Skip to content

Flaky static exec integration test closes broker association during teardown #1412

Description

@wdcui

Summary

The Ubuntu Build and Test job for #1411 intermittently failed in litebox_runner_linux_userland::run::test_static_exec_with_rewriter because the runner closed the active broker association. The unchanged commit passed local reproductions and a rerun of the same workflow, suggesting a runner/broker association teardown race rather than a deterministic regression.

Failure evidence

The failure surfaced during different inner C programs across retries. The first attempt reached thread_exit_exec_rewriter; later attempts failed around execve_exec_rewriter. During an exec retry, the guest printed:

[OK] exec caller became leader; cloexec fd 3 closed; keep fd 4 open

The broker command nevertheless exited with status 1 afterward, which points toward association shutdown/teardown rather than the guest validation itself.

Reproduction attempts

The failure did not reproduce locally on the same commit:

cargo nextest run -p litebox_runner_linux_userland --test run test_static_exec_with_rewriter --no-fail-fast
PASS in 24.317s

cargo nextest run -p litebox_runner_linux_userland --test run --no-fail-fast
15/15 passed; affected test passed in 12.993s

cargo nextest run --no-fail-fast
977/977 passed; affected test passed in 13.353s

Successful unchanged rerun

The failed GitHub Actions job was rerun without any code changes:

We should track the runner/broker association teardown path and improve diagnostics so a future occurrence identifies which side closes the channel and why.

Activity

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

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions