[08:52:28] @seen Hashar [08:52:28] Steinsplitter: Last time I saw Hashar they were quitting the network with reason: Quit: I am a manual virus, please copy me to your quit message. N/A at 9/22/2013 4:55:07 AM (1d3h57m20s ago) [10:56:42] (03PS2) 10Hashar: provide sane configuration defaults [labs/tools/grrrit] - 10https://gerrit.wikimedia.org/r/85404 [10:57:29] (03CR) 10Hashar: "tab to spaces" [labs/tools/grrrit] - 10https://gerrit.wikimedia.org/r/85404 (owner: 10Hashar) [10:57:52] ty, hashar [10:59:28] PissedPanda: :D [10:59:53] PissedPanda: does grrrit supports sending a notification to two channels ? [10:59:59] hashar: it does! [11:00:06] you just add the appropriate repos to both and it'll work [11:00:25] hashar: of course, you must remember to add it to both :) [11:02:25] hashar: do you know if we've anything production written in Ruby? [11:02:27] other than puppet [11:02:59] Parsoid ? [11:04:14] (03PS1) 10Hashar: qa/ci sent to #wikimedia-qa as well [labs/tools/grrrit] - 10https://gerrit.wikimedia.org/r/85653 [11:04:32] hashar: Parsoid is in JS [11:04:53] PissedPanda: ah we have the qa browsertests. The selenium browser testing tool is driven by a ruby software [11:05:11] (03CR) 10Yuvipanda: [C: 032 V: 032] qa/ci sent to #wikimedia-qa as well [labs/tools/grrrit] - 10https://gerrit.wikimedia.org/r/85653 (owner: 10Hashar) [11:06:04] hashar: can you try adding a comment or so to that? [11:06:06] err [11:06:07] hashar: to test? [11:06:24] well the bot has not joined #wikimedia-qa [11:06:26] oh [11:06:28] here it is [11:07:07] my net is slow :/ [11:07:09] reconnecting [11:07:31] (03PS3) 10Yuvipanda: provide sane configuration defaults [labs/tools/grrrit] - 10https://gerrit.wikimedia.org/r/85404 (owner: 10Hashar) [11:07:58] (03CR) 10Yuvipanda: "Going to test and merge in a bit" [labs/tools/grrrit] - 10https://gerrit.wikimedia.org/r/85404 (owner: 10Hashar) [13:24:55] Taking a day off today so that I have a "real" weekend, but if there are blockers that can be solved with a bit of attention now is a good time. [13:25:30] Coren: have a nice weekend :) [13:40:56] [bz] (8NEW - created by: 2Antoine "hashar" Musso, priority: 4Normal - 6enhancement) [Bug 36994] [OPS] Add disk I/O to ganglia reports - https://bugzilla.wikimedia.org/show_bug.cgi?id=36994 [13:50:27] bye Coren [13:52:39] !log bugzilla - creating new instance bugzilla-current, copying prod files .. [14:00:24] mutante: morebots seems down [14:02:03] " runs from Wikimedia Labs under the "morebots" account " [14:02:06] where was that again :p [14:20:11] mutante: Guten Tag :-D [14:20:25] mutante: whenever you find out where morebots is, you might want to update https://wikitech.wikimedia.org/wiki/morebots :D [14:20:29] doc doc doc [14:35:03] hashar: bonjour. but . but . you're telling me when https://wikitech.wikimedia.org/w/index.php?title=Morebots&diff=84064&oldid=46100 [14:35:34] ok, looking at bots project instances [14:35:35] oh [14:35:48] is that still running gin andrew's home dir on wikitech ? :D [14:36:02] i guess not :P [14:36:36] * 09:08 labs-logs-bottie: petrb: restarted morebots * 18:38 andrewbogott: puppetized labs-morebots via role::logbot::wikimedia-labs [14:36:45] bah [14:37:30] eh , i saw that log [14:37:36] but it didnt tell me which instance [14:37:38] mutante: andrew migrated it on September 4th: 9:10 andrewbogott: Moved morebots from wikitech-static to toollabs [14:37:51] yeah blame [14:37:55] :-D [14:37:59] heh [14:38:19] so.. tools-login ? ehmm.. cough [14:40:09] no clue :D [14:41:11] hmm.. and where's the role:) [14:41:39] grepped for logbot [14:44:26] sigh, also i'm project admin but key denied on bots-labs [14:51:34] it's not bots-4, it's not bots-cb, that role is not listed in "puppet class" on instances, tools-login Connection closed by UNKNOWN [14:53:16] petan: ? morebots is dead and no doc around :( [14:53:23] ok sec [14:53:40] ah:) hi [14:53:43] whenever you can, could you also update the lame doc at https://wikitech.wikimedia.org/wiki/morebots ? :D thx [14:53:53] just add which instance this is on, thx [14:54:06] or i would have ran the init script:) [14:54:25] fixing... [14:54:29] eh, didn't even expect it was still here duh [14:54:32] there is README actually in folder [14:54:37] or we could have just looked at IP [14:54:40] I followed it [14:54:51] in which folder? [14:54:52] !log <3 [14:54:53] Message missing. Nothing logged. [14:54:59] in $HOME [14:55:03] become morebots [14:55:05] vi README [14:55:22] ok, but where do i do that [14:55:30] do what [14:55:34] ah [14:55:37] on tools-login [14:55:37] where do i connect to from home to get to morebots [14:55:46] ssh tools-login [14:55:50] become morebots [14:55:51] ok, there we go:) [14:55:53] :P [14:55:59] i tried that, but i think i'm not setup yet [14:56:06] because tools-login doesnt like my key yet [14:56:10] aha [14:56:11] so, thanks for fixing [14:59:36] !log bugzilla - created new instance bugzilla-current, copied prod files (without pass) for migration cleanup [14:59:38] Logged the message, Master [15:00:20] hashar: gs is broken on beta :P [15:01:13] Steinsplitter: fill a bug please :-) [15:02:03] petan: can you copy paste the doc at https://wikitech.wikimedia.org/wiki/Morebots and write down a few lines about how to restart morebots ? :D [15:02:13] ok [15:02:16] every time it dies people waste time attempting to figure out how to restart it hehe [15:02:25] that doc will be a huge time saver! [15:02:26] \O/ [15:02:32] :) yes [15:03:17] although ideally that should be in an upstart job that takes care of restart the lame script [15:03:18] :D [15:03:33] done [15:07:31] https://wikitech.wikimedia.org/w/index.php?title=Morebots&diff=84079&oldid=84078 [15:07:49] "runs from" -> Tool Labs (vs. Labs) [15:08:01] I guess someone will be able to figure it out now. Thx petan [15:08:06] and external to Mediawiki.org .. because .. [15:08:13] https://wikitech.wikimedia.org/wiki/Tool_Labs [15:08:24] is a sof redirect to Nova_Resource:Tools [15:08:30] ? shrug [15:10:28] ssh to tools-dev .. is that after "ssh to tools-login"? [15:11:00] mutante: tools-dev.wmflabs.org works too, btw [15:11:02] it has a public ip [15:11:40] heh, yea, i just noticed, i'll just have to start being a tools labs user:) [15:15:27] Coren, YuviPanda and other people who care about tool labs: https://en.wikipedia.org/wiki/User_talk:Petrb#Tool_Labs_Homepage_.2F_Wiki [15:15:37] some complaints regarding broken docs [15:16:10] no idea what mark stats etc are [15:31:47] petan: me neither :) [15:49:40] Growing whitespace https://wikitech.wikimedia.org/w/index.php?title=Nova_Resource:I-00000601&action=history [15:52:16] [bz] (8NEW - created by: 2Tim Landscheidt, priority: 4High - 6major) [Bug 52560] icinga.wmflabs.org is down: "Error: Could not read host and service status information!" - https://bugzilla.wikimedia.org/show_bug.cgi?id=52560 [15:58:39] !log nagios restarting nagios3 on nagios-main (which is icinga.wmflabs, was down per bug 52560) [15:58:40] Logged the message, Master [16:00:09] labs icinga is back [16:01:50] will it complain about puppet freshness? :P [16:02:23] Coren, pung [16:02:25] Ping [16:03:16] petan, I'm confused by 'there's an init script to start/stop it' -- is that something you just now set up? [16:06:10] mutante, hashar, petan, I documented this a while ago, here: https://wikitech.wikimedia.org/wiki/Admin_Logs and linked from the front page on wikitech... [16:06:22] but apparently that's too obscure since /I/ couldn't find that page just now either :) [16:16:07] andrewbogott: should be on the morebots page on wikitech :) [16:16:48] hasharCall, that page is kind of wrong though… I'm fixing. [16:17:00] thanks andrew:) [16:17:30] yea, searching botname on wikitech is the usual first attempt [16:17:46] but i know we also have bot triggers [16:17:57] but then i end up doing !help :) [16:26:36] !log deployment-prep rebuilding search indecies after new index config deployment [16:26:39] Logged the message, Master [16:33:15] Hi, is there a list of users providing tools through Wikimedia Labs, including contact information? [16:34:20] hmm .. https://wikitech.wikimedia.org/wiki/Special:ListUsers [16:34:55] pajz: If you want to email everyone, though, just use the labs mailing list... [16:35:12] http://www.mediawiki.org/wiki/Wikimedia_Labs/Tool_Labs/List_of_Toolserver_Tools [16:35:18] this is for toolserver / tool labs [16:35:27] just "labs" is a superset of that [16:37:23] pajz: pajz_ you've already looked at http://tools.wmflabs.org/ ? [16:37:42] it does include some user contact info [16:38:11] ah, yea, that link is better [16:38:38] Thanks, mutante and andrewbogott, I wanted to email one specific tools operator, but apparently he hasn't enabled the email functionality on Meta. I was wondering if there is some page listing contact data. [16:38:49] And thanks sumanah, I'll check that page. [16:39:02] did you check his user page on wikitech wiki? [16:43:49] hey Damianz_ how are you? [16:48:09] mutante: There is nothing on the user page (Plus, I'm not registered, so I couldn't use Special:Emailuser even if it were enabled). But I think I just found the email address on Bugzilla. [16:49:28] pajz_: ah, i see, cool [16:50:30] you can register pajz_ - anyone can register an account on wikitech.wikimedia.org [16:51:17] i see, but it's not part of SUL? [16:53:38] pajz: no, it isn't, yet [16:53:43] it may be at some point in the future [16:53:49] ok, thanks. [17:10:49] Technical_13: actually, i don't have a clue if i still have access. wm-bot's been moved to tools by now, right? [17:11:59] sumanah: do you have a countdown clock? anyway, mazel tov [17:12:51] :) [17:12:56] *curtsey* [17:16:06] I don't think so jeremyb (except wm-bota). petan has restarted them already. [17:16:27] Technical_13: yeah, i was making a wild guess that you didn't still need a restart :P [17:16:45] aude, did I leave you hanging last week, or are your wikidata instances mostly doing what you need now? [17:17:25] andrewbogott: looks like it is "restored" to where i had it [17:17:42] i might still try rebuilding, but am running into insufficient resources to make a new instance [17:17:56] Oh, hitting a quota limit? I can fix that. [17:17:59] yes [17:18:09] ok, hang on... [17:18:45] Hm, you should have 7 cores left to go. [17:18:52] i don't know exactly which thing we are hitting quota for [17:18:56] Oh, RAM, probably. [17:20:06] try now? [17:20:50] ok [17:23:02] logging in.... [17:23:50] andrewbogott: it works :) [17:24:03] ok! [17:24:18] probably won't be tonight but i'll try enabling the puppet modules there [17:24:26] and maybe can delete the old instance [17:24:40] is the link from http://tools.wmflabs.org/ to "create new tool" wrong? [17:25:08] looks to me like it should point to Special:NovaServiceGroup and not Special:NovaProject [17:26:20] Nettrom, I think you're right. [17:26:28] I will fix (unless you can...) [17:26:52] I have no idea how to fix that [17:26:57] Oh, hm, not a wiki page... [17:29:04] [bz] (8NEW - created by: 2Andrew Bogott, priority: 4Unprioritized - 6normal) [Bug 54474] Wrong link on Tool Labs doc page - https://bugzilla.wikimedia.org/show_bug.cgi?id=54474 [17:29:32] Nettrom, thanks for catching that. Bug entered ^ [17:29:49] andrewbogott: no problem, thanks for making the bug report! [18:17:11] can someone confirm that http://pinklake.wmflabs.org is slow because labs instances are slow, or did I do something suboptimal in the mw install? [18:22:14] cscott: almost definitely something suboptimal [18:22:23] cscott: where did you put mediawiki? did you use the role to install it? [18:22:33] are you using apc? memcache? [18:23:02] (you should probably use both apc and memcache) [18:24:26] Ryan_Lane: i used 'sudo git clone' like all the power-hackers do [18:24:41] we have a puppet role that will do everything for you [18:24:45] and do it optimally [18:24:55] i'm told there's vagrant/puppet magic to give me a nice mw+visual editor install [18:25:20] there's vagrant for it. not sure about it existing in labs' puppet yet, though [18:25:26] if someone could give me a new instance to play with and maybe some pointers on how to do puppet incantations, i'd be interested in trying to do things The Right Way (tm) [18:25:26] but there's a mediawiki role [18:25:37] and installing visual editor past that shouldn't be too hard [18:25:56] you can create your own instances, can you not? [18:26:11] i don't know, can I? i haven't really flexed my ops powers yet. [18:26:22] lessee [18:26:24] how did you get the instance you are using? [18:26:34] (labs is self-service) [18:26:40] marktraceur made it for me [18:26:43] I see [18:26:50] maybe they didn't give you projectadmin in the project [18:27:08] go to wikitech, log in, then go to "Manage instances" in the sidebar [18:27:16] nope, https://wikitech.wikimedia.org/wiki/Nova_Resource:Visualeditor says i'm a project admin [18:27:28] @mediawiki [18:27:33] @search mediawiki [18:27:33] Results (Found 16): morebots, labs-home-wm, labs-nagios-wm, labs-morebots, gerrit-wm, extension, revision, info, bots, labs-project, openstack-manager, wl, deployment-prep, coren, sumanah, dependency, [18:27:54] -_- [18:28:03] how the fuck does the key I keep making keep disappearing [18:28:32] @search singl [18:28:32] No results were found, remember, the bot is searching through content of keys and their names [18:28:34] @search single [18:28:34] No results were found, remember, the bot is searching through content of keys and their names [18:28:52] !mediawiki is https://wikitech.wikimedia.org/wiki/Help:Single_Node_MediaWiki [18:28:52] Key was added [18:28:56] !mediawiki [18:28:56] https://wikitech.wikimedia.org/wiki/Help:Single_Node_MediaWiki [18:29:13] look at that documentation for how to get puppet to install mediawiki and its dependencies [18:30:48] cool, thanks. [18:30:54] yw [18:35:08] Ryan_Lane: oh, and YuviPanda changed pinklake from gluster to nfs -- how do i do that on my new instance? [18:36:57] there's a role for that as well [18:37:30] oh, role::labsnfs::client looks like it's that [18:37:34] you need to apply the role, run puppet, then reboot the instance [18:37:45] that's theo ne [18:37:47] *the one [18:41:02] [bz] (8NEW - created by: 2Peter Schlömer (dapete), priority: 4Unprioritized - 6enhancement) [Bug 54478] Install graphviz on exec nodes and -dev - https://bugzilla.wikimedia.org/show_bug.cgi?id=54478 [18:43:54] ok, i need to understand how puppet and vagrant interact better. ori-l once told me, "vagrant enable-role visualeditor ; vagrant provision" would get me mw + ve + parsoid. and YuviPanda said (wrt my togetherjs extension) "yeah, but I want to puppetize it so 1. it is easy to spin out more instances, 2. it is easy for other people to hack on via vagrant" [18:44:17] yeah, so they both use puppet, but different sets of code [18:44:25] they haven't been merged [18:44:25] so now i understand how to get puppet roles installed on my instance --- but ve is a vagrant role? i don't really understand. [18:45:26] oh, hm. so labs instances can't directly take advantage of vagrant roles? [18:45:39] no [18:49:58] erm. [18:50:18] so there's some *other* category of servers which can use vagrant roles? [18:50:23] no [18:50:34] you use vagrant on your local hardware [18:50:50] I've asked folks to try to merge the vagrant code and the puppet code ;) [18:51:07] hm. [18:51:46] ok, so the existing towtruck instance was using the puppet mediawiki role. so that's not responsible for slowness. [18:52:08] what about enabling apc and memcache? are those puppet roles? (ie, easy to enable?) [19:07:44] chrismcmahon: are you around ? :D [19:07:49] cscott, mediawiki_singlenode might enable memcache by default, let me check [19:08:40] hi hashar [19:09:22] chrismcmahon: I got a very lame question regarding browser tests :D I cant find how to run them using ruby1.9.3 instead of the default 1.8.7 :D [19:09:44] hashar: yes, you need 1.9.x or higher [19:10:00] yeah I got it on the box [19:10:02] http://paste.openstack.org/show/47392/ [19:10:09] using bundle exec cucumber [19:10:16] cscott, most likely apc is already enabled. To turn on memcache just check the 'memcached' box in your puppet config page on wikitech. [19:10:30] chrismcmahon: I don't know how to have bundle/cucumber to use the ruby1.9.3 install :/ [19:11:32] chrismcmahon: will look it up [19:11:47] hashar: have you read https://github.com/wikimedia/qa-browsertests/blob/master/README.md ? [19:11:57] chrismcmahon: yeah I installed using README.md [19:12:05] hashar: if you can use RVM, that's probably the way to go [19:13:08] bundle: You’ll need to install 1.2.0 of bundler to use the ruby keyword. || https://devcenter.heroku.com/articles/ruby-versions [19:13:09] yeahhh [19:20:27] ^demon: CirrusSearch on wikitech? [19:20:38] <^demon> Yessir. [19:20:55] ^demon: doesn't really look like it, though.https://wikitech.wikimedia.org/w/index.php?title=Special:Search&search=check&fulltext=Search&redirs=0&profile=default&srbackend=CirrusSearch has all kinds of wiki markup in the highlighting [19:21:08] <^demon> Yeah, that's probably because we use SMW. [19:21:16] huh [19:21:22] <^demon> Semantic MediaWiki. [19:21:25] <^demon> Is my guess. [19:21:32] maybe it is just not resolving templates at all. [19:21:33] stabstabstab [19:21:34] <^demon> But yes, it's on and set as $wgSearchType. [19:22:27] <^demon> Ah, we indexed before it was default or something. [19:22:33] huh. I'll file a low priority bug to look at it with SMW. [19:22:40] <^demon> So we indexed wrong text in ES. They'll slowly purge out. [19:22:43] <^demon> It's not SMW. [19:22:59] <^demon> Refresh that search you linked above, I null edited the first two docs. [19:23:03] <^demon> No more wikitext. [19:24:01] weird. I thought that we didn't have to be default to purge wikitext [20:20:41] [bz] (8NEW - created by: 2Mark A. Hershberger, priority: 4Low - 6enhancement) [Bug 54427] Create ping.wmflabs.org - https://bugzilla.wikimedia.org/show_bug.cgi?id=54427 [20:42:59] Coren: https://gerrit.wikimedia.org/r/#/c/84969/ [22:22:26] ohai cscott [22:25:20] hey cscott [22:25:30] hello again [22:25:35] oh, where were we [22:25:54] oh, yeah, StevenW was complaining that pinklake was slow, and i was trying to figure out why [22:26:04] cscott: ah, right. [22:26:14] i enabled memcache, and the consensus seemed to be that apc was already enabled. and we turned on nfs. [22:26:17] cscott: just generally slow? [22:26:24] cscott: is it slower than, say, blue-dragwon.wmflabs.org? [22:26:30] (blue-dragon is setup via vagrant puppet) [22:27:54] i don't know, blue-dragon doesn't seem to have any content in it. [22:28:11] it could be that the slowness is because i chose to import 'realistic' articles from enwiki, and parsing them is slow. [22:28:24] that's highly possible [22:28:42] certainly https://pinklake.wmflabs.org/wiki/Wikipedia displays much faster the *second* time it is loaded [22:28:45] cscott: are you putting mediawiki in /data/project? [22:29:05] it's in /srv/mediawiki, i think. wherever the default puppet role puts it [22:29:10] ah. ok [22:29:18] it's fast for me [22:29:56] random page is slow... [22:30:08] incredibly slow [22:30:10] Ryan_Lane: btw, this is also using dynamicproxy. Potentially another aspect, although I guess that the culprit is far more likely to be mediawiki [22:30:12] YuviPanda: anyway, to complete your backlog catchup, i was trying to figure out if things would be faster if i used the standard vagrant visualeditor install instead of installing via 'sudo git clone', and found out to my sadness that i couldn't install from vagrant on an wmflabs instance because puppet != puppet [22:30:22] YuviPanda: yeah [22:30:24] indeed, cscott. [22:30:36] cscott: let me see if I can fix that [22:30:39] Ryan_Lane: yeah, i suspect it's just the first parse is really really slow, although i don't know why (other than than enwiki pages are complex) [22:30:50] cscott: that [22:30:59] parsing can take ages [22:31:11] what instance is this? [22:31:44] towtruck. https://wikitech.wikimedia.org/wiki/Nova_Resource:I-000008d0 [22:32:22] i was trying to populate with 'interesting' pages to edit. [22:33:29] $wgMainCacheType = CACHE_NONE; [22:33:29] $wgMemCachedServers = array(); [22:33:46] ow [22:33:53] you realize why that might be a problem. right? :) [22:34:23] no memcache and no caching at all [22:35:09] i just turned on the memcache puppet role, probably i need to rerun puppet before that takes effect [22:35:15] btw, when you configure memcache, notice that the system is running it on port 11000 [22:35:20] (well, "just turned on" == hours ago, but i haven't rerun puppet since) [22:35:22] memcache puppet role doesn't really add it to wiki [22:35:28] indeed [22:35:29] cscott: puppet runs automatically every 30m or so [22:35:44] you need to configure mediawiki to use memcache [22:35:51] you also need to enable caching [22:35:51] YuviPanda: i was hoping that was the case, which was why i was being lazy about trying to figure out if i needed to do more work. ;) [22:36:00] :D [22:36:35] so part of the answer is, "the default mediawiki role is uncached and rather slow" i guess. which comes back to my initial question where i was trying to figure out if there was a 'better' way to do the setup. [22:36:54] i guess not, i should turn on the caching and whatnot manually? [22:37:13] reparsing after an edit is likely to still be slow, though, right? [22:37:42] [bz] (8PATCH_TO_REVIEW - created by: 2Kunal Mehta (Legoktm), priority: 4Unprioritized - 6normal) [Bug 54444] Conflicting versions and paths of 'java' on -dev and -exec nodes - https://bugzilla.wikimedia.org/show_bug.cgi?id=54444 [22:37:56] yes, there's no way to make it automatic [22:38:05] reparsing after an edit will be slow, yes [22:38:13] but it is in production, too [22:40:17] ok, any suggestions for where i should look for 'sane' cache settings? [22:40:42] i'd steal the LocalSettings.php of wiki.laptop.org, but i don't think i have access to it anymore. ;) [22:41:01] https://www.mediawiki.org/wiki/Manual:Cache [22:41:07] thx [22:41:18] well, that's not amazingly helpful [22:41:36] https://www.mediawiki.org/wiki/Memcached [23:25:52] Ryan_Lane: I'm investigating a 'labsmediawiki' module to ops/puppet that'll bring the vagrant stuff to labs in a more easy way :) hope to have something ready tomorrow [23:26:14] heh. ok [23:26:26] I actually think paravoid's proposal is best, though [23:27:00] though it's quite a bit of work and bringing vagrant in is probably a good intermediate step [23:27:03] Ryan_Lane: oh sure, I complerely agree [23:30:36] sorry, Ryan_Lane - my bouncer has gone kaput [23:30:38] * yuvipanda_ checks logs [23:30:39] !logs [23:30:40] raw text: http://bots.wmflabs.org/~wm-bot/logs/%23wikimedia-labs/ cute html: http://tools.wmflabs.org/wm-bot/logs/index.php?display=%23wikimedia-labs [23:31:18] YuviPanda: I hadn't said anything since your last response [23:31:51] Ryan_Lane: heh, I sent about 5 lines which have been lost [23:32:15] Ryan_Lane: so, yeah - vagrant in labs is a 1 day hack, while paravoid's suggestions were a lot nicer but more-work-is-needed [23:32:24] back! [23:32:40] yeah [23:33:05] so, really it could be relatively quickly done [23:33:16] I guess dealing with settings is harder, though [23:33:25] like passwords and such [23:33:31] true [23:33:38] currently vagrant just sets them to 'vagrant' / 'admin' :P [23:33:41] but otherwise it's easy [23:34:07] Ryan_Lane: yeah, I manually set one up at puppet-dawg, and it took me about 3 mins (not counting puppet run time) [23:34:23] I'd say create 3-4 app servers [23:34:32] Ryan_Lane: i'm currently making sure that the virtualbox specific stuff in vagrant puppet are put behind a variable guard. after that should be easy. [23:34:48] then, we'd use service groups [23:34:52] Ryan_Lane: this is a different use case, no? One instance per project, as it runs now. [23:34:59] Ryan_Lane: true, we'd migrate to that at some point [23:35:04] a user creates a service group, we have a service that notices that and does a mediawiki clone [23:35:10] but it does a clone that's based on a shared repo [23:35:14] (or set of repos) [23:35:46] when a user creates a service group, it would also create a dns entry, and add a proxy rule [23:36:16] Ryan_Lane: ideally, I'd liek to think that the user creates a service groups, specifies what roles he wants, and then we do a puppet run with those params (user / roles) and it sets it up [23:36:17] this also lets it update nicely [23:36:46] well, in the setup I'm talking about, we'd have shared app servers [23:36:49] true [23:36:51] wait [23:36:54] and they'd use redis with hash prefixes [23:36:57] Ryan_Lane: why would that restrict us from using puppet? [23:37:04] puppet isn't great for this [23:37:09] salt would be better [23:37:16] and it would also fit in with the next step [23:37:26] which is adding your gerrit stream watcher [23:37:54] we'd have gerrit give permissions to remote branches for the service groups [23:37:59] based on the users in the service group [23:38:18] the users would push into the remote branch, and your bot would trigger a git-deploy [23:39:05] we don't really need a puppet run at all anywhere, because the initial creation of a deployment is just a git-deploy [23:39:15] I should really read up on git-deploy :P [23:39:49] users would manage their LocalSettings.php (minus passwords and other private info) in their remote branch [23:39:49] Ryan_Lane: the reason I suggested puppet / roles was because - how would you manage which version of core goes with which version of which extension? [23:39:49] plus deployment config stuff [23:39:54] along with a set of submodules [23:40:00] ah, hmm. [23:40:05] would this be a branch of mediawiki/core? [23:40:07] like we do deployment branches [23:40:10] right [23:40:26] yeah. we should make sure this isn't going to piss off the MW folks :) [23:40:43] 'WHY ARE THERE SO MANY BRANCHES I DO NOT CARE ABOUT!!1!?' [23:40:47] I should write up docs on this :) [23:40:51] YuviPanda: indeed [23:40:54] Ryan_Lane: good idea! [23:40:56] I doubt there'll be more than 300 or so [23:41:08] but that's still a lot of branches [23:41:24] can puppet do conditional 'include's? [23:41:32] yes [23:41:33] Ryan_Lane: should be fine if we prefix them [23:41:36] but puppet is slow [23:41:46] and git-deploy already handles basically everything we need [23:41:53] is there docs on git deploy, Ryan_Lane? [23:42:03] users wouldn't need to worry about that ;) [23:42:14] they'd just need to push stuff into their branch, and it would get deployed [23:42:48] maybe we should only trigger on tags, though [23:43:27] I need to make git deploy changes for this too, since we'd want repos to share a common git base [23:43:33] otherwise we'd eat up shit-loads of disk space [23:43:49]