Last modified: 2014-09-11 09:33:56 UTC
Until now gerrit time to review for open reviews show the time from submission and also from last update. The metric asked by Wikimedia is the time to review from last uploaded patch.
Metric implemented and added to korma panel: http://korma.wmflabs.org/browser/gerrit_review_queue.html The list of repositories is now ordered using this new metric.
(In reply to Alvaro from comment #1) > Metric implemented and added to korma panel: > > http://korma.wmflabs.org/browser/gerrit_review_queue.html Good, thank you! You can remove the line for "Most recent updates" in "Age of open changesets", since it is not really relevant. > The list of repositories is now ordered using this new metric. Maybe this is something temporary, but TorBlock appears as #1 when (according to the data shown) it shouldn't.
How complex would it be to compute "Date of submission of open changesets" by the date of the last upload, instead of the initial upload? If it's not too complicated, the data based on last upload would be more useful, because we could clearly see the old stuff that is there, just rotting.
(In reply to Quim Gil from comment #2) > (In reply to Alvaro from comment #1) > > Metric implemented and added to korma panel: > > > > http://korma.wmflabs.org/browser/gerrit_review_queue.html > > Good, thank you! You can remove the line for "Most recent updates" in "Age > of open changesets", since it is not really relevant. > Removed. > > > The list of repositories is now ordered using this new metric. > > Maybe this is something temporary, but TorBlock appears as #1 when > (according to the data shown) it shouldn't. Let me check ... TorBlock data: http://korma.wmflabs.org/browser/data/json/gerrit.wikimedia.org_mediawiki_extensions_TorBlock-scr-rep-static.json has the "review_time_pending_upload_days_median": 495.29124999999999" used for ordering. And it is the higher one. I need to cross check it with evolution in time. We will updated this ticket once we have a detailed answer.
(In reply to Quim Gil from comment #3) > How complex would it be to compute "Date of submission of open changesets" > by the date of the last upload, instead of the initial upload? If it's not > too complicated, the data based on last upload would be more useful, because > we could clearly see the old stuff that is there, just rotting. In this graph we are showing the distribution of current open changesets using the submission time. We can play with it to use the upload time, yes. Let's open a new ticket to follow it. https://bugzilla.wikimedia.org/show_bug.cgi?id=70650
(In reply to Alvaro from comment #4) > Let me check ... TorBlock data: > > http://korma.wmflabs.org/browser/data/json/gerrit.wikimedia. > org_mediawiki_extensions_TorBlock-scr-rep-static.json > > has the "review_time_pending_upload_days_median": 495.29124999999999" used > for ordering. And it is the higher one. This is not what Gerrit says: https://gerrit.wikimedia.org/r/#/q/status:open+project:mediawiki/extensions/TorBlock,n,z Only one patch waiting for review, with a las version uploaded on 21 August 2014 (only 20 days ago) https://gerrit.wikimedia.org/r/#/c/62026/
(In reply to Quim Gil from comment #6) > (In reply to Alvaro from comment #4) > > Let me check ... TorBlock data: > > > > http://korma.wmflabs.org/browser/data/json/gerrit.wikimedia. > > org_mediawiki_extensions_TorBlock-scr-rep-static.json > > > > has the "review_time_pending_upload_days_median": 495.29124999999999" used > > for ordering. And it is the higher one. > > This is not what Gerrit says: > > https://gerrit.wikimedia.org/r/#/q/status:open+project:mediawiki/extensions/ > TorBlock,n,z > > Only one patch waiting for review, with a las version uploaded on 21 August > 2014 (only 20 days ago) > > https://gerrit.wikimedia.org/r/#/c/62026/ Yes, we have fixed a problem in time series for upload (aggregated information used for ordering was right). We are rechecking the logic to detect that a review is waiting for Submitter. Right now rechecking the data about TorBlock.
(In reply to Alvaro from comment #7) > (In reply to Quim Gil from comment #6) > > (In reply to Alvaro from comment #4) > > > Let me check ... TorBlock data: > > > > > > http://korma.wmflabs.org/browser/data/json/gerrit.wikimedia. > > > org_mediawiki_extensions_TorBlock-scr-rep-static.json > > > > > > has the "review_time_pending_upload_days_median": 495.29124999999999" used > > > for ordering. And it is the higher one. > > > > This is not what Gerrit says: > > > > https://gerrit.wikimedia.org/r/#/q/status:open+project:mediawiki/extensions/ > > TorBlock,n,z > > > > Only one patch waiting for review, with a las version uploaded on 21 August > > 2014 (only 20 days ago) > > > > https://gerrit.wikimedia.org/r/#/c/62026/ > > Yes, we have fixed a problem in time series for upload (aggregated > information used for ordering was right). We are rechecking the logic to > detect that a review is waiting for Submitter. > > Right now rechecking the data about TorBlock. Yes, in the SQL we were not using the min upload time. It is fixed and regenerating the metrics now. It is hard to include all the logic in SQL that we are reaching. I hope that with this fix the resulting metric is right. Once the new metrics are available I will recheck them with several repositories to be sure about them.
Quim, we have updated the JSON data for: http://korma.wmflabs.org/browser/gerrit_review_queue.html Now the upload time is in general lower than before, and the data is more accurate. You can consult TorBlock data at: http://korma.wmflabs.org/browser/repository.html?repository=gerrit.wikimedia.org_mediawiki_extensions_TorBlock
I have checked exhaustively the current top 10 at "Ranking of repositories" and the numbers make sense now. Thank you!