[01:19:57] <inflatador>	 whoohoo, looks like it finished
[09:24:16] <moritzm>	 ryankemper, inflatador: FYI, I ran the nginx restart cookbook on wcqs/wdqs hosts and it's working fine
[13:18:58] <inflatador>	 moritzm that's great news, thanks for the update
[13:20:25] <inflatador>	 re: our blazegraph/neptune discussion yesterday, looks like Brad Bebee, current Neptune GM at AWS, used to be the CEO of Systap https://www.prweb.com/releases/2015/11/prweb13083392.htm
[13:21:36] <dcausse>	 inflatador: yes
[13:23:51] <inflatador>	 (sorry if this is old news)  I updated the office wiki page with more info about him
[13:25:00] <dcausse>	 thanks!
[13:25:39] <inflatador>	 That press release says stuff about GPU acceleration for BG, is that something worth looking at or is it buggy/enterprise-only or anything?
[13:29:55] <dcausse>	 inflatador: I have no clue...
[13:31:17] <inflatador>	 dcausse no worries, I started grepping thru the source code it has press releases in the actual git repo. LOL 
[13:32:18] <dcausse>	 lol
[13:32:26] <dcausse>	 not seeing anything related to GPU acceleration in Neptune so not sure that went very far
[13:38:41] <inflatador>	 yeah, sounds like it was an enterprise-only feature based on https://github.com/blazegraph/database/issues/53
[14:36:23] <elukey>	 hello folks!
[14:36:44] <elukey>	 I found https://wikitech.wikimedia.org/wiki/Search/articletopic on wikitech and I was wondering if it is still active or not
[14:37:13] <elukey>	 if so I created https://phabricator.wikimedia.org/T317768 to propose a new set of streams, and I'd like your opinion :)
[14:39:35] <gehel>	 Yes, those are still active. I'll let Erik and David comment on the task for the specifics, but it looks like you want us to migrate to a new stream / hive table. I suspect there are a few hurdles if the naming of those topics change with the new version...
[14:58:51] <elukey>	 gehel: yep yep this would be the plan if everybody agrees :)
[16:07:21] <inflatador>	 workout, back in ~30
[16:45:34] <inflatador>	 back
[17:13:46] <gehel>	 I just learned about https://office.m.wikimedia.org/wiki/Slack/Channels. Quite interesting! Might be useful for pfischer and as part of our onboarding checklist.
[17:14:55] <gehel>	 There is a surprisingly large number of social channels!
[17:43:48] <inflatador>	 lunch, back in ~45
[18:05:43] <ebernhardson>	 meh, dropping historical subgraph data failed with a null pointer exception in the hive command, no fun :P
[18:16:56] <ebernhardson>	 unrelated, but for the randomly curious, the regex acceleration takes /\<ref( [^\>]*)?\>https?:\/\/[^ \<\>\{\}]+ *\<\/ref/i and turns it into this as a pre-filter: htt AND (("ef>" AND "f>h") OR ("ef " AND "f >" AND " >h") OR "ef ") AND ">ht" AND "://" AND (("tp:" AND "p:/") OR ("tps" AND "ps:" AND "s:/")) AND "<re" AND "ttp" AND "ref"
[18:20:48] <ebernhardson>	 (i might not quite have the AND/OR in the right place, the profile output doesn't actually say)
[18:25:24] <ebernhardson>	 and the npe exception is...hive has terrible error handling :P `alter table foo drop partition (foo='bar', baz='bang')` NPE's if one of those columns isn't an actual column in the table
[18:57:39] <inflatador>	 back
[19:41:13] <Trey314159>	 ebernhardson: that regex acceleration is pretty slick. I had assumed it would ignore optional characters (and lose trigrams spanning them) so including the OR clauses is cool to see.
[19:44:22] <ebernhardson>	 there is still some meh in there,  `("ef " AND "f >" AND " >h") OR "ef " could probably be collapsed into only "ef ", but i suppose not bad :)
[19:44:43] <ebernhardson>	 school-run time, back in a bit
[20:04:12] <inflatador>	 quick break, back in ~20
[20:18:28] <ebernhardson>	 back
[20:22:33] <inflatador>	 back
[20:41:41] <ebernhardson>	 meh, test failures in CI because 8.0225368120364E-5 != 8.022536812036404E-5
[20:41:54] <ebernhardson>	 wonder why it got two extra decimals of precision.  We probably only need like 8.022 anyways
[20:42:02] <ebernhardson>	 8.022E-5
[20:42:30] * ebernhardson has some memory that we set decimal precision somewhere before...
[20:52:29] <ebernhardson>	 hmm, it's using the same docker container as a run from yesterday, implies it's not like a php minor version bump or something caused issues
[21:40:00] <ebernhardson>	 meh, this portion of the test suite doesn't use the standard fixture handling stuff we put together, so it doesn't get the decimal precision fix. No clue why it decided today was the time to start failing...working up a patch to make it use the new stuf
[21:59:22] <ryankemper>	 brian and I did an audit of the 10G networking status
[22:00:33] <ryankemper>	 codfw will be 100% 10G once we decom the remaining codfw hosts, so that's very straightforward, but eqiad will require re-racking 15 hosts: https://phabricator.wikimedia.org/T317816
[22:46:21] <ebernhardson>	 meh, cross-repo dependencies :( almost have this fixed but now need more patches in more places :P
[22:46:34] * ebernhardson seems some of the allure of mono-repos :P