[CI] TestCuVSAcceleratedHNSWGaps.testVectorSearchWithAlternatingDocuments is flaky

Open Beginner friendly
#2,551 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
2/5
Estimated time
1-3 hours
Newbie friendliness
72/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
java
Domain
search, testing-qa

Research direction

Start with TestCuVSAcceleratedHNSWGaps.testVectorSearchWithAlternatingDocuments and its calculateExpectedTopK helper. Reproduce the failure with the supplied tests.seed, locale, and timezone, then verify the assertion reflects approximate HNSW results rather than requiring exact top-k membership. Done means the test remains meaningful and passes for legitimate approximate results.

Written by the indexing model from the issue text.

Description

ci Lucene

calculateExpectedTopK computes the exact Euclidean top-k and the test asserts that every result returned by the HNSW search must appear in that set:

assertTrue("Result " + id + " should be in expected top-k results", expectedIds.contains(id));

Since Lucene99AcceleratedHNSWVectorsFormat uses an approximate nearest-neighbor algorithm (CAGRA-built HNSW graph), it is not guaranteed to return the exact top-k. The assertion fails for legitimately correct approximate results.

Reproduced with:

-Dtests.seed=FFE4A5B93F63EAEF -Dtests.locale=fr-BE -Dtests.timezone=SystemV/PST8PDT

Failure:

Result 858 should be in expected top-k results
Dominant language
Cuda
Stars
854
Forks
236
Avg merge
2d 23h
Merged PRs (30d)
64

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from NVIDIA/cuvs

All issues in NVIDIA/cuvs

Similar issues

More Search issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.