[fix][ml] Fix eviction trigger race that cleared the in-progress marker - #25988
Merged
Merged
Conversation
triggerEvictionWhenNeeded used updateAndGet with a lambda returning null when another eviction was already in progress, which overwrote the running eviction's marker with null. The looping thread would then install its own future and submit a duplicate eviction task, while the first cycle's cleanup cleared the second's marker. Use compareAndSet so only the winning thread installs the marker and losers re-read the current owner's future.
void-ptr974
approved these changes
Jun 10, 2026
dao-jun
approved these changes
Jun 10, 2026
lhotari
added
release/4.2.3
release/4.0.12
release/4.1.4
and removed
release/4.0.12
labels
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
4 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Motivation
RangeEntryCacheManagerImpl.triggerEvictionWhenNeeded()usesupdateAndGetwith a lambda that returnsnullwhen another eviction is already in progress. SinceupdateAndGetstores the lambda's result, a thread losing the race overwrites the running eviction's in-progress marker withnull. On its next loop iteration that thread installs its own future and submits a duplicate eviction task, while the first cycle's cleanup then clears the second cycle's marker.With the cache sitting at the eviction trigger threshold and many managed-ledger threads inserting concurrently, this happens routinely: redundant tasks are queued on the single eviction executor (each re-checks the size, so the wasted work is bounded, but timer-driven eviction passes get delayed behind the no-op tasks), and the marker no longer reliably reflects whether an eviction is running.
Modifications
Use
compareAndSet(null, newEvictionFuture)so only the winning thread installs the in-progress marker; losing threads loop and pick up the owner's future viaget(). The owner remains the only one clearing the marker intriggerEvictionToMakeSpace'sfinallyblock.