[08:42:25] I thought the deepcat limit was around 8000 category... but it's around 256 actually, mainly because we run a bool wrapping match queries, we could perhaps increase that limit and run a "terms" query, but that requires switching to a "normalized" keyword field [08:42:45] probably quite some work to make it happen but not impossible [08:48:50] FYI a couple of cumin aliases are reporting not matching any host anymore, maybe some puppet refactoring that require to update them? are the wdqs-scholarly and wdqs-main aliases, respectively matching on puppet role wdqs::scholarly and wdqs::main [08:49:59] volans: I think this is waiting for https://gerrit.wikimedia.org/r/c/operations/puppet/+/1054342? (cc stevemunene, ryankemper) [08:50:57] ah so the aliases have been put in advance of the clusters, got it [08:51:26] Indeed, sorry for the noise [08:54:40] we have a check that ensures that all aliases match hosts and all hosts are matched by sum of all the A:DC_NAME aliases for all DCs [08:55:06] also because if an alias doesn't match any host if you do A:foo or A:bar and A:foo is empty it fails [09:09:47] We can temporarily remove the aliases and re add them later on. [09:11:52] if it's just a couple of days no worries, if it's 3 weeks maybe better :) [10:13:03] lunch [12:32:33] ebernhardson: NICE! [12:46:07] Did anyone else get an email from gitlab saying `Your Group access token sonarcloud will expire in 7 days or less` ? That was 5 days ago so LMK if I need to take action [12:53:19] o/ [12:54:06] inflatador: yes, unsure what we need to do about it, looking at the token it says that it was never used... [13:05:51] dcausse thanks. I guess I'll reach out to releng. Sounds like we don't need it though [13:18:16] \o [13:23:07] o/ [14:31:47] * ebernhardson notices we are still using DeprecationLoggedHttp, guess it doesn't hurt anything [14:47:01] quick break, back in ~10 [15:04:10] ryankemper we're in SRE standup if you are around [15:21:37] dcausse: can meet in 5' (or later) if you're around [15:22:00] ryankemper: in 5min is great! [15:23:09] sent a quick invite [15:23:39] feel free to move it some time later if that does not work [16:05:02] been back, but now working out, back in ~40 [17:06:39] back [17:31:10] dinner [18:03:01] lunch, back in ~40 [18:07:05] * ebernhardson mutters at presto and it being awkward with struct fields for some reason [18:15:36] dr0ptp4kt: to be clear, on T370662 when you say slice do you mean groupby, and a line per thing, or a filter? [18:15:37] T370662: Slice Search metrics dashboards by project / language - https://phabricator.wikimedia.org/T370662 [18:21:54] ebernhardson: guillaume wrote that. but, it feels like a filter is best, at least assuming that it isn't dramatically more effort. it seems like the charts would get too busy otherwise. [18:24:40] dr0ptp4kt: alright, makes sense. I was also thinking filter, but then i read your comment about not having separate lines for logged in /logged out and wondered if i was thinking something different [18:30:08] ebernhardson: oh, right. yeah, for that i can see how we'd actually be okay doubling the number of lines. even there, though, i suppose we could have a filter. like you had said, we may not always have the data available for the logged in /logged out determination...and i should split that logged in / logged out stuff to a different task anyway. that's a separate thing, but i see why you ask. [18:40:25] * ebernhardson wishes superset would give the full failing query instead of the limited `Column \'normalized_host."project"\' cannot be resolved` when every involved table has that [18:47:44] suspecting the answer is superset doesn't understand how escaping is supposed to work... [19:11:42] * ebernhardson sometimes wishes things were still implemented as forms where i could hand-edit the controls and submit the corrected values... [20:00:14] anyone from Search going to Wikimania? [20:02:02] not this year. you? [22:08:54] see ya tomorrow!