[09:49:04] lunch [13:14:55] o/ [13:20:31] dcausse I one-offed wdqs-categories1001 so it can start wdqs-categories service. I still need to update the reload cookbook to work with that hostname, but we should be able to get a reload started today. If you have anything in particular you want me to check during the reload LMK [13:22:31] inflatador: nice! were you able to provision the host with a single blazegraph instance running? [13:56:52] \o [14:01:52] o/ [14:03:59] dcausse yes...it's all one-offed which isn't great, but I didn't want to deal with puppet so I did some rsyncing [16:15:43] lunch, back in ~1h [16:24:08] dinner [17:06:21] * ebernhardson finds it mildly amusing there are separate disambiguation pages for SERP and SERPS [17:06:39] of course, mostly because i thought it was a single thing and of course there are others with the same acronym :P [17:23:07] back [17:49:44] found a minor problem with abandonment rate, we have a problem where not all expected click events are recieved. In the abandonment rate we only consider clicks, but if we also consider if the sessions had a satisfied page visit the numbers for an example hour drop from 209k to 192k. Not huge, still gives 48% abandon, but a minor correction [17:50:12] (to that end, in that hour we recorded 198k sessions with a satisfied visit, but only 189k sessions with a clickthrough) [17:50:33] err, it's a day not an hour [17:51:54] from 52.5% -> 48.3%. Might be worthwhile to better understand why we don't get all the click events, but might be immaterial [18:09:16] and then some bad news, some autocomplete data was sneaking into the fulltext_sat_visit calculation. Redoing the calculations with the fixed numbers brings it back to 52.4% :P [18:21:20] which also means, we don't graph it anywhere so it's not visible, but sat visits are lower than previously estimated [19:05:54] ebernhardson: do you know/remember which jenkins pipeline it is to build the wdqs release these days? the most likely one seems to be https://integration.wikimedia.org/ci/job/wikidata-query-rdf-maven-release/ but i can't tell if that's just for building streaming updater stuff [19:06:08] it used to be https://integration.wikimedia.org/ci/blue/organizations/jenkins/wikidata-query-rdf-maven-release-docker/ FWIW but that doesn't exist anymore [19:08:12] hmm, i always found it in the documentation. sec [19:09:06] ryankemper: i suspect https://integration.wikimedia.org/ci/job/wikidata-query-rdf-maven-release/ ? but thats from my browser history [19:09:18] last build is yesterday, which is promising :) [19:10:11] ok good was thinking so but started second guessing :D [19:10:12] ty [19:11:11] as for where -docker went, usually i would go spelunking in the integration/config, but unfortunately unless someone mentioned it in their commit message it's tedious to figure out [19:13:27] apparently in april -docker suffix was removed from everything: https://gerrit.wikimedia.org/r/c/integration/config/+/1016827 [19:14:39] Might be revealing my own ignorance, but TIL where all the jenkins config lives ;) [19:15:22] inflatador: there is a 'jjb' directory in that repo, has all the fun stuff [19:15:55] also looks like the documented link works: https://wikitech.wikimedia.org/wiki/Wikidata_Query_Service#Initial_Preparation [19:55:27] * ebernhardson wonders about how to nicely move hive schemas into discolytics, it's a bit tedious how many different repos i have to commit to for the virtualpageview...but then we need to consider hive migrations in discolytics which sounds not fun :P