Improvements from live testing#6404
Merged
Merged
Conversation
guilload
approved these changes
May 8, 2026
nadav-govari
force-pushed
the
nadav/pr12
branch
2 times, most recently
from
4a63122 to
175f802
Compare
nadav-govari
commented
May 12, 2026
| .with_split_state(SplitState::Published) | ||
| .retain_immature(OffsetDateTime::now_utc()) | ||
| .with_update_timestamp_gte(self.cursor); | ||
| .sort_by_maturity_timestamp() |
Collaborator
Author
There was a problem hiding this comment.
investigate sending NOT IN in flight merges
Refactor migrations
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
There were a number of observations made when testing.
When deploying >2 compactors, the split cache becomes an impediment rather than an optimization. We end up with a bunch of splits on disk that we never end up using in a merge. Rather than fix the split cache, remove it (for now. I might come back and fix this down the road)
The merge policy
operationsonly produced one merge per level per call, which resulted in huge split backlogs buildings. We now construct as many merge operations as we can on every tick/The splits table wasnt sorted the way we needed, so our scans were expensive. Added an index to make this easier.
Got rid of the scan cursor. We can do this naturally with a scan limit and a maturity greater than. We'll always fetch the <= 5000 oldest immature splits. As splits get merged and marked for deletion, the head will move naturally.
How was this PR tested?
Extensive testing on a test cluster that sees meaningful traffic, including a backfill of over 900k immature splits using the compactor architecture.