorId, metrics) => modifiedMetrics.update(executorId, metrics) }
Contributor
Author
There was a problem hiding this comment.
I'll update the code according to the design doc. I think the code is not that correct. Please refer it in design doc |
||
|
|
||
| // Events that do not trigger a flush | ||
| override def onStageSubmitted(event: SparkListenerStageSubmitted): Unit = { | ||
| updateAndLogExecutorMemoryMetrics() | ||
| logEvent(event) | ||
| } | ||
|
|
||
|
|
@@ -185,7 +191,7 @@ private[spark] class EventLoggingListener( | |
|
|
||
| // Events that trigger a flush | ||
| override def onStageCompleted(event: SparkListenerStageCompleted): Unit = { | ||
| updateAndLogExecutorMemoryMetrics() | ||
| logEvent(event, flushLogger = true) | ||
| } | ||
|
|
||
|
|
@@ -218,8 +224,8 @@ private[spark] class EventLoggingListener( | |
| } | ||
|
|
||
| override def onExecutorRemoved(event: SparkListenerExecutorRemoved): Unit = { | ||
| executorIdToLatestMetrics.remove(event.executorId) | ||
| executorIdToModifiedMaxMetrics.remove(event.executorId) | ||
| logEvent(event, flushLogger = true) | ||
| } | ||
|
|
||
|
|
@@ -228,7 +234,7 @@ private[spark] class EventLoggingListener( | |
|
|
||
| // No-op because logging every update would be overkill | ||
| override def onExecutorMetricsUpdate(event: SparkListenerExecutorMetricsUpdate): Unit = { | ||
| executorIdToLatestMetrics.update(event.execId, event) | ||
| updateModifiedMetrics(event.execId) | ||
| } | ||
|
|
||
|
|
@@ -258,10 +264,10 @@ private[spark] class EventLoggingListener( | |
| * @param executorId the executor whose metrics will be modified | ||
| */ | ||
| private def updateModifiedMetrics(executorId: String): Unit = { | ||
| val toBeModifiedEvent = executorIdToModifiedMaxMetrics.get(executorId) | ||
| val latestEvent = executorIdToLatestMetrics.get(executorId) | ||
| if (toBeModifiedEvent.isEmpty) { | ||
| if (latestEvent.isDefined) executorIdToModifiedMaxMetrics.update(executorId, latestEvent.get) | ||
| } else { | ||
| val toBeModifiedMetrics = toBeModifiedEvent.get.executorMetrics.transportMetrics | ||
| if (toBeModifiedMetrics.isDefined) { | ||
|
Contributor
There was a problem hiding this comment. won't private def updateModifiedMetrics(executorId: String, latestEvent: SparkListenerExecutorMetricsUpdate): Unit = {
executorIdToModifiedMaxMetrics.get(executorId) match {
case None => executorIdToModifiedMaxMetrics.update(executorId, latestEvent)
case Some(toBeModEvent) =>
val toBeModMetrics = toBeModEvent.executorMetrics.transportMetrics
...
}
}and depending on whether or not we need to keep
Contributor
Author
There was a problem hiding this comment.
Thank you for you cleaner code style example, I've updated in my code. |
||
|
|
@@ -270,29 +276,23 @@ private[spark] class EventLoggingListener( | |
| val toBeModTransMetrics = toBeModifiedMetrics.get | ||
| var timeStamp: Long = toBeModTransMetrics.timeStamp | ||
| // the logic here should be the same with that for memoryListener | ||
| val onHeapSize = if (latestTransMetrics.onHeapSize > toBeModTransMetrics.onHeapSize) { | ||
| timeStamp = latestTransMetrics.timeStamp | ||
| latestTransMetrics.onHeapSize | ||
| } else { | ||
| toBeModTransMetrics.onHeapSize | ||
| } | ||
| val directSize = if (latestTransMetrics.directSize > toBeModTransMetrics.directSize) { | ||
| timeStamp = latestTransMetrics.timeStamp | ||
| latestTransMetrics.directSize | ||
| } else { | ||
| toBeModTransMetrics.directSize | ||
| } | ||
| toBeModifiedEvent.get.executorMetrics.setTransportMetrics( | ||
| Some(TransportMetrics(timeStamp, onHeapSize, directSize))) | ||
| } | ||
| } | ||
| } | ||
| } | ||
|
|
||
| private[spark] object EventLoggingListener extends Logging { | ||
|
|
||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
whats the point of this? sorry if we discussed it earlier ... if its just to test the
ExecutorMetricsreally is serializable, that would be better in a test caseThere was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is due to SPARK-3465. Currently we do not have any aggregation operations for
ExecutorMetrics, we can remove this. We can add it back when we do some aggregation.There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ah, this is a great point. In that case, I think you can leave it in for now, but add a comment that the serialization & deserialization is just to make a copy, for to SPARK-3465