test: stabilize QueryVirtualStorageTest metric assertions

[FrankChen021]

Summary

• Stabilize the cumulative virtual storage metric assertions in QueryVirtualStorageTest.testQueryPartials.
• Replace race-prone waitForNextEvent calls with LatchableEmitter.waitForEventAggregate waits where the test reads cumulative metric totals.
• Preserve the existing metric thresholds and avoid sleeps, timeout increases, or weakened assertions.

Root cause

StorageMonitor emits virtual storage metrics on periodic monitor ticks and resets the interval statistics after each tick. The test previously waited for only one future matching event with waitForNextEvent, then read a cumulative total. If that event represented an interval before the final segment load completed, the test could observe 23 loads even though a later metric event would bring the cumulative total to the required 24.

waitForEventAggregate evaluates already-processed matching events and continues collecting future matching events until the requested cumulative threshold is reached. The fix therefore removes the timing race while retaining the expected thresholds.

Evidence

• PR #19879 failed unit job: QueryVirtualStorageTest.testQueryPartials failed at the first load assertion with expected 24 but only got 23.
• Independent PR #19876 run: the same failure occurred on an unrelated commit.
• Later PR #19876 run: the same test passed, demonstrating intermittent behavior across unrelated commits.

This is a test-stability fix found while validating the JUnit 5 migration. It does not migrate JUnit, change production behavior, or change the expected metric thresholds. Related to #13948.

Validation

• mvn -ntp test -pl embedded-tests -am -Dtest="org.apache.druid.testing.embedded.query.QueryVirtualStorageTest#testQueryPartials" -Dsurefire.failIfNoSpecifiedTests=false -Pskip-static-checks -Dweb.console.skip=true -T1C — 1 test, 0 failures, 0 errors, 0 skipped; full reactor BUILD SUCCESS.
• mvn -ntp -pl embedded-tests -am -DskipTests -Dweb.console.skip=true -Dcheckstyle.skip=false -Dspotbugs.skip=false checkstyle:check spotbugs:check -T1C — BUILD SUCCESS; Checkstyle reported 0 violations and SpotBugs reported no bugs/errors.
• git diff --check — passed.

[Copilot]

Pull request overview

This PR stabilizes QueryVirtualStorageTest.testQueryPartials by replacing race-prone “wait for one next event” latches with aggregate-based waits when asserting cumulative virtual storage metrics, aligning the waiting strategy with how StorageMonitor emits and resets interval stats.

Changes:

• Replaced LatchableEmitter.waitForNextEvent waits with LatchableEmitter.waitForEventAggregate for cumulative metric thresholds (e.g., load begin count, hit count).
• Added aggregate waits for several additional VSF metrics before reading cumulative sums to reduce timing-related flakes.

Suppressed comments (1)

embedded-tests/src/test/java/org/apache/druid/testing/embedded/query/QueryVirtualStorageTest.java:276

• Similar to the read metrics, StorageMonitor emits VSF_EVICT_COUNT and VSF_EVICT_BYTES as separate events. Waiting only for VSF_EVICT_COUNT can release before the bytes event is processed, making the immediate VSF_EVICT_BYTES > 0 assertion flaky. Add an aggregate wait for VSF_EVICT_BYTES before asserting.

        event -> event.hasMetricName(StorageMonitor.VSF_EVICT_COUNT),
        aggregate -> aggregate.hasSumAtLeast(1)
    );
    Assertions.assertTrue(emitter.getMetricEventLongSum(StorageMonitor.VSF_EVICT_COUNT) > 0);
    Assertions.assertTrue(emitter.getMetricEventLongSum(StorageMonitor.VSF_EVICT_BYTES) > 0);

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[FrankChen021]

Severity	Findings
P0	0
P1	0
P2	1
P3	0
Total	1

Severity	Findings
P0	0
P1	0
P2	1
P3	0
Total	1

Reviewed 1 of 1 changed files.

---

This is an automated review by Codex GPT-5.6-Sol