[19:43:18] Amir1: so.. why does ShortPages need a union anyway? [19:43:42] Krinkle: in case of many content ns [19:43:48] I'm struggling to see the difference between a union of three clones that only add a where clause, and a single query with a combined where clause (IN or OR) [19:43:49] which is quite rare tbh [19:44:23] I know why actually [19:44:34] I think mariadb wasn't very smart back then [19:44:40] so the query was rewritten into union [19:44:59] https://phabricator.wikimedia.org/T168010 [19:45:57] a simpler approach to make it a where condition if unionSupportsOrderAndLimit is not true [19:46:16] so production stays as unions and sqlite becomes where() [19:46:26] could do yeah. [19:46:48] awesome [19:46:56] might be worth running the query that would exist if we revert https://gerrit.wikimedia.org/r/c/mediawiki/core/+/359502/ which added the override. and try that on enwiki and commonswiki by hand or something. [19:47:23] anyway, Im fine with the dual version as well if you think it's unlikely to be fast in prod [19:48:06] my understanding of indexes must be wrong as I don't get why a force index didn't fix this. It seems we used to force page_redirect_namespace_len here and that didn't work either? [19:52:58] oh and yeah, we now exclude NS_FILE on Commons [19:53:00] so that's changed as well [19:57:10] "joy" [22:15:27] MatmaRex: If you have time already for MW stuff and interested in diving in MW-Statsd/MW-BagOStuff, I could use some ideas on where to take T338189 next. [22:15:27] T338189: WANCache stats missing for 'CirrusSearchParserOutputPageProperties' - https://phabricator.wikimedia.org/T338189 [22:22:40] Krinkle: not really D: also i don't think i have access to the things (and not sure if i want it) [22:24:19] ack, I confirmed the shell stuff already so this would be about debugging the mw code or coming up with hypothesis, not tracing it on prod shell. though if you think it might need more tracing, I am curious where you think I could look next. In any case, no problem.