On Mon, 2023-06-19 at 18:28 +0200, Adam Williamson wrote:
Hey folks! So, as per https://pagure.io/fesco/issue/3011 , we plan to enable gating of Rawhide updates on openQA test results on Wednesday.
This is now enabled! Please let me and/or releng know if it causes any terrible consequences, and in the worst case, we'll turn it off again.
I did notice while doing final checks that we still have a bit of a result display inconsistency. If you check https://bodhi.fedoraproject.org/updates/FEDORA-2023-3c2931ff23 - the bad sqlite update - you will see that Bodhi/Greenwave still think several tests are "running". They are actually cancelled.
This is because they're jobs that depend on other tests that failed; when that happens, openQA cancels the dependent jobs, but does not publish any event about it in its internal event messaging system. Our system for publishing Fedora messages and reporting results to resultsdb triggers off those internal events, so because we don't get an internal openQA "job_done" or "job_cancel" message, we never send any Fedora messages or file a failed/cancelled result to ResultsDB, so Greenwave and hence Bodhi don't know the job has been cancelled. They'll just think it's 'queued' forever.
I'm working on a fix for this upstream now. The issue doesn't have any particularly terrible consequences - any time we land in this situation, we'll have the record that the *parent* job failed, so the update will be gated as it should be - but it does look a bit confusing in Bodhi.
On Thu, Jun 22, 2023 at 12:27 PM Adam Williamson adamwill@fedoraproject.org wrote:
On Mon, 2023-06-19 at 18:28 +0200, Adam Williamson wrote:
Hey folks! So, as per https://pagure.io/fesco/issue/3011 , we plan to enable gating of Rawhide updates on openQA test results on Wednesday.
This is now enabled! Please let me and/or releng know if it causes any terrible consequences, and in the worst case, we'll turn it off again.
🎉🎉🎉
Thanks a lot for your work on this, Adam.