[SPARK-50372][CONNECT][SQL] Make all DF execution path collect observed metrics #48920
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
xupefei
commented
Nov 21, 2024
| CollectMetrics(name, metrics.map(_.named), transformRelation(rel.getInput), planId) | ||
| CollectMetrics(name, metrics.map(_.named), input, planId) |
There was a problem hiding this comment.
This fixes a bug where the input of a CollectMetrics can be processed two times, once in Line 1190 and once here/below.
When the input contains another CollectMetrics, transforming it twice will cause two Observation objects (in the input) to be initialised and registered two times to the system. Since only one of them will be fulfilled when the query finishes, the one we'll be looking at may not have any data.
This issue is highlighted in the test case Observation.get is blocked until the query is finished ..., where we specifically execute observedObservedDf, which is a CollectMetrics that has another CollectMetrics as its input.
vicennial
approved these changes
Nov 21, 2024
connector/connect/client/jvm/src/main/scala/org/apache/spark/sql/SparkSession.scala
Outdated
Show resolved
Hide resolved
HyukjinKwon
changed the title
HyukjinKwon
approved these changes
Nov 21, 2024
Done! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What changes were proposed in this pull request?
This PR fixes an issue that some of DataFrame execution paths would not process
ObservedMetrics. The fix is done by injecting a lazy processing logic into the result iterator.The following private execution APIs are affected by this issue:
SparkSession.execute(proto.Relation.Builder)SparkSession.execute(proto.Command)SparkSession.execute(proto.Plan)The following user-facing API is affected by this issue:
DataFrame.write.format("...").mode("...").save()This PR also fixes an issue in which on the Server side, two observed metrics can be assigned to the same Plan ID when they are in the same plan (e.g., one observation is used as the input of another). The fix is to traverse the plan and find all observations with correct IDs.
Another bug is discovered as a byproduct of introducing a new test case. Copying the PR comment here from SparkConnectPlanner.scala:
Why are the changes needed?
To fix a bug.
Does this PR introduce any user-facing change?
Yes, this bug is user-facing.
How was this patch tested?
New tests.
Was this patch authored or co-authored using generative AI tooling?
No.