squito commented on a change in pull request #22874: [SPARK-25865][CORE] Add GC
information to ExecutorMetrics
URL: https://github.com/apache/spark/pull/22874#discussion_r258672092
##########
File path:
core/src/main/scala/org/apache/spark/metrics/ExecutorMetricType.scala
##########
@@ -99,6 +102,56 @@ case object ProcessTreeMetrics extends ExecutorMetricType {
}
}
+case object GarbageCollectionMetrics extends ExecutorMetricType with Logging {
+ override val names = Seq(
+ "MinorGCCount",
+ "MinorGCTime",
+ "MajorGCCount",
+ "MajorGCTime"
+ )
+
+ /* We builtin some common GC collectors which categorized as young
generation and old */
+ private[spark] val YOUNG_GENERATION_BUILTIN_GARBAGE_COLLECTORS = Seq(
+ "Copy",
+ "PS Scavenge",
+ "ParNew",
+ "G1 Young Generation"
+ )
+
+ private[spark] val OLD_GENERATION_BUILTIN_GARBAGE_COLLECTORS = Seq(
+ "MarkSweepCompact",
+ "PS MarkSweep",
+ "ConcurrentMarkSweep",
+ "G1 Old Generation"
+ )
+
+ private lazy val youngGenerationGarbageCollector: Seq[String] = {
+
SparkEnv.get.conf.get(config.EVENT_LOG_GC_METRICS_YOUNG_GENERATION_GARBAGE_COLLECTORS)
+ }
+
+ private lazy val oldGenerationGarbageCollector: Seq[String] = {
+
SparkEnv.get.conf.get(config.EVENT_LOG_GC_METRICS_OLD_GENERATION_GARBAGE_COLLECTORS)
+ }
+
+ override private[spark] def getMetricValues(memoryManager: MemoryManager):
Array[Long] = {
+ val gcMetrics = new Array[Long](names.length) // minorCount, minorTime,
majorCount, majorTime
+ ManagementFactory.getGarbageCollectorMXBeans.asScala.foreach { mxBean =>
+ if (youngGenerationGarbageCollector.contains(mxBean.getName)) {
+ gcMetrics(0) = mxBean.getCollectionCount
+ gcMetrics(1) = mxBean.getCollectionTime
+ } else if (oldGenerationGarbageCollector.contains(mxBean.getName)) {
+ gcMetrics(2) = mxBean.getCollectionCount
+ gcMetrics(3) = mxBean.getCollectionTime
+ } else {
+ logDebug(s"${mxBean.getName} is an unsupported garbage collector." +
Review comment:
sorry one more thing -- with SPARK-26329
https://github.com/apache/spark/pull/23767 this may show up a lot more. Can
you update this so the msg only gets printed once? should be pretty simple.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
