[unknownhackerworld]

Issue number: closes #8645

Summary

Changes

The half-open probe success streak lives in the module-level _LOCAL_SUCCESSES dict in aws_lambda_powertools/utilities/circuit_breaker/base.py. Before this change it was reset in only two places:

• _call_probe, when a probe fails in this environment
• _call_probe, when this environment reaches success_threshold and closes the circuit

Nothing reset it when a worker won a new probe election in handle(). So a streak from an earlier recovery window carried into the next one. With success_threshold=3:

1. Env A wins the election and gets 2 successful probes.
2. Env A goes idle, its probe lease lapses, env B takes over, B's probe fails and the circuit reopens. A's counter is still 2, because B runs in a different execution environment.
3. In the next window env A wins the election again, and its first success brings the count to 3 and closes the circuit.

That breaks the documented rule ("If success_threshold consecutive probes succeed, the circuit closes") and sends full traffic back to a downstream that just failed a probe.

The fix adds a small _reset_probe_successes() helper, which sets _LOCAL_SUCCESSES[name] = 0 under _COUNTERS_LOCK. handle() calls it right after try_acquire_half_open succeeds, on both paths:

• the OPEN path (recovery_timeout elapsed, fresh election)
• the HALF_OPEN expired-lease takeover path

The thread that already owns the lease skips the election (record.half_open_owner == _probe_owner_id()), so a streak within one window is unchanged. Only a newly won election starts from zero. This matches how _LAST_OBSERVED_STATE already throws away a stale failure streak after an external close (test_stale_local_failures_reset_on_external_close).

The lock is held only for the dict write, never across persistence I/O or callbacks, as the existing comment on _COUNTERS_LOCK requires.

Tradeoffs

• On a lease takeover, the new owner now needs the full success_threshold successes and no longer inherits the old owner's partial streak. That means a few more probes before closing. It follows the "correctness over recovery latency" choice from fix(circuit_breaker): make probe election per-thread and synchronize local counters #8323, which noted the inherited streak as pre-existing and left it unchanged.
• No changes to the persistence schema, the DynamoDB layer, the public API or the docs. The docs already describe the intended behaviour.

Alternatives considered

• Reset the success streak whenever this env sees OPEN or CLOSED in handle(), like _LAST_OBSERVED_STATE does for failures. This misses the lease takeover path, where the record goes from HALF_OPEN (old owner) to HALF_OPEN (new owner) without passing through OPEN or CLOSED. Tying the reset to winning the election covers both paths in one place.
• Persist the success count in the store. This would make the count shared across environments, but it adds a write per probe and changes the record schema. That is a much bigger change for the same guarantee.

User experience

Before (success_threshold=3, after the scenario above):

call()  # first probe of the new recovery window
store.db["c"].state  # CLOSED after one probe

After:

call()  # first probe of the new recovery window
store.db["c"].state  # HALF_OPEN, 1 of 3
call(); call()
store.db["c"].state  # CLOSED after 3 consecutive successes

Single-environment functions that never lose their lease behave the same as before.

Tests

Two new regression tests in tests/functional/circuit_breaker/test_circuit_breaker.py, next to the existing bug regressions:

• test_new_probe_election_does_not_reuse_successes_from_an_earlier_recovery_window: 2 successes, then an external reopen, then a new election. The new window needs its own 3 successes.
• test_probe_lease_takeover_starts_a_fresh_success_streak: a takeover of an expired lease in an environment that has a leftover streak of 2. The first success counts as 1 and the circuit stays HALF_OPEN.

Both fail on develop (the circuit closes after one probe) and pass with this change. Locally:

• pytest tests/functional/circuit_breaker tests/unit/circuit_breaker: 85 passed
• ruff check and ruff format --check on the changed files: clean
• mypy aws_lambda_powertools/utilities/circuit_breaker/base.py: no issues

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

Disclaimer: We value your time and bandwidth. As such, any pull requests created on non-triaged issues might not be successful.

[boring-cyborg]

Thanks a lot for your first contribution! Please check out our contributing guidelines and don't hesitate to ask whatever you need.
In the meantime, check out the #python channel on our Powertools for AWS Lambda Discord: Invite link

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[leandrodamascena]

Thanks @unknownhackerworld, this is great work! I ran it through the same multi-environment setup as yesterday, including 4 processes against a real DynamoDB table, and the circuit never closed early anymore, while develop still did. The two tests cover exactly the two election paths. Approving now, this will wait here a bit until the release, no need to rebase or touch anything, and you'll get the credit in the release notes. Thank you!

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.93%. Comparing base (d82a695) to head (f586a9e).

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #8646   +/-   ##
========================================
  Coverage    96.93%   96.93%           
========================================
  Files          318      318           
  Lines        16031    16036    +5     
  Branches      1423     1423           
========================================
+ Hits         15540    15545    +5     
  Misses         349      349           
  Partials       142      142           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[boring-cyborg]

Awesome work, congrats on your first merged pull request and thank you for helping improve everyone's experience!

[powertools-for-aws-oss-automation]

Awesome work, congrats on your first merged pull request and thank you for helping improve everyone's experience!