I checked a few of these at random. They appear to have been moved from one of the valid namespaces.

Good find. Thank you for that MPGuy2824.

Here's a query of the counts of all pages in pagetriage_pages over one month since ptrp_reviewed_updated was last touched, excluding draftspace. Why one month? This is a common setting for some of the PageTriage delete cron jobs.

https://quarry.wmcloud.org/query/68541

And a list of pages, excluding draftspace.

https://quarry.wmcloud.org/query/68542

Some of these are not from page moves, but many are.

I think these are just cluttering up the table. They are skipped by our normal cleanup cron jobs and just swell the size of the SQL tables. I think the path forward is...

• Change move hook to delete from pagetriage_page and pagetriage_page_tags when moving a page to an unsupported namespace
• Ask DBAs to do a one time deletion of these entries, probably by running maintenance/cleanupPageTriage.php
• Keep an eye on this / query this after the DBA deletion. If it starts swelling again, track down the offending code and fix it. Rinse and repeat.

Looks like the # of buggy rows increased by 600 in 1 month.

I think there's two bugs here:

1. Hooks.php -> onPageMoveComplete, line 95 should check if page is being moved out of a namespace PageTriage patrols. If so, run $pageTriage->deleteFromPageTriage()
2. cron job at cron/updatePageTriageQueue.php should delete from all namespaces after 30 days but doesn't. The bug is on line 77, where the $secondaryNamespaces variable is declared. Despite a comment stating otherwise, it contains an incomplete list of namespaces.

I think I'll refactor updatePageTriageQueue.php, then we can squash the two bugs with small patches.

Change 865581 had a related patch set uploaded (by Novem Linguae; author: Novem Linguae):

[mediawiki/extensions/PageTriage@master] [WIP] Refactor updatePageTriageQueue.php

https://gerrit.wikimedia.org/r/865581

Change 868635 had a related patch set uploaded (by Novem Linguae; author: Novem Linguae):

[mediawiki/extensions/PageTriage@master] Extract cron job class into its own file

https://gerrit.wikimedia.org/r/868635

Change 868635 merged by jenkins-bot:

[mediawiki/extensions/PageTriage@master] Extract cron job class into its own file

https://gerrit.wikimedia.org/r/868635

Change 865581 abandoned by Novem Linguae:

[mediawiki/extensions/PageTriage@master] [WIP] Refactor updatePageTriageQueue.php

Reason:

I ended up splitting this into smaller patches

https://gerrit.wikimedia.org/r/865581

Note that there appears to be a maintenance script that cleans these up: maintenance/cleanupPageTriage.php. But I still think it'd be best to fix the root causes. See above for the 2 bugs I suspect are causing this.

Hi @Novem_Linguae , has this bug been fixed ?

In T321982#9675185, @Destiny wrote:

Hi @Novem_Linguae , has this bug been fixed ?

Nope. Not yet. I've updated the task to be more accurate about what needs to be fixed.