xuanyuanking commented on a change in pull request #30427:
URL: https://github.com/apache/spark/pull/30427#discussion_r528546502
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/streaming/ui/StreamingQueryStatisticsPage.scala
##########
@@ -126,6 +126,58 @@ private[ui] class StreamingQueryStatisticsPage(parent:
StreamingQueryTab)
<br />
}
+ def generateWatermark(
+ query: StreamingQueryUIData,
+ minBatchTime: Long,
+ maxBatchTime: Long,
+ jsCollector: JsCollector): Seq[Node] = {
+ // This is made sure on caller side but put it here to be defensive
+ require(query.lastProgress != null)
+ if (query.lastProgress.eventTime.containsKey("watermark")) {
+ val watermarkData = query.recentProgress.flatMap { p =>
+ val batchTimestamp = parseProgressTimestamp(p.timestamp)
+ val watermarkValue =
parseProgressTimestamp(p.eventTime.get("watermark"))
+ if (watermarkValue > 0L) {
+ // seconds
+ Some((batchTimestamp, ((batchTimestamp - watermarkValue) / 1000.0)))
Review comment:
Fully agree with `knowing the gap between actual wall clock and
watermark looks more useful than the absolute value.` and thanks for the super
useful watermark info!
I only have one concern same with @viirya
https://github.com/apache/spark/pull/30427#issuecomment-730852396. Maybe we can
address both scenarios(event time is/isn't close to clock time) by using the
max event time received in this batch? I mean:
```
if (watermarkValue > 0L) {
// seconds
- Some((batchTimestamp, ((batchTimestamp - watermarkValue) /
1000.0)))
+ val maxEventTime = parseProgressTimestamp(p.eventTime.get("max"))
+ Some((batchTimestamp, (maxEventTime - watermarkValue) / 1000.0))
} else {
None
}
```
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
