Summary
The "Final summary" emitted by the Helix Job Monitor at the end of a run reports work item counts that don't match what was actually processed, and the failed count appears to undercount failures.
Observed behavior
The periodic status line correctly reports:
info: ℹ️ Status: 6 processed / 6 completed / 0 running / 0 waiting jobs
669 processed / 669 completed / 0 running / 0 waiting work items
But the final summary reports:
info: 📊 Final summary:
Jobs: 6 submitted / 0 resubmitted / 6 processed
Work items: 236 submitted / 0 resubmitted / 1 failed
Issues:
-
Work item count mismatch: The status line shows 669 work items were processed, but the final summary says 236 were submitted. It's unclear what "236 submitted" refers to vs. the 669 that were actually tracked and completed.
-
Failed count appears incorrect: The summary reports 1 failed, but the AzDO Test tab shows test failures from at least two different legs (TestBuild: windows x64 and FullFramework: windows x64), which must correspond to at least 2 failed work items. Additionally, only one failed work item console log is listed in the ❌ Failed work item console logs section.
Expected behavior
- The final summary's work item counts should be consistent with the status line counts (669 in this case).
- The failed count should accurately reflect the total number of work items that exited with a non-zero exit code across all jobs.
Context
Observed on dotnet/sdk PR #54583, build https://dev.azure.com/dnceng-public/public/_build/results?buildId=1451937
Monitor tool version: 11.0.0-beta.26303.111
Summary
The "Final summary" emitted by the Helix Job Monitor at the end of a run reports work item counts that don't match what was actually processed, and the failed count appears to undercount failures.
Observed behavior
The periodic status line correctly reports:
But the final summary reports:
Issues:
Work item count mismatch: The status line shows 669 work items were processed, but the final summary says 236 were submitted. It's unclear what "236 submitted" refers to vs. the 669 that were actually tracked and completed.
Failed count appears incorrect: The summary reports
1 failed, but the AzDO Test tab shows test failures from at least two different legs (TestBuild: windows x64 and FullFramework: windows x64), which must correspond to at least 2 failed work items. Additionally, only one failed work item console log is listed in the❌ Failed work item console logssection.Expected behavior
Context
Observed on dotnet/sdk PR #54583, build https://dev.azure.com/dnceng-public/public/_build/results?buildId=1451937
Monitor tool version:
11.0.0-beta.26303.111