test: fix intermittent CI hang in the unit test matrix#1921
Open
WilliamBergamin wants to merge 3 commits into
Open
test: fix intermittent CI hang in the unit test matrix#1921WilliamBergamin wants to merge 3 commits into
WilliamBergamin wants to merge 3 commits into
Conversation
The "Unit tests (3.x)" matrix job intermittently hung on "Run tests" until its 15-minute timeout on a random Python version, passing on retry. Root cause was in the test infrastructure, not the SDK: - No per-test timeout, so a hung test gave a blind 15-min job timeout with no indication of which test stuck. - The shared mock web API server used a single-threaded HTTPServer on a non-daemon thread with unbounded join()/wait(). With HTTP/1.1 keep-alive it could leak a thread holding port 8888, hanging the next test's setUp. - Two socket-mode interaction tests leaked servers/clients: test_interactions_websockets omitted stop_socket_mode_server in tearDown and shared port 3001 with the aiohttp test; test_interactions_builtin had its per-iteration client.close() commented out, leaking thread pools. Fixes (test/CI only, no shipped SDK change): - Add pytest-timeout with a thread-method timeout so a hang fails fast with a full thread dump instead of stalling the job. - Make the mock web API server threads daemon + ThreadingHTTPServer, and bound join()/server_started.wait() so a failed bind fails loudly. Apply the same to the live duplicate in tests/rtm and drop the dead MockServerThread in tests/slack_sdk/web/mock_web_api_handler.py. - Fix the two leaking socket-mode tests: add the missing teardown, move off the shared port, and re-enable per-iteration client.close(). Co-Authored-By: Claude <svc-devxp-claude@slack-corp.com>
Follow-up to the CI-hang fix, addressing review findings: - Fix a shutdown race: stop() deleted self.server.queue before server.shutdown(), so under ThreadingHTTPServer an in-flight handler thread could hit AttributeError. Reorder to shutdown() -> join() -> del, after which no handler threads remain. - Unify stop()/stop_unsafe() into a single stop(). The split only existed because asyncio.Queue lacks .mutex; with the safe ordering it is unnecessary. Make the queue optional and drop redundant str(). - Warn when the server thread outlives join(timeout=5) instead of silently abandoning it, so a re-emerging leak is visible. - Dedupe tests/rtm/mock_web_api_server.py to reuse the canonical MockServerThread instead of a near-identical copy. - Wrap the async websockets tearDown in try/finally so the socket-mode server is always stopped even if mock-server cleanup raises. Co-Authored-By: Claude <svc-devxp-claude@slack-corp.com>
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #1921 +/- ##
==========================================
- Coverage 84.17% 84.14% -0.03%
==========================================
Files 118 118
Lines 13423 13423
==========================================
- Hits 11299 11295 -4
- Misses 2124 2128 +4 ☔ View full report in Codecov by Harness. |
WilliamBergamin
commented
Jul 23, 2026
| self._handle() | ||
|
|
||
|
|
||
| class MockServerThread(threading.Thread): |
Contributor
Author
There was a problem hiding this comment.
Keeps implementation DRY
WilliamBergamin
commented
Jul 23, 2026
| test.thread.start() | ||
| test.server_started.wait() | ||
| if not test.server_started.wait(timeout=5): | ||
| raise RuntimeError( |
Contributor
Author
There was a problem hiding this comment.
Fails with noise
WilliamBergamin
commented
Jul 23, 2026
| self._handle() | ||
|
|
||
|
|
||
| class MockServerThread(threading.Thread): |
Contributor
Author
There was a problem hiding this comment.
Keeps implementations DRY
WilliamBergamin
commented
Jul 23, 2026
| try: | ||
| cleanup_mock_web_api_server_async(self) | ||
| finally: | ||
| stop_socket_mode_server(self) |
Contributor
Author
There was a problem hiding this comment.
We were not always stopping the socket mode server I think this was leading to test hanging sometimes
WilliamBergamin
commented
Jul 23, 2026
| self.server = HTTPServer(("localhost", self.port), self.handler) | ||
| self.server.queue = self.queue | ||
| self.test.server_url = f"http://localhost:{str(self.port)}" | ||
| self.server = ThreadingHTTPServer(("localhost", self.port), self.handler) |
Contributor
Author
There was a problem hiding this comment.
The ThreadingHTTPServer is only available in python 3.7+, this is why we were using HTTPServer
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Fixes an intermittent hang in the Unit tests CI matrix that occasionally stalled the job until the workflow-level timeout.
Root cause: the tests' mock web API server ran on a single-threaded
HTTPServerwith a non-daemon thread, and teardown diddel self.server.queuebeforeserver.shutdown(). An in-flight request handler (whose first action isqueue.put_nowait(...)) could race the teardown, and a non-daemon thread that never joined could wedge the whole run.Changes (test infra only):
ThreadingHTTPServerand run it on adaemon=Truethread so a stuck worker can never block interpreter exit.shutdown()→join(timeout=5)→del queue, closing the delete-before-shutdown race (server_close()in the run loop'sfinallyjoins workers, so no handler touches the queue after the join).stop()/stop_unsafe()into a singlestop()that works for both the syncqueue.Queueand asyncasyncio.Queuepaths (the mutex is no longer needed), and make the queue optional so RTM can reuse the class.server_started.wait(timeout=5)now raises a clearRuntimeError(e.g. "port already in use") instead of blocking forever.MockServerThreadcopies intests/rtm/andtests/slack_sdk/web/; both now import the canonical implementation.tearDownintry/finallyso the socket-mode server is always stopped even if mock-server cleanup raises.client.close()intest_interactions_builtin.py(was commented out).pytest-timeout(timeout=180,timeout_method="thread") so a future hang fails fast with a thread dump instead of stalling CI. Thethreadmethod is required here becausesignalcannot interrupt hangs on non-main threads (asyncio/socket-mode).Testing
Reviewers can verify with:
Verified locally: full suite 988 passed, 5 skipped (4m17s) with the per-test timeout armed (no hang),
black --checkclean, and the async socket-mode test stable across 5 consecutive runs.Category
/docs(Documents)/tutorial(PythOnBoardingBot tutorial)tests/integration_tests(Automated tests for this library)Requirements
python3 -m venv .venv && source .venv/bin/activate && ./scripts/run_validation.shafter making the changes.