Last modified: 2014-09-05 14:59:43 UTC
Something has happened in the past days that has made the column "Update time for pending reviews waiting for reviewer in days" at http://korma.wmflabs.org/browser/gerrit_review_queue.html show numbers clearly wrong. TorBlock shows a median of 2085 days, which is simply impossible. The two open reviews were submitted/updated about eleven months ago, so the median should be around 330 days. Parsoid shows two open reviews but then 0 median time, also impossible. What is also weird is that even with 0 is listed in the #4 position. DonationInterface shows 2514 (also impossible) and it is listed after "tools"with 684 (also not true, but a smaller number than 2514). In other words, that column seems to be completely broken now, even if the classification is still relevant somehow.
Working on it. Thanks for the detection Quim.
Quim, the problem is related with the update of messages in this page. I think you worked in an HTML not in sync with master, and once I accepted your pull request (my fault not to review it with more care), the problem was introduced. Also, for this metric before having our actual testing procedures, we normally compare visually all graphs to be sure about the data. But in this case, the shape of the graph was right, but the scale not. A sum of issues! Problem fixed now in production.
(In reply to Alvaro from comment #2) > Quim, the problem is related with the update of messages in this page. I > think you worked in an HTML not in sync with master, and once I accepted > your pull request (my fault not to review it with more care), the problem > was introduced. After thinking more about where the issue was introduced, I think it is not related with the new strings, but with the removal of a minichart. The minichart metric was removed, but not the options for it (start date, aggregate result), so the options for this minichart removed were applied to the "Update time for pending reviews waiting for reviewer in days", including the aggregation of data. So the time was aggregated and this fact introduced the error. > > Also, for this metric before having our actual testing procedures, we > normally compare visually all graphs to be sure about the data. But in this > case, the shape of the graph was right, but the scale not. > > A sum of issues! > > Problem fixed now in production.
Ok, I'm sorry. I shouldn't send patches going beyond edits in text strings without testing them in an environment. Do you know why Parsoid appears at #2 with 0 days? Same for #12 SmashPig, #13 gerrit.wikimedia.org_integration_docroot, #16 gerrit.wikimedia.org_analytics_kraken, and maybe more. Except this problem, now the ranking and the numbers looks sane. Thank you.
Also, DataValues is now the #1 repo, but I can't find any repo by this name, neither open reviews in the three related projects: https://gerrit.wikimedia.org/r/#/admin/projects/?filter=DataValues I realized that there is a https://github.com/DataValues repo (that has no open pull requests). Maybe DataValues in Gerrit was an old repo that was closed with two open changesets? CCing Jeroen.
Any idea of what is happening with DataValues? Also, projects like Parsoid, SmashPig, and gerrit.wikimedia.org_integration_docroot appear in the top positions, but declaring 0 days of median age.
DataValues still has a copy on Gerrit?! This stuff was moved to GitHub ages ago https://github.com/DataValues/
(In reply to Jeroen De Dauw from comment #7) > DataValues still has a copy on Gerrit?! Was it ever requested somewhere to remove it from WM Git/Gerrit? References welcome.
The problem with (In reply to Quim Gil from comment #4) > Ok, I'm sorry. I shouldn't send patches going beyond edits in text strings > without testing them in an environment. > > Do you know why Parsoid appears at #2 with 0 days? Same for #12 SmashPig, > #13 gerrit.wikimedia.org_integration_docroot, #16 > gerrit.wikimedia.org_analytics_kraken, and maybe more. The problem was that we are not filtering the WIP reviews in some metrics. We have fixed that and now this problem is fixed! > > Except this problem, now the ranking and the numbers looks sane. Thank you. Great!
(In reply to Jeroen De Dauw from comment #7) > DataValues still has a copy on Gerrit?! This stuff was moved to GitHub ages > ago https://github.com/DataValues/ No it doesn't have a copy in Gerrit, and this is probably why the data got stuck somehow. Mmm... if a Key Wikimedia Project is not using the Wikimedia infrastructure, then I think it is safe to assume that we can skip it in our metrics. Hopefully this case will be a single exception. Álvaro, we can resolve this report as soon as DataValues has been removed from the list.
Great, we will remove it tomorrow and it will disappear from the dashboard. Thanks!
The project DataValues does not appear anymore!
This is solved now. Thank you.