test(internal): add lifecycle stress / fuzzing harness for PeriodicThread

[r1viollet]

Description

Adds tests/internal/test_periodic_stress.py — a randomized lifecycle
fuzzer for PeriodicThread that interleaves
create / start / stop / join / awake / drop-last-ref / fork / gc /
thread-churn across a pool of PeriodicThread and PeriodicService
objects.

This is the harness that surfaced #17707 and #18040
(PeriodicThread.awake() blocking forever after stop()).

What's in the file

• test_periodic_thread_lifecycle_stress — the randomized fuzzer.
  Default budget kept short for CI (DD_STRESS_ITERS=200, ~10s).
  Documented env vars for soaks (DD_STRESS_ITERS, DD_STRESS_SECONDS,
  DD_STRESS_SEED, DD_STRESS_POOL, DD_STRESS_FORK_EVERY,
  DD_STRESS_TRACE_FILE). Seed is printed on every run so a failing
  random seed can be pinned with DD_STRESS_SEED=N.
• test_periodic_thread_concurrent_dealloc_race — focused regression
  for fix: periodic thread start race #17485 (refcount TOCTOU between std::thread creation and the
  lambda acquiring the GIL).
• test_periodic_thread_stop_without_join_then_fork_repeat — focused
  regression for fix: periodic thread after fork cleanup #16955 (pthread_t recycling after stop-without-join
  then fork).

How to run it when modifying the periodic thread framework

The module docstring documents the exact commands. tl;dr:

# ~30s soak
DD_STRESS_ITERS=10000 scripts/run-tests -- -- tests/internal/test_periodic_stress.py

# 120s wall-clock soak (recommended under ASan/TSan)
DD_STRESS_SECONDS=120 scripts/run-tests -- -- tests/internal/test_periodic_stress.py

# Reproduce a failing seed (the seed is printed to stderr by every run)
DD_STRESS_SEED=12345 scripts/run-tests -- -- tests/internal/test_periodic_stress.py

Testing

All three tests pass locally on Python 3.13 in the testrunner image:

tests/internal/test_periodic_stress.py::test_periodic_thread_lifecycle_stress PASSED
tests/internal/test_periodic_stress.py::test_periodic_thread_concurrent_dealloc_race PASSED
tests/internal/test_periodic_stress.py::test_periodic_thread_stop_without_join_then_fork_repeat PASSED

Risks

None — test-only addition. Existing test discovery and CI suites are
unchanged.

changelog/no-changelog — test-only.

🤖 Generated with Claude Code

[datadog-prod-us1-3]

 

✨ Fix all issues with BitsAI

⚠️ Warnings

🚦 2 Pipeline jobs failed

Changelog | Validate changelog     

🛟 This job is unlikely to succeed on retry. Please review your pipeline configuration.

Release note not found during changelog validation. Use 'reno new <slug>' to add a new note to 'releasenotes/notes'.

DataDog/apm-reliability/dd-trace-py | download win_arm64 wheels     

🛟 This job is unlikely to succeed on retry. Please review your pipeline configuration.

RUN_ID not found. Check if the GitHub build jobs were successfully triggered on your PR.

ℹ️ Info

No other issues found (see more)

🧪 All tests passed
❄️ No new flaky tests detected

🚧 81 tests that failed were ignored due to quarantine

Useful? React with 👍 / 👎

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: f853f3b | Docs | Datadog PR Page | Give us feedback!}