Skip to content

parquet tests fails inconsistently in CI聽#813

Description

@deependujha

馃悰 Bug

Parquet tests fail inconsistently in CI, mostly due to timeout, waiting for index file to be available. Possibly, multiple parquet tests leaking the environment.

To Reproduce

Steps to reproduce the behavior...

Code sample
# Ideally attach a minimal code sample to reproduce the decried issue.
# Minimal means having the shortest code but still preserving the bug.

Alternatively, you can share a fully reproducible Lightning Studio environment:

A simple guide on how to create such a studio can be found here.

  1. Create a Studio.
  2. Reproduce the issue in the Studio.
  3. Publish the Studio.
  4. Paste the Studio link here.

Expected behavior

Additional context

Environment detail
  • PyTorch Version (e.g., 1.0):
  • OS (e.g., Linux):
  • How you installed PyTorch (conda, pip, source):
  • Build command you used (if compiling from source):
  • Python version:
  • CUDA/cuDNN version:
  • GPU models and configuration:
  • Any other relevant information:

Activity

  1. snakefood3232 commented on Apr 29, 2026

    @snakefood3232

    I've noticed your parquet tests are failing inconsistently due to timeouts related to index file availability. I'd handle this by isolating the test environment or ensuring proper teardown to prevent leaks causing these issues. I can have a PR ready for review within 48 hours. I've previously optimized data pipeline tests in CI environments.

  2. kostasrigatos commented on Oct 9, 2026

    @kostasrigatos

    I looked into this. #810 already fixed the cause described here for the tests that share
    ~/.lightning/chunks, but three optimize_hf tests in tests/streaming/test_parquet.py still use the
    shared $TMPDIR/chunks and $TMPDIR/data scratch dirs, and two of them aren't in the xdist group.
    Every optimize() call deletes and recreates those dirs (_cleanup_cache), so the tests can race when
    they land on different workers. A macOS py3.10 CI job failed this way after #810 (FileNotFoundError
    inside shutil.rmtree). The other recent parquet failure I found (assert 3 == 2) was on a branch
    before #897 and is already fixed on main.

    I'll open a PR that applies #810's pattern (per-test DATA_OPTIMIZER_CACHE_FOLDER /
    DATA_OPTIMIZER_DATA_CACHE_FOLDER) to those three tests.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't workinghelp wantedExtra attention is needed

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions