Skip to content

Flaky test: partition_order_runs_at_most_the_given_number_of_partitions assumes concurrent partitions start in order #7676

Open

Description

@dwsmith1983

Describe the bug

partition_order_runs_at_most_the_given_number_of_partitions in native/local/tests/execution.rs, added in #6883, fails intermittently in the merge queue:

thread 'partition_order_runs_at_most_the_given_number_of_partitions' panicked at local/tests/execution.rs:493:5:
assertion `left == right` failed
  left: [1, 0]
 right: [0, 1]

It has failed merge queue runs for #7556 (run), #7464 (run, run) and #7386 (run) since 2026-11-17, and it evicted #7556 and #7464. None of these PRs touches native/local.

The query admits two partitions at once on a four-worker runtime, and execute() records each partition in executions as it starts, so 0 and 1 can arrive in either order. assert_executed_once sorts before comparing, but the check at line 493 does not.

Steps to reproduce

Run cargo test -p datafusion-comet-local --test execution partition_order_runs_at_most_the_given_number_of_partitions repeatedly; it depends on which of the first two partitions reaches execute() first.

Expected behavior

The check verifies that only two partitions have started, not which one started first, so it should compare the sorted list.

Activity

  1. dwsmith1983 commented on Nov 18, 2026

    @dwsmith1983
    ContributorAuthor

    @andygrove there seems to be a flaky tests failing CI. It depends on the order so some can survive and others fail.

  2. dwsmith1983 commented on Nov 18, 2026

    @dwsmith1983
    ContributorAuthor

    take

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

Metadata

Metadata

Assignees

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions