[CI Failure] Flaky: AspireRabbitMQLoggingTests.TestExceptionWithoutInnerException fails with empty collection #18801

Activity

1. github-actions

   added

   test-failure

   ci-failure-cause

   on Jul 15, 2026
2. radical commented on Aug 29, 2026

   

   Ankit Jain (radical)

   on Aug 29, 2026

   Member

   More actions

   [automated] The CI shepherd is watching this failure.

   Current assessment: Watch this single-test-occurrence for another independent failure of the same test on a different day.

   Evidence reviewed:

   • issue:18801
   • run:29456677302
   • pr:18799

   Evidence still needed:

   • verified-fix-or-current-recurrence-check

   Reassess when: After another independent failure of the same test on a different day or positive recovery.

   No quarantine, retry, closure, or investigation has been started.
3. davidfowl commented on Oct 31, 2026

   

   David Fowler (davidfowl)

   on Oct 31, 2026 – with Copilot SWE Agent

   Collaborator

   More actions

   I suggest closing this as obsolete. The failing leg no longer exists, and there is no recorded or reproducible recurrence on the current code.

   Evidence

   • The failing leg was removed. The only occurrence (run 29456677302, 2026-07-15) was in Tests / RabbitMQ.Client.v6 (windows-latest), which ran the shared AspireRabbitMQLoggingTests against RabbitMQ.Client 6.x through the #if RABBITMQ_V6 code paths. Remove Aspire.RabbitMQ.Client.v6 and Aspire.MongoDB.Driver.v2 #20608 (9365348, 2026-09-30) deleted the Aspire.RabbitMQ.Client.v6 package, its test project, and all RABBITMQ_V6 branches. The remaining test runs only against RabbitMQ.Client 7.2.2.
   • The original diagnostics are gone. The job log and test-result artifacts for run 29456677302 have expired (GET .../jobs/87491646494/logs returns HTTP 410). The CI shepherd's watch on this issue recorded no second occurrence in the 3.5 months since.
   • No local reproduction on current main (8487570): 40 consecutive bounded runs of the full Aspire.RabbitMQ.Client.Tests suite (net10.0 and net11.0, 98 tests per run) all passed, which is 80 executions of TestExceptionWithoutInnerException. Only the Docker-dependent EndToEndLoggingTest was excluded, because this environment has no Docker.

   Code analysis. I looked for a shared-code cause that could still affect v7. The one candidate is a theoretical construction race in RabbitMQEventSourceListener:

   • The base EventListener constructor releases EventListenersLock before the derived constructor assigns _log and replays _eventSources.
   • An EventSource first created on another thread inside that window could be queued after the replay and never enabled.

   The window is only a few instructions wide, though. A 2000-iteration stress test that raced Start() against creating a fresh rabbitmq-client EventSource missed 0 events. And TestExceptionWithoutInnerException itself first touches the event source after Start(). So I can't attribute this failure to that race, and I'm not proposing a speculative change.

   If the test fails again on v7, please reopen with the new run and I'll take it from there.

   🤖 Posted by AspireAgent (AI Agent)
4. davidfowl commented on Nov 1, 2026

   

   David Fowler (davidfowl)

   on Nov 1, 2026 – with Copilot SWE Agent

   Collaborator

   More actions

   Closure recommendation: no recurrence in 7 days of CI

   • Window: ci.yml attempt-1 runs from 2026-10-25T10:57Z to 2026-11-01T09:51Z (824 runs). I grepped every failed test-job log in that window (968 of 968 downloaded).
   • Test: AspireRabbitMQLoggingTests.TestExceptionWithoutInnerException. It's a plain [Fact], still present on main (1bccada).
   • RabbitMQ.Client lane jobs: main 63 Linux / 63 Windows, PR 70 Linux / 70 Windows, plus 4+4 RabbitMQ.Client.v6 jobs.
   • Failures: 0.
   • Last recorded occurrence: 2026-07-15, run 29456677302.
   • Fixing PR: none identified.

   I recommend closing this. Reopen it if the signature comes back.

   🤖 Posted by AspireAgent (AI Agent)