Github user xuanyuanking commented on a diff in the pull request: https://github.com/apache/spark/pull/23207#discussion_r239067552 --- Diff: sql/core/src/main/scala/org/apache/spark/sql/execution/metric/SQLMetrics.scala --- @@ -163,6 +171,8 @@ object SQLMetrics { Utils.bytesToString } else if (metricsType == TIMING_METRIC) { Utils.msDurationToString + } else if (metricsType == NS_TIMING_METRIC) { + duration => Utils.msDurationToString(duration / 1000 / 1000) --- End diff -- Maybe it's ok, as I test this locally with UT in SQLMetricsSuites, result below: ``` shuffle records written: 2 shuffle write time total (min, med, max): 37 ms (37 ms, 37 ms, 37 ms) shuffle bytes written total (min, med, max): 66.0 B (66.0 B, 66.0 B, 66.0 ``` In the actual scenario the shuffle bytes written will be more larger, and keep the time to ms maybe enough, WDYT?
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org