18:00:05 #startmeeting Infrastructure (2015-02-26) 18:00:05 Meeting started Thu Feb 26 18:00:05 2015 UTC. The chair is nirik. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:05 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:00:06 #meetingname infrastructure 18:00:06 The meeting name has been set to 'infrastructure' 18:00:06 #topic aloha 18:00:06 #chair smooge relrod nirik abadger1999 lmacken dgilmore mdomsch threebean pingou puiterwijk 18:00:06 Current chairs: abadger1999 dgilmore lmacken mdomsch nirik pingou puiterwijk relrod smooge threebean 18:00:22 morning 18:00:35 hi. I'm here, but also in another meeting 18:00:36 * danielbruno is here 18:00:38 hi 18:00:38 * sborza waves hello 18:00:39 * lmacken 18:00:52 * pingou here 18:00:57 I am here 18:01:02 though sleeping I guess 18:01:16 :) 18:01:37 * relrod here 18:01:48 * tflink is here 18:02:16 morning everyone. ;) 18:02:21 hey hey 18:02:25 #topic New folks introductions and Apprentice tasks 18:02:32 any new folks like to introduce themselves? 18:02:38 or apprentices with questions or comments? 18:03:36 Hey, I have started with some packaging, and looking for a reviewer and sponsor for my first package. 18:03:47 And I would like to be added to the apprentice group again :) 18:03:55 kushalk124: cool. have a link? sure, we can add you... 18:04:04 Also, I would like to contribute to some python apps. 18:04:24 Hey nirik, https://bugzilla.redhat.com/show_bug.cgi?id=1194576 18:04:29 excellent. 18:04:38 :) 18:04:54 good day, folks. 18:05:02 kushalk124: see me in #fedora-admin after the meeting and I'll get you added... 18:05:09 morning danofsatx 18:05:17 nirik, sure 18:05:34 hey 18:05:40 morning dgilmore 18:06:10 #topic Applications status / discussion 18:06:20 any applications side news this week? or upcoming? 18:06:35 #info new nuancier deployed 18:06:40 #info new anitya deployed 18:06:56 isn't there a nuancier election going on now? 18:06:57 (with a nice bunch of fixes and new features :)) 18:07:02 nirik: yes it is :(: 18:07:04 :( 18:07:42 pingou: ok. 18:07:56 I wasn't happy about it, but I learned about the new election when it was already started 18:08:21 while there was 6 months work piled for the release :/ so we tested it carefully with gnokii before pushing to prod 18:08:26 as long as the update didn't break them... 18:08:32 (and so far, no reports so looks like it went ok :)) 18:09:04 ok, good. 18:09:43 ok, anything else? I know since we are in freeze things are a bit quiet. ;) 18:09:56 yeah, a new fmn release went out right before freeze 18:09:57 did I miss the new comers' introduction? 18:10:09 contained changes that made the format of some emails nicer (mostly). 18:10:11 intrk: yeah... but feel free to jump in now with one. :) 18:10:19 an update of the-new-hotness went out too with similar cosmetic changes. 18:11:46 #info fedora notifications and the-new-hotness had releases just before freeze. 18:12:41 okay, well... Hi! I'm a fairly young engineer, 25yo. Started at the Helpdesk, went through AV/Security support tech job and now I'm a Network Security technical support engineer. Got my RHCSA last year but since I work mostly on firewalls at the moment I want to have some more experience with Linux and thought this might be the place. 18:12:54 intrk: cool. welcome. ;) 18:13:17 I guess from that background you are more interested in sysadmin type things? or are you also interested in application devel stuff? 18:13:53 Both to be honest, however, I haven't dabbled in programming too much 18:14:31 I think I can read basic code and have passed an introductory course for Python and would love to get my feet with that also 18:14:32 cool. :) We have both.... 18:15:03 see me in #fedora-admin after the meeting and I can get you added to our apprentice group... and see #fedora-apps for the application side of things. 18:15:09 and again welcome. ;) 18:15:30 #topic Sysadmin status / discussion 18:15:40 :) superb, thank you 18:15:50 lets see... I wish I could say it was a quiet week with the freeze and all, but it's not really been. ;) 18:16:13 we have a possibly ongoing issue with out squid proxy. (I'll explain that more in a bit here) 18:16:53 I did manage to clean up some more playbooks monday... our ansible check/diff output is getting lower. 18:17:06 smooge reinstalled osuosl02 (one less machine in puppet) 18:17:13 cool 18:17:19 how many left? 18:17:27 (after I live migrated guests off it.) 18:17:38 For the record, virsh migrate is pretty awesome. ;) 18:17:59 I reinstalled sb10 but it crashed out and we are in a email trail of dead letters on getting it back 18:18:07 25 hosts left in puppet. 18:18:24 nirik: any big one remaining in there? 18:19:05 pingou: most of the ones left will be tricky. I sent a list to the list. 18:19:16 oh :( 18:20:22 I also looked into torrent software. There's no good news there really. Opentracker (what we use now) has some new commits, but nothing that seems very useful. 18:21:17 I did have the thought to see if we could switch from offering people torrent files to offering them magnet: uris... in that you can list hosts you can download from, so we could list our download.fedoraproject.org, etc. 18:22:05 I'll look at proposing something around that sometime. ;) 18:22:08 lets see... 18:22:27 #topic nagios/alerts recap 18:22:27 http://ur1.ca/jr7j4 18:22:28 googling how to make magnet uri isn't really helpful :/ 18:22:49 pingou: http://en.wikipedia.org/wiki/Magnet_uri 18:23:06 better than google :D 18:23:17 so basically that would allow us to not seed. We would just track and get some people to download the images from https first and seed. 18:23:57 oh, another thought I had was to make another torrent server so we could have both listed in the magnet link and can reboot one anytime we need to without outage... 18:24:55 anyhow, nagios has been somewhat less noisy... 18:25:14 I see proxy07 is high on the list, but I don't think any of those caused actual notifications 18:26:03 #topic Upcoming Tasks/Items 18:26:03 https://apps.fedoraproject.org/calendar/list/infrastructure/ 18:26:12 anything upcoming anyone would like to schedule or note? 18:26:56 ok, I have 2 items now... 18:27:00 #topic Meeting process 18:27:28 so, we have had this same meeting process for a long time, I read something the other day (which I cannot for the life of me find now) that had a interesting process... 18:27:37 basically the way it worked was: 18:27:51 * Open a shared document (for us this would be on gobby or something) 18:28:25 * Over the next week people add things to it. There's a 'status/info' section, there's a 'discuss' section and a 'todo/actions' 18:28:44 * next meeting people read that and we don't talk about status/info stuff... unless it needs discussion. 18:29:07 then we discuss the things on the list to discuss, and then make todos/actions which get added to the next shared doc 18:29:35 So, just a thought... I might send this to the list first and get more feedback before we change anything. 18:29:54 but I wanted to toss it out here too. ;) 18:30:09 interesting 18:30:26 yeah, the meetings take on a kind of pre-fab feel since the agenda is the same each week. 18:30:46 We do something similar at my current job with a share OneNote document. It works out fairly well. 18:30:48 yeah, and we all forget what we worked on really. 18:30:53 yeah :P 18:31:03 if we add stuff as we do it/think of it... it makes it easier. 18:31:10 I like it. The zodbot output is nice, but having a rw doc could be more collaborative if people have a week to poke at it 18:31:13 interestingly enough, we've been talking about something similar for qadevel meetings 18:31:31 tflink: cool. did you read something similar to this process recently? 18:31:42 it's driving me crazy that I cannot find the blog post or whatever I read. ;( 18:31:52 nirik: no, just trying to figure out how to make the meetings go faster 18:32:01 * tflink wasn't the person who suggested the shared document 18:32:27 https://sprint.ly/blog/kanban/ ? 18:32:29 * nirik wonders if it could help blocker meetings too. 18:32:47 we decided to try suggesting prepared #infos first because it's less complex 18:32:48 nope, that wasn't it. :( oh well. 18:33:02 nirik: we've wanted to support async blocker debate for a long time 18:33:04 * oddshocks pops in 18:33:12 so, should we try this for next week? or discuss on list first and try week after? 18:33:58 * tflink is fine with either but suspects that there may not be much feedback from a list discussion 18:33:59 I think having status/info stuff as #info in the doc would be nice... could cut and paste it into the meeting 18:34:25 tflink: true. Most people on irc meetings are... in the irc meeting. ;) 18:34:50 ok, so let me do this: after the meeting I will make a doc, and send info about it to the list... 18:34:57 then we can see what we can do with it next week. ;) 18:35:00 cool :) 18:35:11 #info will try adding a shared doc and see if it makes our meetings more helpfull. 18:35:21 #action nirik to make doc and mail list. 18:35:33 So, the other thing I wanted to go over was: 18:35:45 #topic Learn about an app: squid / kojipkgs 18:35:50 oooo 18:36:01 So, we have currently having some issues with squid. So, I thought I would describe how it's setup 18:36:13 and how we might do it better or replace it or find out how to fix it. 18:36:42 so, we have a machine that is 'kojipkgs01' internally, and kojipkgs.fedoraproject.org externally. 18:36:51 It runs apache/httpd on port 8080 locally only. 18:36:59 It has a nfs mount to the koji packages store. 18:37:13 it runs squid listening on port 80 and 443. 18:37:36 squid has it setup that the httpd on port 8080 is a 'parent' cache, so it queries it for the actual files 18:37:56 then squid caches in memory or disk for requests. 18:38:07 This worked without problem for a long time. 18:38:31 Then recently, I nuked the old rhel6 version and replaced it with a shiny rhel7 one. 18:38:52 I gave the new one tons and tons of memory with the idea it could memory cache most things. 18:39:08 But we are running into an issue where after a few days/week it just gets really really slow. 18:39:18 restarting it 'fixes' it 18:39:27 takes to long to search in its cache? 18:39:33 I adjusted the memory cache down today to see if that would help. 18:39:51 it's not clear. There's 0 errors in squid logs or messages or dmesg on the host. 18:40:16 squid is also running in it's default mode of 1 process. 18:40:29 squid has a smp mode, but I cannot seem to get it to function at all. 18:41:14 Some ideas: make a kojipgks02 setup the same way and make kojipkgs a round robin to both... that way at least when we restart builds don't die. 18:41:25 Replace with varnish? but not sure how that would work. 18:41:33 nirik: maybe we could use active/active 18:41:35 Replace with memcached? but again not sure how that would work 18:41:49 replace with mod_cache 18:42:08 I think I looked at that and couldn't get it working, but I could try again. 18:42:15 replace with something else 18:42:45 3) Run the RHEL-6 version on the RHEL-7 system and see if it is a kernel/architecture problem? 18:43:14 run a Fedora 22 version 18:43:23 or fedora 21 18:43:27 I can't get smp mode working here on f21. ;) 18:43:31 Or see if we built a RHEL-6 version with equivalent memory if it still happened 18:43:39 maybe its a bug in rhel 7's squid that is fixed in fedora 18:43:52 I still have the old rhel6 disk, but that is really not something I want to go back to. 18:44:06 I can see if there's any differences. 18:44:35 there's a big version difference. 18:45:16 squid-3.3.8-12.el7_0.x86_64 vs squid-3.4.9-3.fc22.x86_64 18:45:48 nirik: maybe a quick test is rebuild fedora's squid for rhel7 18:45:48 ok, some things to try. ;) 18:46:06 sure, although I hate to have to maintain it... for security updates forever 18:46:34 indeed 18:46:47 or we run Fedora server 18:47:16 perhaps. 18:47:22 #topic Open Floor 18:47:30 anyone have any items for open floor? 18:47:36 questions, comments, ideas? 18:47:51 oh 18:48:06 I was wondering if there would some more people interested in testing progit 18:48:27 afaics it starts behaving nicely 18:48:50 UI could be improved, but functionality is there 18:49:41 cool. Perhaps a call to list? 18:49:44 how would people test it? 18:49:51 at one point I'll start looking at packaging it and writing some doc on how to deploy it 18:50:04 smooge: for the moment the easiest is to use the dev instance: http://209.132.184.222 18:50:17 make a project, fork one, check the pull-requests, issues 18:50:24 see what needs improvement 18:50:49 and flood me with requests and bug report :] 18:51:19 okie dokie 18:52:18 pingou: or perhaps devel list at this point? 18:52:24 nirik: sure thing 18:52:29 cool. 18:52:33 thanks :) 18:52:40 #info everyone go test progit. :) http://209.132.184.222 18:52:49 ok, if nothing else will close out in a minute. 18:52:52 I should blog about it as well :) 18:54:03 yes. :) 18:54:09 ok, thanks for coming everyone! 18:54:12 #endmeeting