18:00:23 #startmeeting Infrastructure (2018-02-22) 18:00:23 Meeting started Thu Feb 22 18:00:23 2018 UTC. The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:23 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:00:23 The meeting name has been set to 'infrastructure_(2018-02-22)' 18:00:23 #meetingname infrastructure 18:00:23 #topic aloha 18:00:23 #chair smooge relrod nirik dgilmore threebean pingou puiterwijk pbrobinson maxamillion 18:00:23 The meeting name has been set to 'infrastructure' 18:00:23 Current chairs: dgilmore maxamillion nirik pbrobinson pingou puiterwijk relrod smooge threebean 18:00:32 .hello2 18:00:33 puiterwijk: puiterwijk 'Patrick "マルタインアンドレアス" Uiterwijk' 18:00:45 .hello2 18:00:46 tflink: tflink 'Tim Flink' 18:00:57 .hello pingou 18:00:58 pingou: pingou 'Pierre-YvesChibon' 18:00:59 ay Tim! 18:01:17 .hello2 18:01:18 smooge: smooge 'Stephen J Smoogen' 18:01:24 .hello2 18:01:26 keitellf: keitellf 'Lee Keitel' 18:01:27 hello o/ 18:02:07 How's everyone? 18:02:35 hello 18:02:42 ok nirik is running a bit late 18:02:44 * relrod waves 18:02:51 * sayan is here 18:03:06 #topic New folks introductions 18:03:06 #info This is a place where people who are interested in Fedora Infrastructure can introduce themselves 18:03:55 hello to anyone new 18:04:26 #chair tflink 18:04:26 Current chairs: dgilmore maxamillion nirik pbrobinson pingou puiterwijk relrod smooge tflink threebean 18:06:20 ok next up then 18:06:41 #topic announcements and information 18:06:41 #info Infrastructure hackfest in april: https://fedoraproject.org/wiki/Infrastructure_Hackathon_2018 18:06:41 #info Debugging 503's on mirrormanager dropped from 80,000 to 10,000 day (thanks nirik) 18:06:41 #info Rebuild mirrormanager container to be more resilient 18:06:54 There will be a hackfest in April 18:09:25 We have a problem with mirrormanager giving out 503's under heavy load. Out of 135 million requests we were seeing 80-100k failures 18:09:26 .hello2 18:09:27 lorbus: lorbus 'Christian Glombek' 18:09:45 we are looking at how to deal with this a bit better 18:10:05 do we know what's actually causing the errors? 18:10:09 I think that the mirrormanager stuff was related to how we restart things, and with some changes we got it significantly down, right? 18:10:53 keitellf, so the errors occur usually on the start of the hour from :00 -> :15 after 18:11:17 this is also when 50% of all mirror requests come in per day 18:11:36 That makes sense, cronjobs. 18:12:17 as puiterwijk said part of the problem was that it was related to how we restart things, and part of it was the number of dockers we had at a time. 18:12:50 changing those helped bring it down, but we have 2 hosts which consistently show multiple thousands during the 05:00 -> 05:15 UTC time 18:13:44 that is the busiest hour of the day 18:14:18 so next steps are to look at how many threads the containers have and how many are running at that time frame 18:14:30 and moving one of the proxies to a hopefully faster box 18:15:23 ok any other announcements for the last week? 18:16:15 #topic Apprentice Open office minutes 18:16:33 #info A place where apprentices can get help on any tickets or problems they are having 18:17:53 Any questions here? 18:18:44 Alright then.. 18:18:47 #topic Open Floor 18:19:14 OK I would like to get people to send stuff they would like talked about or updated to the list for next meeting 18:20:16 Anything else ? 18:20:31 .. is never sure if he is in a netsplit or everyone is being quiet 18:21:12 ok I am going to say "thank you all for coming today" 18:21:14 ..crickets 18:21:25 and I hope we have more next meeting 18:21:28 #endmeeting