Github user rdblue commented on a diff in the pull request:
https://github.com/apache/spark/pull/17540#discussion_r112606599
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/ui/SQLListener.scala ---
@@ -353,11 +353,19 @@ class SQLListener(conf: SparkConf) extends
SparkListener with Logging {
}
val driverUpdates = executionUIData.driverAccumUpdates.toSeq
- val totalUpdates = (accumulatorUpdates ++ driverUpdates).filter {
- case (id, _) => executionUIData.accumulatorMetrics.contains(id)
+ // filter out updates that aren't Longs
+ val totalUpdates = (accumulatorUpdates ++ driverUpdates).filter {
update =>
+ update._2.isInstanceOf[Long]
}
+ // Hack alert!
+ // metrics may be missing from the executionUIData because metrics
are taken from the
+ // SparkPlan passed to withNewExecutionId, but some nodes in that
plan link in LogicalPlan
+ // nodes that don't expose the actual SparkPlan or metrics because
the SparkPlan is built
--- End diff --
I think it's better to have the metrics than to discard them. Why discard
metrics if we know that they can be safely summarized and used?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
