Fix distributor rebatch bug#5186
Merged
mdisibio merged 3 commits intografana:mainfrom May 29, 2025
Merged
Conversation
… only for ring sharding (as required)
joe-elliott
reviewed
May 29, 2025
joe-elliott
reviewed
May 29, 2025
| maxAttributeBytes := d.getMaxAttributeBytes(userID) | ||
|
|
||
| keys, rebatchedTraces, truncatedAttributeCount, err := requestsByTraceID(batches, userID, spanCount, maxAttributeBytes) | ||
| ringTokens, rebatchedTraces, truncatedAttributeCount, err := requestsByTraceID(batches, userID, spanCount, maxAttributeBytes) |
joe-elliott
approved these changes
May 29, 2025
mdisibio
requested review from
carles-grafana,
electron0zero,
javiermolinar,
mapno,
stoewer,
yvrhdn and
zalegrala
as code owners
mdisibio
added a commit
to mdisibio/tempo
that referenced
this pull request
Jun 3, 2025
* Fix distributor rebatch bug, by not using a 32-bit hash for deduping, only for ring sharding (as required) * lint * changelog
carles-grafana
pushed a commit
to carles-grafana/tempo
that referenced
this pull request
Jun 4, 2025
* Fix distributor rebatch bug, by not using a 32-bit hash for deduping, only for ring sharding (as required) * lint * changelog
3 tasks
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 this PR does:
Distributors rebatch incoming writes by trace ID, in order to ensure that all spans for a given trace end up on the same ingesters over time. It's using a 32-bit hash which can lead to collisions. When 2 traces in the incoming write request have the same hash (collide), their spans will get intermixed, i.e. bad data.
The core issue is that the distributor logic was conflating hashing for the ring, and hashing for dedupe. These don't need to be the same hash, nor should they be. Because the ring requires 32-bit hashes and collisions there don't matter, it just means 2 traces go the same ingesters (normal and unavoidable).
But collisions for dedupe must be avoided, and we do this by swapping that part to a 64-bit hash. Same approach as in trace combiner for spans, and there are test cases for known trace ID collisions under the old method and a test to check the collision rate of the new method.
Which issue(s) this PR fixes:
Fixes #
Checklist
CHANGELOG.mdupdated - the order of entries should be[CHANGE],[FEATURE],[ENHANCEMENT],[BUGFIX]