MMapDirectory loses shared-arena grouping when a closing arena removes a newer live arena

Finding LFV-037Area: store / MMapDirectoryPriority: lowImpact: minorNovelty: known upstream

AI-assisted finding, confirmed by automated checks. Upstream status: Fixed, by Uwe Schindler (Lucene PMC) on 2026-10-28. Uwe Schindler said it looks very relevant: Uwe had long felt something was wrong with shared arenas, and Gianluca Graziadei had privately reported the same issue (Uwe on #16931, Chris Hegarty agreeing; Uwe's first dev@ message said closely related). Chris Hegarty fixed it in #16931 (merged 2026-10-29; Uwe Schindler cherry-picked it to branch_11x) and backported it for 10.6.1 in #16945. On #16931 Uwe Schindler noted that a single thread reproduces it and that it breaks nothing (no unclosed arenas, only some slowdown after many NRT reopens), so 10.6.0 was not respun for it. Source. This site files nothing upstream.

What it shows

You use MMapDirectory with its default shared-arena grouping while files of one group are opened and closed concurrently. You expect the files of a group to share one arena. Instead onClose of an old arena removes the newer live arena's entry, so grouping is lost (577 arenas instead of 10 in the repro).

The test shows that in a default MMapDirectory, closing a group's old arena (A1, permits used up) unregisters the newer arena A2 that is still open, so the next open of the same segment gets a third arena (TestMMapDirectoryArenaGrouping#testOpenAfterOldArenaClosedSharesNewerArena, one file, single thread).

Expected vs actual

Expected: the open after A1 closed shares the open arena A2 (63 permits left). Actual: it gets a new arena; fails with -ea and -da (performance only).

Files

src/org/apache/lucene/store/TestMMapDirectoryArenaGrouping.java (JUnit test)

Run it

Tested on Lucene main 645d4fe14a20 (2026-10-28): reproduced.
Recorded outcome of run 1 on Lucene main 645d4fe14a20: 1 test(s) run, 1 failed.
How to read the outcome: a failing test is the finding reproducing (the failure message states expected against actual).
Seeds: a rerun with the same seed replays the same random choices. Assertions off (-da -dsa -Dtests.asserts=false) is the production-like setting.

# Lucene checkout at the tested commit, built with: ./gradlew --no-daemon assemble   (JDK 25)
LUCENE=/path/to/lucene; V=11.0.0-SNAPSHOT
CP=$LUCENE/lucene/core/build/libs/lucene-core-$V.jar
CP=$CP:$LUCENE/lucene/codecs/build/libs/lucene-codecs-$V.jar
CP=$CP:$LUCENE/lucene/test-framework/build/libs/lucene-test-framework-$V.jar
CP=$CP:randomizedtesting-runner-2.9.2.jar:junit-4.13.2.jar:hamcrest-3.0.jar   # test-framework dependencies (Maven Central)
CP=$CP:junit-platform-console-standalone-6.1.3.jar   # JUnit Platform console launcher (Maven Central)

javac -d classes -cp "$CP" $(find src -name '*.java')

# run 1: assertions on, seed DEADBEEF
java -ea -esa --add-modules jdk.incubator.vector --enable-native-access=ALL-UNNAMED -Dtests.seed=DEADBEEF \
     -cp "classes:$CP" org.junit.platform.console.ConsoleLauncher \
     execute --select-class org.apache.lucene.store.TestMMapDirectoryArenaGrouping

Gradle alternative: copy the test into the matching module's src/test/ tree of the checkout and run ./gradlew -p lucene/<module> test --tests <class> -Ptests.seed=<seed> (add -Ptests.asserts=false for a run without assertions).

Feedback

Open an issue in lucene-ai-bug-hunt. Please do not open issues or pull requests in apache/lucene about this finding unless a Lucene committer agrees it should be filed.

Snapshot of 2026-11-04. AI-assisted; triaged by a Lucene maintainer as noted above. Feedback: open an issue in lucene-ai-bug-hunt.