Fix terms aggregation doc_count_error_upper_bound for already reduced results (batched query phase) #134645
+22
−21
Conversation
This file contains hidden or 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
benchaplin
added
>non-issue
:Analytics/Aggregations
|
Pinging @elastic/es-analytical-engine (Team:Analytics) |
|
Pinging @elastic/es-search-foundations (Team:Search Foundations) |
6 tasks
jimczi
reviewed
Sep 18, 2025
|
|
||
| protected abstract AggregationReduceContext forSubAgg(AggregationBuilder sub); | ||
|
|
||
| public boolean doesFinalReduceHaveBatchedResult() { |
There was a problem hiding this comment.
nit: that feels weird since it depends on isFinalReduce? Maybe rename it into hasBatchedResult so that callers have to check isFinalReduce and hasBatchedResult?
| return finalReduceHasBatchedResult; | ||
| } | ||
|
|
||
| public void setFinalReduceHasBatchedResult(boolean finalReduceHasBatchedResult) { |
There was a problem hiding this comment.
This should be final and set in the ForFinal ctr?
| // If we are reducing only one aggregation (size == 1), the doc count error should be 0. | ||
| // However, the presence of a batched query result implies this is a final reduction and a partial reduction with size > 1 | ||
| // has already occurred on a data node. The doc count error should not be 0 in this case. | ||
| docCountError = size == 1 && reduceContext.doesFinalReduceHaveBatchedResult() == false ? 0 : sumDocCountError; |
There was a problem hiding this comment.
Does that also handle the case where the partial reduction happens on the coord node (when reaching reduce batch size)?
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Outdated suggestions cannot be applied.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.
There's logic in AbstractInternalTerms that sets the top-level doc_count_error_upper_bound to 0 if only one aggregation is being reduced. However, now that we have batched query execution, it's possible a reduction of multiple aggregations has already occurred on a data node, and is now undergoing a final reduction on the coordinating node.
I discovered this case while attempting to remove the settings override turning off batched query execution in TermsDocCountErrorIT (testFixedDocs with -Dtests.seed=ABFC03388645940D):
I've attempted a lightweight fix here: flag the existence of a batched query result in the AggregationReduceContext, then check this flag when potentially setting doc_count_error_upper_bound to 0. If a batched query result is present, it implies multiple shards were reduced remotely (a single shard on a data node is not batched).