[00:21:56] * ebernhardson wonders if importing JvmOptionsParser (transforms jvm.options file into a CLI arg) from elastic would simplify blazegraph conf. both use apache 2.0 making it reasonably easy [09:01:27] ejoseph: meeting? https://meet.google.com/ukb-kgxq-gvq [09:05:16] ejoseph: ping me when you're around [09:07:34] gehel: internet [09:09:09] ejoseph: not sure I understand, are you having internet issues? [09:09:21] If that's the case, ping me when those are resolved [09:09:38] It's resolved [09:09:42] I am on the meet [09:10:04] I'm back in there as well! [09:31:19] Petrol scarcity here in Nigeria today, I can't power my Gen and there is no power supply from the power company since Monday. I would see if i can get a place to work [09:31:21] https://www.channelstv.com/2022/02/08/petrol-supply-with-excess-methanol-cause-of-fuel-scarcity-says-fg/ [09:32:12] :/ [09:33:04] Make una no too miss me sha (Pidgin for don't miss me much, I'll do what i can to be available soon) [09:44:51] good luck, hope everything goes back to normal soon [09:51:58] agreed that a jvm.options file would make this "options mess" a bit cleaner but I wonder where we would call that since the jvm will bootstrap jetty-runner [10:26:35] errand [10:29:53] dcausse, zpapierski: while reviewing https://docs.google.com/spreadsheets/d/1mLhLz0fqK8JaCssFM1mQDiftGNaCg5fnABTKgsAVNwQ/edit I got distracted and stumbled on https://www.rdfhdt.org/ [10:30:18] I tried it [10:30:51] it sadly required loading the full dataset in mem to be built [10:30:57] Would it make any sense to publish dumps in that format? One of the complain is that the dumps are huge and slow to download. [10:31:19] Ouch, that's a pretty big isse! [10:31:26] s/isse/issue/ [10:31:44] there's a hadoop version that I should try which I hope fix this problem [10:32:19] I looked at it for the categories RDF data and it was already too big for my machine :( [10:32:53] I don't think we have any server with 1T of RAM available :/ [10:39:37] And it seems to be limited to 2G triples [10:40:08] (using arrays internally to store the full S/P/O list [10:50:35] there are newer versions of it IIRC but not sure they're suited for very large dataset [10:51:39] It seems to me that a compact representation should be especially focused on large datasets. But I'm probably biased by our own use case. [10:53:40] yes, I guess it depends on what you consider "large" :) [11:12:08] lunch [11:17:03] lunch+errand [13:24:22] I am back guys [13:24:48] dcausse: Let me know when you are available [14:09:26] ejoseph: I'm around [14:11:22] Emmanuel - David [14:11:22] Wednesday, February 9 · 11:00 – 11:50am [14:11:22] Google Meet joining info [14:11:22] Video call link: https://meet.google.com/eek-ziau-djx [14:11:22] Or dial: ‪(FR) +33 1 87 40 43 41‬ PIN: ‪819 227 343‬# [14:11:22] More phone numbers: https://tel.meet/eek-ziau-djx?pin=8564650775234 [14:14:58] Greetings! I have to run to the doctor in ~30m , shouldn't be gone more than an hour [14:38:57] dcausse: is the 6.8 extra search plugin already ready? ( I'm seeing mixed things) context https://phabricator.wikimedia.org/T297002#7697247 [14:39:07] addshore: yes [14:39:41] but not at https://archiva.wikimedia.org/repository/releases/org/wikimedia/search/extra/ ? [14:40:43] addshore: no they're at maven central https://repo1.maven.org/maven2/org/wikimedia/search/extra/6.8.23-wmf1/ [14:40:50] thanks! [14:41:02] yw! [15:43:07] back [15:57:30] I wonder if we are OK building deb using a docker image [15:58:00] or if we can install openjdk on the machine building the deb package [15:59:00] dcausse we can take a look at that today if you like [15:59:16] you surely want mo.r.itz for this but on the build machine all debs are built in a chroot so the deps are installed inside it, it usually doesn't need deps installed on teh build host apart very special cases [15:59:40] inflatador: ok, thanks [16:00:05] volans: it requires openjdk as a dependency for compiling [16:00:09] Also thanks volans for the context [16:00:27] dcausse: and that happens during the build? [16:00:35] some related info https://wikitech.wikimedia.org/wiki/Git-buildpackage [16:01:03] it does things like: INCLUDE = -I"$(JAVA_HOME)/include" -I"$(JAVA_HOME)/include/linux" [16:01:05] CFLAGS =-Wall -Werror -fPIC -flto -shared $(INCLUDE) -lrt [16:02:11] shouldn't be a prob, openjdk is probably a build-depends [16:02:38] ok cool [16:11:30] dcausse: which package is that? [16:11:42] gehel: jvmquake [16:11:49] Oh, interesting! [16:12:04] * gehel is now excited! [17:50:02] quick break, back in ~15 [18:36:36] dcausse: ok if i switch you out for zbyscko in the MediaWiki event carried state transfer RACI? [18:36:38] https://docs.google.com/document/d/1LyOXA0W8FJGk65ku4W_ePS9MkkbANP7qKgKrLTOH6l8/edit [18:36:48] or uh, switch zbysko out for you [18:40:40] quick lunch, back before retro [18:57:39] ~5 mins late to retro [19:01:56] ejoseph, dcausse, cbogen_: WCQS retro starting: https://meet.google.com/xbj-eswc-aig [19:51:57] thanks team! this was a super interesting retrospective for me! [20:05:54] lunch [20:47:08] meh, the eternal question. oauth_settings can merge into extra_jvm_opts at any level from hiera to the tempate and everywhere in between. So where is the right place? [20:47:15] *template [20:49:43] I'd say hiera, but it's not like I know what I'm doing (yet... ;P ) [21:03:25] hmm, i suppose why not. I was ponder how to deal with that hiera wants to name things consumer_key but we are talking to java, so its consumerKey which requires silly transliteration somewhere in the middle as well. should just keep it simpler [21:05:16] * ebernhardson checks the definition of transliteration and realizes this doesn't count, but close enough :P [21:44:21] luncj [22:26:18] back