Skip to content

[SPARK-24637][SS] Add metrics regarding state and watermark to dropwizard metrics #21622

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
HeartSaVioR wants to merge 2 commits into from
+23 −0
Closed

[SPARK-24637][SS] Add metrics regarding state and watermark to dropwizard metrics #21622

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
147c98a
SPARK-24637 Add metrics regarding state and watermark to dropwizard m…
HeartSaVioR Jun 23, 2018
722e6a0
Address minor review comment
HeartSaVioR Jun 26, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 20 additions & 0 deletions sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/MetricsReporter.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,10 +17,13 @@

package org.apache.spark.sql.execution.streaming

import java.text.SimpleDateFormat

import com.codahale.metrics.{Gauge, MetricRegistry}

import org.apache.spark.internal.Logging
import org.apache.spark.metrics.source.{Source => CodahaleSource}
import org.apache.spark.sql.catalyst.util.DateTimeUtils
import org.apache.spark.sql.streaming.StreamingQueryProgress

/**
Expand All @@ -39,6 +42,23 @@ class MetricsReporter(
registerGauge("processingRate-total", _.processedRowsPerSecond, 0.0)
registerGauge("latency", _.durationMs.get("triggerExecution").longValue(), 0L)

private val timestampFormat = new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss.SSS'Z'") // ISO8601
timestampFormat.setTimeZone(DateTimeUtils.getTimeZone("UTC"))

registerGauge("eventTime-watermark",
s => convertStringDateToMillis(s.eventTime.get("watermark")), 0L)

@arunmahadevan arunmahadevan Jun 26, 2018

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. nit: rename s => progress to make it clear.
  2. The eventTime-watermark metrics needs to be reported only if the map is not empty (event time). Could be skipped if the map is empty (processing time) to avoid confusion.

@HeartSaVioR HeartSaVioR Jun 26, 2018

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. will address
  2. We don't know whether the map will be empty when calling registerGauge, and once we register the metric, getValue in Gauge is called from Dropwizard so I'm not sure we can control whether reporting the value or not.


registerGauge("states-rowsTotal", _.stateOperators.map(_.numRowsTotal).sum, 0L)
registerGauge("states-usedBytes", _.stateOperators.map(_.memoryUsedBytes).sum, 0L)

@HeartSaVioR HeartSaVioR Jul 6, 2018

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We can add more metrics like "providerLoadedMapSizeBytes" after adopting SPARK-24441, so that actual memory usage of state store provider could be tracked via time-series manner.

@tdas tdas Aug 1, 2018

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Those are custom metrics, which may or may not be present depending on the implementation of state store. I dont recommend adding them here directly.

@HeartSaVioR HeartSaVioR Aug 1, 2018 •
edited
Loading

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the input! I'll keep the patch as it is.

Could you suggest some approaches to extend the maintained metrics? I would like to expand more, and newer things might be coming from custom metrics (like from source and sink) so might be worth to have extension point.

My question is beyond of this PR, so please continue reviewing the patch. Thanks!

private def convertStringDateToMillis(isoUtcDateStr: String) = {
if (isoUtcDateStr != null) {
timestampFormat.parse(isoUtcDateStr).getTime
} else {
0L
}
}

private def registerGauge[T](
name: String,
f: StreamingQueryProgress => T,
Expand Down
3 changes: 3 additions & 0 deletions sql/core/src/test/scala/org/apache/spark/sql/streaming/StreamingQuerySuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -462,6 +462,9 @@ class StreamingQuerySuite extends StreamTest with BeforeAndAfter with Logging wi
assert(gauges.get("latency").getValue.asInstanceOf[Long] == 0)
assert(gauges.get("processingRate-total").getValue.asInstanceOf[Double] == 0.0)
assert(gauges.get("inputRate-total").getValue.asInstanceOf[Double] == 0.0)
assert(gauges.get("eventTime-watermark").getValue.asInstanceOf[Long] == 0)
assert(gauges.get("states-rowsTotal").getValue.asInstanceOf[Long] == 0)
assert(gauges.get("states-usedBytes").getValue.asInstanceOf[Long] == 0)
sq.stop()
}
}
Expand Down