Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/21221#discussion_r207004013
--- Diff:
core/src/main/scala/org/apache/spark/status/AppStatusListener.scala ---
@@ -669,6 +686,34 @@ private[spark] class AppStatusListener(
}
}
}
+
+ // check if there is a new peak value for any of the executor level
memory metrics
+ // for the live UI. SparkListenerExecutorMetricsUpdate events are only
processed
+ // for the live UI.
+ event.executorUpdates.foreach { updates: ExecutorMetrics =>
+ liveExecutors.get(event.execId).foreach { exec: LiveExecutor =>
+ if (exec.peakExecutorMetrics.compareAndUpdatePeakValues(updates)) {
+ maybeUpdate(exec, now)
+ }
+ }
+ }
+ }
+
+ override def onStageExecutorMetrics(executorMetrics:
SparkListenerStageExecutorMetrics): Unit = {
+ val now = System.nanoTime()
+
+ // check if there is a new peak value for any of the executor level
memory metrics,
+ // while reading from the log. SparkListenerStageExecutorMetrics are
only processed
+ // when reading logs.
+ liveExecutors.get(executorMetrics.execId)
+ .orElse(deadExecutors.get(executorMetrics.execId)) match {
+ case Some(exec) =>
+ if
(exec.peakExecutorMetrics.compareAndUpdatePeakValues(executorMetrics.executorMetrics))
{
+ maybeUpdate(exec, now)
--- End diff --
this is only for replaying from logs, right? but then, `maybeUpdate` will
be a no-op, as `live` is false. So, what is the desired effect here? But
maybe I'm missing something, because this is covered by the some of the api
tests. I think maybe you don't need to call update at all here, and it'll just
get written to the kvstore as part of the final `flush` call when the kvstore
is closed after reading the whole log.
also style nit: indentation is off
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
