[00:00:54] module stuff is autoloaded [00:01:01] you need a role class, yes [00:01:14] urgh why do I need a role class if it's autoloaded [00:01:20] the role class will include the classes from the modules [00:01:28] then the node will include the role class [00:01:35] sigh [00:01:38] it's only autoloaded if you include the module [00:01:44] err [00:01:46] that makes sadface [00:01:52] well, it makes sense [00:01:55] Should the https://labsconsole.wikimedia.org/wiki/Help:Using_debs_in_labs contain a mention of how to create the Package and Release file? Or should those be automatically created? [00:02:02] you have to tell puppet to use the class, obviously [00:02:24] apmon: well, that would be a really long guide [00:02:29] we should just link to debian's guide [00:03:11] "apt-ftparchive release . > Release" and "apt-ftparchive packages . > Package" seems to have done it for me [00:03:33] in /data/project/repos/ [00:04:07] Ryan_Lane: So https://gerrit.wikimedia.org/r/#/c/26441/10/manifests/role/labs-bots.pp is needed eventhough it seems a crazy waste of space and typing [00:04:16] Which is why these where role classes to start with [00:05:37] For signing you need an additional "gpg -abs -o Release.gpg Release". But overall that doesn't seem to complicated to not include it on the wikipage (assuming what I did was correct) [00:05:57] * Damianz kicks puppet and sees if this runs [00:06:13] nope can't find class. same as last time [00:08:22] * Damianz tries class {'':} [00:08:30] would all of these be included on a single node? [00:08:43] or would these all be included on different nodes? [00:09:09] Each class is basically a diferent role with 1 common class included on all of them [00:09:35] ie mysql, apache, common application and a generic symlink+locals class. [00:10:15] Using class { } gives an Puppet::Parser::AST::Resource failed with error ArgumentError, using include gives a Could not find class labs::bots::common for *sigh* [00:12:59] yeah [00:13:11] you can't have role::blah::blah2 and include blah::blah2 [00:13:18] because, again, puppet is a piece of shit [00:13:23] wtf [00:13:26] yeah [00:13:28] it's a bug [00:13:33] so I have to name things retardadly [00:13:37] yep [00:13:43] what a pile of shit [00:13:48] well... [00:13:56] if we make the roles modules too, then that's not as much of an issue [00:14:03] we should probably make the roles modules [00:14:12] err [00:14:15] one module role [00:14:19] wait [00:14:20] sorry [00:14:20] What, put all in 1 module and include it in site? [00:14:22] one role module [00:14:34] then each role would be a class in the module [00:14:48] because that's not going to get huge [00:14:57] no larger than it currently is [00:15:10] also [00:15:21] why the bastard pissing fuck can't I rename classes in the gropu manager [00:15:28] heh [00:15:32] I can change which group they are in which is pointless [00:15:43] you can delete and recreate ;) [00:15:51] I have an open bug on that [00:16:03] yeah because that is like 2 more pages per item :( [00:16:27] 3 clicks to delet something is enough to give someone RSI [00:16:40] ideally, I'd turn that entire page into two fancy multi-selects [00:16:50] but not an actual multi-select, because those suck [00:17:03] also, if I could rename them I wouldn't then have to reconfigure all the instances [00:17:13] yeah [00:17:18] I know. it's not ideal [00:17:31] patches welcome :) [00:17:41] though it's also on my agenda at some point [00:18:56] oh come on puppet [00:19:07] So I renamed the role classes and it still throws an error grrr [00:22:06] same error? [00:22:18] thanks for working on this btw [00:22:29] bots was kind of the wild west ;) [00:22:41] err: Could not retrieve catalog from remote server: Error 400 on SERVER: Puppet::Parser::AST::Resource failed with error ArgumentError: Could not find declared class labs::bots::common at /etc/puppet/manifests/role/labs-bots.pp:2 on node bots-apache-test.pmtpa.wmflabs [00:22:45] is the latest [00:23:23] ah [00:23:24] I know why [00:24:01] added in a review [00:24:35] heh [00:24:41] please use the GPL for the license :) [00:24:49] or some other OSI approved license [00:25:35] GPL is so mainstream, WTFPL is more amusing [00:25:53] * Damianz wonders if he can sed wtfpl to gpl3 [00:26:26] urgh [00:26:33] it takes the module name not the class name in the module? [00:31:34] New patchset: DamianZaremba; "License change" [labs/nagios-builder] (master) - https://gerrit.wikimedia.org/r/26729 [00:31:47] Change merged: DamianZaremba; [labs/nagios-builder] (master) - https://gerrit.wikimedia.org/r/26729 [00:35:34] Ryan_Lane: Something happened to the "Instances for this project" table recently. [00:35:49] FQDN is no longer just the usable hostname for ssh [00:35:53] it now contains both [00:36:11] which means double click selects both, rather annoying when trying to copy it for ssh [00:36:35] e.g. https://labsconsole.wikimedia.org/wiki/Nova_Resource:Integration [00:36:52] integration-wikibugs i-0000049c.pmtpa.wmflabs [00:36:54] it used to be "i-0000049c.pmtpa.wmflabs", not it is " [00:37:04] " [00:37:22] holy fucking shit [00:37:24] this works [00:37:27] heh [00:37:40] Krinkle: hm. that's weid [00:37:43] weird [00:37:46] * Damianz should swear less so sumannah doesn't tell him off [00:38:02] https://labsconsole.wikimedia.org/wiki/Special:Browse/Nova_Resource:I-2D0000049c [00:38:09] that's because FQDN is being populated twice [00:38:27] * jeremyb waves sleepily [00:38:28] because a host can totally have 2 fqdns :D [00:38:48] that too [00:38:54] it does actually have two FQDNs [00:38:54] wow, so many accts to create [00:39:15] we need to work towards eliminating the instance id hostname [00:39:31] so that only the instance name is used [00:39:42] we need to purge puppet keys for that, though [00:40:11] yeah, btw that causes issues with the snmp traps in nagios as well as sudo policies etc [00:40:36] and it can't really have 2 because the ptr will only ever be one (unless you're like some of the servers I use to manage that had 3 ptrs...) [00:40:44] fucking useless company those where colo'd at [00:40:57] were* ;) [00:41:13] jeremyb: sp333leeeeeeing doesn't count at nearly 2am :D [00:41:24] it's not nearly 2am though [00:41:34] It's 1:41, that's nearly 2am! [00:42:08] 05 00:41:34 < Damianz> It's 1:41, that's nearly 2am! [00:42:19] it's going to be a decent amount of effort to only use the instance names [00:42:20] You're just in the past [00:42:29] TZ=wikipedia.org [00:42:39] sudo policies in labsconsole are broken because we're inbetween the two right now [00:42:39] Ryan_Lane: I'm sure nova will change them in 6months anyway =D [00:42:43] hahaha [00:42:44] true [00:43:00] Though [00:43:10] Now you have salt on most instances, sorting puppet should be easyish. [00:44:23] So these work, if anyone wants to see crappy manifests I'll happily take general review of https://gerrit.wikimedia.org/r/#/c/26441/ (or opinions on what software we should support as standard on bots). [00:44:39] * Damianz goes to find ice cream before bashing out some misc fixes [00:53:02] hah, Test User! [00:53:29] Test User is an actual person, he looks a lot like Ryan. [00:53:41] :) [00:53:47] he is in fact Ryan [00:54:04] I need to test things live every once in a while [00:54:22] jeremyb: While you might have a lot I made it sexy simple now! It's like 3 clicks instead of 6 clicks. [00:54:41] Damianz: what? [00:55:54] (doesn't apply to base accounts, but then they're boring) [00:55:59] Change on 12mediawiki a page Developer access was modified, changed by Jeremyb link https://www.mediawiki.org/w/index.php?diff=590652 edit summary: all done [00:56:16] Damianz: i still don't know what you mean [00:57:06] Auto adding to users to bastion on adding them to the shell group, which saves clicks. Only really applies for the next account steps, base accounts are boring. [01:07:28] Ryan_Lane: Do you have anything against changing the page titles on the wiki to be 'Project' and 'Instance' over the silly nova related name they are now? [01:07:46] well... [01:07:50] it's a different namespac [01:08:00] we'd have to change a lot of crap [01:08:10] a few bots. OpenStackManager [01:08:15] the nova configuration [01:08:56] a bunch of templates [01:08:59] urgh, well changing the namespace and using 2 would be awesome but I thought it was possible to just change the titles and leave the namespace alone [01:09:03] let's make a list of things we need to do [01:09:08] I'd rather use no namespcae [01:09:11] though moving the namespace would be nice long term [01:09:34] is that even possible? due to what we have to pull from the api [01:10:10] something lile //instances //docs //sal would be nice structure wise but I didn't think that was do-able due to the way mw uses pages. [01:11:26] Krinkle: the property is defined in DNS name and in FQDN [01:11:32] I can fix that [01:12:07] I was looking at the stuff we're pulling from the api for the instance pages currently, which I'd rather move to SMW includes and keep that data updated so we don't split it. Though it seems rather urgh currently. [01:12:33] instance page (correct) does | [[FQDN::{{{FQDN|}}}]] [01:12:41] Project page does ?FQDM [01:12:48] how can they give different results [01:13:03] hm. [01:13:08] https://labsconsole.wikimedia.org/w/index.php?title=Template:InstanceStatus&action=edit [01:14:26] fixed [01:14:33] now it'll just show the instance-id fqdn [01:14:37] which is kind of non-ideal [01:14:53] I'd prefer it show the instance name fqdn [01:16:17] that's fine two, as long as it is a fqdn for ssh (with the local config to bind wmflabs to bastion.wmflabs) [01:16:21] too* [01:16:29] i.e. one that work [01:16:30] i.e. one that works [01:16:46] yeah [01:16:55] Ryan_Lane: wait, so integration-wikibugs.pmtpa.wmflabs should work? [01:16:59] (or already works?) [01:17:00] of course [01:17:07] that's awesome [01:17:13] it's also annoying [01:17:16] I never even tried, because that string isn't shown anywhere [01:17:23] (from an integration point of view) [01:17:31] I can actually remember those :D [01:17:34] Damianz: what do you mean? [01:17:41] Krinkle: yep. that's the idea [01:18:01] Ryan_Lane: Multiple names for instances that's not simple to programatically tell which is being used == bleh [01:18:21] andrewbogott: can you make the fqdn field in your template put the FQDN rather than just the hostname? [01:18:29] on the bright side nagios is using the fqdn now so won't break when we bring eqiad up [01:18:31] he's likely not around at this hour :) [01:18:36] Damianz: true [01:18:58] well, the goal is to kill off the instance id host names [01:18:58] I swear I can't tell when andrewbogott is around and not, when he's here he randomly doesn't answer and when he's afk or in a meeting he randomly answers! [01:19:05] heh [01:19:39] he pops in with work magically done, and that makes me happy [01:20:14] well they do say if you work in engineering users should never know you exist :D [01:22:15] * Damianz wonders how hard it would be to make a generic proxy in puppet [01:23:45] I have a fairly generic one already make [01:23:46] *made [01:23:48] in nginx [01:24:01] look at protoproxy [01:24:16] it takes a hash as configuration [01:24:25] it's meant for ssl tunneling [01:24:40] hm. that's actually one way we could provide ssl for labs [01:24:51] we could get a *.wmflabs.org cert [01:24:58] and use that at the proxy service [01:25:16] I was thinking that the other day, it's the only secure way to do it [01:25:18] it would need to be separated, tough [01:25:22] *though [01:25:26] yeah [01:26:33] that's actually not terribly easy. [01:26:45] I was thinking a hash with that takes and and generates nginx configs and if hostname matches /.+\.wmflabs\.org/ auto support ssl or such. The thing I really dislike about that is there isn't a simple form someone can use to set stuff up and edit it. It has to be reviewed, done in puppet etc. [01:26:53] because we allow any service to use .wmflabs.org [01:27:15] Hm.. `/etc/wmflabs-instancename` isn't on all instances (I was using it in my dotfiles repo PS1). [01:27:19] Ah, $INSTANCENAME [01:27:21] wonderful [01:27:33] Krinkle: new instances don't have that problem [01:27:53] new instances' hostnames are the actual hostname, not i- [01:28:07] we'll switch the old ones over in the future [01:28:27] I suppose the default .bashrc will also be updated then? [01:28:35] (as it would no longer have to work around it) [01:28:45] doubt it [01:28:52] nah [01:28:56] we'll leave that there [01:28:57] it currently has a workaround in place that grabs it [01:29:01] wouldn't want to break people's stuff [01:29:12] I mean for new instances [01:29:42] but you're saying HOSTNAME and \h in PS1 will equal the instance name? [01:30:43] yeah [01:30:52] it already does on new instances [01:36:39] hmm you know what, I can do this bots crap with 1 template if I re-gigg the manifest for userweb [01:36:42] * Damianz notes todo tomorrow [01:46:15] http://i-000003a9.pmtpa.wmflabs/ is pretty sexy, shame about the dhcp spam [02:57:34] Damianz: If my name says 'afk' that means my laptop is closed, and if it doesn't say 'afk' that means my laptop is open. But of course this evening my laptop was open but I was miles away from it, so the system could use some refining. [02:57:45] I also keep semi-normal 9-5 CST hours, if that helps :) [02:58:28] We could tag you with a gps becon and plot your movements on OSM to solve this issue. [02:59:35] Oh hey you're only like -5 rather than -7, awesome [02:59:38] Well, you'd need a beacon on me /and/ on my laptop, because the interesting thing is the delta. [03:00:55] Also, the andrewbogott_afk means that conversation is logged and I get beeped when I come back if you talk to/about me [03:03:04] Dunno, I've had to fix a server using ssh from my phone before... that was annoyingly slow. [03:06:42] * Damianz should stop looking at skills courses or this is going to get expensive =/ [03:18:26] * Hazard-SJ sighs [03:19:09] For SSH I should use bastion.wmflabs.org as the host name and 22 as the port, right? [03:27:12] Hydriz? [03:27:24] yes? [03:27:36] For SSH I should use bastion.wmflabs.org as the host name and 22 as the port, right? [03:28:13] yes. :) [03:28:21] usually you don't have to define the port [03:28:42] Hydriz: But I'm having a problem logging on :( [03:28:59] here to help, tell me what you are getting [03:30:05] No, let me rephrase that: Tell me the error that outputs when you SSH in [03:30:49] 10/05/2012 - 03:30:49 - Creating a project directory for stackfarm [03:30:50] 10/05/2012 - 03:30:49 - Created a home directory for hydriz in project(s): stackfarm [03:30:53] When it says "login as: ", I enter "hazard-sj", then a message pops up saying "Disconnected: No supported authentication methods available" [03:31:03] oo [03:31:12] that looks like something different I have seen [03:31:57] have you tried Googling that? [03:32:01] No [03:32:26] Google is your best friend, use it :P [03:35:43] 10/05/2012 - 03:35:43 - User hydriz may have been modified in LDAP or locally, updating key in project(s): stackfarm [03:37:43] !log stackfarm Project created for testing of an extension to manage wikifarms (both existing and new). [03:37:44] Logged the message, Master [03:38:47] Hydriz: I got in some other way [03:39:20] heh [03:39:22] magic? :P [03:39:46] that means your key isn't being used [03:40:01] oh Damianz! [03:40:04] Yes, I asked Hermione Granger for the spell :P [03:40:06] I got a message for you from ops [03:40:08] well usually, -vv would tell you [03:40:19] Does it say remove the fucking crons? [03:40:24] yeah [03:40:26] If so, Ariel asked me earlier [03:40:28] bingp :P [03:40:30] *bingo [03:40:50] Got to leave a mark in ops' inboxes :D [03:41:00] I spammed them too [03:41:07] I had one per minute cron [03:41:19] I have a 1min, 5min and once a day crons going [03:41:22] s/have/had/ [03:41:24] but I think Gmail can handle these emails well I believe [03:41:26] I logged on from WinSCP then opened PuTTY from there [03:41:35] oh my god [03:41:41] talk about going around the houses [03:41:47] just configure puppet/pagent [03:41:54] * Hazard-SJ wonders how he'll get petan [03:42:13] Just pray for a few days and he appears like magic [03:44:29] I sent him a memo on the 28th of September, he read it the 2nd of this month, but no reply. Doesn't he usually idle, btw? [03:44:39] sometimes [03:44:53] sadtimes when the only indoor climbing club you can find is for lonely gay guys [03:51:58] the long-suppressed beatles album, a follow up to sgt. pepper's [03:52:13] "The Lonely Gay Guys' Indoor Climbing Club" [03:54:35] They're first line on the frontpage was more like 'The UK's most social group for gay and bisexual men who love the great outdoors.' [03:55:56] On a better note I just found out if I take out gym membership I get free access to the climbing wall *profit* [05:32:05] and the toolserver is offline again, yay *sigh* [07:34:14] aude: I think I nearly have a first version of puppetizing the OSM rendering stack done [07:36:23] For the moment it is still a mess and I still need to verify that it was the puppet rules and no manual steps, but the maps-test2 instance does seem to render and server tiles now [07:44:05] hey [07:44:13] @trusted [07:44:13] I trust: petan!.*@wikimedia/Petrb (2admin), .*@wikimedia/.* (2trusted), .*@mediawiki/.* (2trusted), .*@wikimedia/Ryan-lane (2admin), .*@wikipedia/.* (2trusted), .*@nightshade.toolserver.org (2trusted), .*@wikimedia/Krinkle (2admin), .*@[Ww]ikimedia/.* (2trusted), [07:44:23] effective way to ping all people :D [07:44:35] Krinkle|detached ping [07:58:36] Change on 12mediawiki a page Developer access was modified, changed by Luft-on link https://www.mediawiki.org/w/index.php?diff=590790 edit summary: [07:58:49] apmon: nice [07:59:20] owwie my head hurts [08:00:04] Once I have it fully working, someone needs to tell me how to do it properly with roles and classes and templates... [08:00:11] Change on 12mediawiki a page Developer access was modified, changed by Luft-on link https://www.mediawiki.org/w/index.php?diff=590792 edit summary: [08:00:12] ok [08:01:03] and how to proceed after that [08:01:09] * aude nods [08:01:40] Hopefully I'll be able to get the first stage done by the weekend [08:01:53] * aude is willing to take a look [08:01:54] Change on 12mediawiki a page Developer access was modified, changed by Luft-on link https://www.mediawiki.org/w/index.php?diff=590794 edit summary: /* User:Luft-on */ [08:02:13] with mod_tile? [08:02:19] yes [08:02:25] ok [08:04:03] It is all currently on maps-test2 [08:04:20] that's fine [08:04:40] * aude will be back in ~30 min [08:05:15] I should be going to sleep now. So I won't be back till about 6 hours or so. [08:05:29] ok [08:52:23] http://securityreactions.tumblr.com/ [09:35:38] !log wikidata-dev wikidata-dev-3: is up and running again with editing / account creation restrictions and a fresh database. Note to self: When installing: git clone MW core, then set up MW in browser, THEN git clone the extensions (and not before!). Otherwise the Wikidata extension might not be seen during MW installation, afterwards no import of items possible. [09:35:40] Logged the message, Master [09:46:52] @seenrx Quent [09:46:52] petan: Last time I saw Quentinv57 they were quiting the network N/A at 10/5/2012 5:57:13 AM (03:49:38.8665940 ago) (multiple results were found: Quentinv57_, Quentinv57__) [09:51:51] petan: Are you available to help me debug a MySQL error? [09:52:22] what's the 37707 [09:53:14] ? [09:53:33] error [09:53:55] lol [09:53:58] er... [09:54:14] ERROR 1033 (HY000) at line 472: Incorrect information in file: './wiki/templatelinks.frm' [09:54:43] thats all I am getting [09:54:52] while accessing the wiki [09:55:12] or rather, doing the mysql command itself [09:55:33] urgh, that usually means your schema is corrupt/dodgy [09:55:39] just checking, where is your mysql data stored? [09:55:47] project storage [09:55:52] ok, don't do that [09:56:02] oh? why? [09:56:04] gluster has locking issues that leads to data corruption/loss [09:56:10] sigh [09:56:11] affects mysql [09:56:18] great, many thanks on that :) [09:56:44] Which reminds me, I wonder if Ryan fixed the types so I can create a 50gb instance yet for bots sql [09:57:30] flavours? [09:57:58] oh god, Wikimedia config files are so unusable [09:58:11] I spent almost the whole afternoon trying to fix all the problems [10:00:51] hmm... where is the unconfigured domain file? [10:01:40] mhm [10:02:09] the cluster mediawiki config ... mess... really needs tyding. I think everyone basically went 'I'm not breaking everything' and left it. [10:02:38] It needs _alot_ of configuring and fixing before you can use it for other wikifarms [10:03:15] ...and its magic on how MWVersion (the multiversion script) can be called to CommonSettings.php without other touches. [10:04:20] Imo a large chunk of that config should be handled in puppet and deployed as configs under puppet rather than the weird rsync/nfs/symlink/thing that happens currently but that's probably not going to happen :) [10:05:42] hmm, how do you find the source file for this page: http://jojwq.jqiodwq.beta.wmflabs.org/? [10:06:47] grep -r 'Domain not configured' . [10:07:06] Which reminds me [10:07:12] config viewing stuff is broken on beta [10:07:44] hasher isn't here [10:07:52] * Damianz shrug [10:07:57] Going shopping, bbiab [10:08:06] oh great, thanks for that, found it in default/index.html [10:08:12] oh, bye :) [11:02:11] Change on 12mediawiki a page Developer access was modified, changed by Zeljko.filipin(WMF) link https://www.mediawiki.org/w/index.php?diff=590825 edit summary: [11:35:52] * Damianz magically re-appears with subway [11:41:40] Change on 12mediawiki a page Developer access was modified, changed by Luft-on link https://www.mediawiki.org/w/index.php?diff=590846 edit summary: /* User:Luft-on */ [11:43:25] Change on 12mediawiki a page Developer access was modified, changed by Luft-on link https://www.mediawiki.org/w/index.php?diff=590847 edit summary: /* User:Luft-on */ [14:30:11] !log wikidata-dev wikidata-dev-2: installed package python-imaging for better captcha [14:30:13] Logged the message, Master [15:10:10] @help RC+ [15:10:10] Info for RC+: Create entry for feed of specified page, example @RC+ wiki page [15:10:58] 10/05/2012 - 15:10:57 - Creating a home directory for louisdang at /export/keys/louisdang [15:15:59] 10/05/2012 - 15:15:59 - Updating keys for louisdang at /export/keys/louisdang [15:18:42] New patchset: Hydriz; "Removing all instances of definitions for "hydrizwiki"." [labs/incubator] (master) - https://gerrit.wikimedia.org/r/26805 [15:20:12] @help [15:20:12] Type @commands for list of commands. This bot is running http://meta.wikimedia.org/wiki/WM-Bot version wikimedia bot v. 1.8.23.8 source code licensed under GPL and located at https://github.com/benapetr/wikimedia-bot [15:20:19] New review: Hydriz; "Merged and deployed to live site." [labs/incubator] (master); V: 1 C: 2; - https://gerrit.wikimedia.org/r/26805 [15:20:20] Change merged: Hydriz; [labs/incubator] (master) - https://gerrit.wikimedia.org/r/26805 [15:21:43] @commands [15:21:43] Commands: channellist, trusted, trustadd, trustdel, info, statistics-off, statistics-on, statistics-reset, configure, infobot-link, infobot-share-trust+, infobot-share-trust-, infobot-share-off, infobot-share-on, infobot-detail, infobot-off, seenrx, refresh, infobot-on, seen, drop, whoami, add, reload, suppress-off, suppress-on, help, RC-, recentchanges-on, language, infobot-ignore+, infobot-ignore-, recentchanges-off, logon, logoff, recentchanges-, recentchanges+, RC+ [15:24:42] ^demon: Are you there [15:36:14] <^demon> Jan_Luca: What's up? [15:38:27] Are you the person for asking about labs-repos in Gerrit? [15:39:39] Or do I have to ask the Labs admins? [15:41:00] <^demon> I can do it. Is there a request on-wiki yet? [15:41:26] Yes: https://www.mediawiki.org/wiki/Git/New_repositories/Requests [15:44:26] <^demon> Created. [15:44:58] Thank you, I didn't know if there is a seperate process for these repis [15:47:46] <^demon> No, it's all the same process. [16:01:50] I need a test process that forces the WM install to recreate the admin user and password. I would've thought that deleting the mediawiki database would do that, but it seems not... [16:01:54] Anyone have other ideas? [16:11:34] The admin user is only created when you install MW [16:13:11] andrewbogott:The method is Installer::createSysop() [16:13:59] Jan_Luca: Hm… I see that my question was phrased backwards. What I want is to clear the existing user and password, so that when I rerun mwinstall it recreates it. [16:14:35] Do you mean install.php? [16:14:35] <^demon> Delete the entry from the `user` table [16:15:06] Do you mean running install.php again? [16:15:47] yep. [16:16:00] Which file does the 'user' table live in? [16:16:12] The user table is in MySQL [16:16:30] in the wiki database [16:17:00] When you run install.php again it should create the user again [16:17:22] Hm, clobbering the wiki database didn't help. I must be doing some other dumb thing :( [16:18:18] What install.php command did you run? [16:20:27] I am running /srv/mediawiki/maintenance/install.php testwiki admin --dbname testwiki --dbuser root --pass bananas --server $mwserver --scriptpath '/srv/mediawiki' --confpath '/srv/mediawiki/orig/' [16:20:37] Which would make me think the password is 'bananas' but it seems not [16:22:47] And the script's output said that there are no problems [16:23:04] I think so… running it via puppet so, many levels of abstraction [16:23:28] OK, well, this time I removed the record by hand rather than wiping out the db file, and the behavior is more reasonable. [16:23:35] So, I don't know what that's about, but can move ahead now :) [16:24:17] What I /really/ want is to have the password in the db not encrypted so I can see exactly what's happening. That would probably involve excessive hacking though. [16:25:46] You could add a command in Installer::createSysop() to output the password that goes to db into a file [16:27:12] Yeah, that's probably the next step. [16:27:19] What I'm /really/ trying to do is this: [16:27:28] /srv/mediawiki/maintenance/install.php testwiki admin --dbname testwiki --dbuser root --pass `cat /srv/mediawiki/orig/adminpass` --server $mwserver --scriptpath '/srv/mediawiki' --confpath '/srv/mediawiki/orig/' [16:27:41] Which is to say, setting up mediawiki whilst pulling the password out of a pre-existing file. [16:28:11] There's probably some quoting problem with that [16:28:52] Do not forget the quotes around `cat /srv/mediawiki/orig/adminpass` [16:32:24] hm… that gets me further but is still setting the password to an unknown mystery. [16:33:36] * andrewbogott will now learn enough php to write to a file [16:36:39] who can grant shell these days? [16:36:57] 05 02:21:56 < drdee> jeremyb: can you give the loginviashell right to user 'louisdang' in labs? [16:37:11] * jeremyb isn't 100% sure. i guess either sysop or cloudadmin [16:37:20] jeremyb: I can do it; one moment [16:37:36] andrewbogott: right, but as a more general question as well [16:37:56] Oh… I'm not sure. [16:38:06] btw, louisdang has shell already. You probably knew that :) [16:38:55] Jan_Luca: OK, writing the password out explains everything… the password is now being set to "cat /srv/mediawiki/orig/adminpass" [16:39:01] So my entire approach is flawed :( [16:40:35] andrewbogott: no, didn't know (drdee's response made me think it was still needed) [16:40:45] lemme check bastion [16:41:10] drdee: anyway, keep in mind there is an application process ;) usually good to try that first [16:41:13] andrewbogott, i tried adding louisdang to a project late last night and it gave me that error [16:41:37] well jeremby i always follow the process but things change quite often within labs [16:42:48] drdee: well i didn't think it was so often but i guess i could understand you thinking otherwise [16:42:49] Hm… indeed, when I try to add louisdang to Bastion I get an error. But -- we just added code that should automatically have added them when the shell right was set. [16:42:53] So it might be that all is well, already. [16:43:04] k [16:43:06] that or something interesting is broken :( [16:43:35] ok it works [16:43:37] now [16:43:41] not sure what changed [16:43:51] $ groups louisdang [16:43:51] louisdang : wikidev project-bastion [16:45:45] 10/05/2012 - 16:45:45 - Created a home directory for louisdang in project(s): hadoop [16:50:45] 10/05/2012 - 16:50:45 - User louisdang may have been modified in LDAP or locally, updating key in project(s): hadoop [17:00:02] !log wikidata-dev wikidata-dev-2: tried to configure ConfirmEdit in a way that external links in the user NS trigger a captcha - which doesn't seem to be possible. [17:00:03] Logged the message, Master [17:11:30] Hello. I would like to change my "Full Name" on gerrit (not the username). My current username is "martineznovo", and the "full name" is the same. Is this channel the right place to ask? [17:13:08] Vulpix: A number of people have asked for a name change, including myself, but I was told it's not possible right now [17:13:50] damn... Ok [17:14:43] It's strange since some users have a different full name... maybe they got it right when the account was created :S [17:20:44] andrewbogott: that change looks fine to me [17:20:55] I put an alternative way to create passwords, but I like the openssl way of doing it better [17:21:05] pwgen isn't installed everywhere [17:21:15] Ryan_Lane: OK, thanks. I'm mostly aghast at the triple-layer quoting in that exec. [17:21:35] triple-layer quoting? [17:21:57] oh. in mediawiki_setup? [17:22:05] I think I have a "\"`'string'`\"" in there. [17:22:14] Which makes me think I'm Doing It Wrong [17:22:18] But *shrug* it works. [17:22:23] it's fine [17:22:37] why the need to put php through sh? [17:22:49] To force the backtick evaluation [17:22:56] that looks pretty eveil and potentially dangerous [17:22:58] Otherwise you just get a password that contains backticks [17:23:10] why not put it in double quotes? [17:23:13] then the backtick will work [17:23:31] you can also use $(cat /srv/mediawiki/orig/adminpass) ? [17:23:50] s/?// [17:24:04] DanielK_WMDE: For context, we're talking about this: https://gerrit.wikimedia.org/r/#/c/26812/ [17:24:40] Ryan_Lane: I'm intrigued by but don't understand either of your suggestions. Is the $(cat blahblah) something that puppet extracts and evaluates? [17:25:18] "/usr/bin/php /srv/mediawiki/maintenance/install.php testwiki admin --dbname testwiki --dbuser root --pass \"$(cat /srv/mediawiki/orig/adminpass)\" --server $mwserver --scriptpath \"/srv/mediawiki\" --confpath \"/srv/mediawiki/orig/\"" [17:25:36] no. $() is a bash thing [17:26:19] OK, but… minus the original /bin/sh then bash isn't involved [17:26:29] https://groups.google.com/forum/?fromgroups=#!topic/puppet-users/HXbnqn7yf4E [17:26:38] really? [17:26:46] wow [17:26:47] lame [17:27:13] Lame [17:27:16] well, I guess the current way is fine, then [17:27:20] that's fucking stupid [17:27:24] It seems vaguely correct to me that the command would be split up and passwed to subp (or something) rather than just shoved into sh as a lump. [17:27:37] every single day I get to complain about yet another stupid puppet thing [17:27:37] Although of course I assumed it was doing the latter. [17:28:04] I think it's right the way it is, but exec should take a 'shell' flag. [17:28:12] install.php can't read from a file? [17:28:17] that's annoying [17:28:35] It maybe can, i just don't know how to do it. I was hoping it would pull the password from stdin but it doesn't like that at all. [17:28:46] RoanKattouw: passwords on the command line? [17:28:47] tsk tsk [17:28:56] * andrewbogott reads more about install.php [17:29:43] bah [17:29:44] dbpassfile [17:29:52] there's one for the database password, but not the admin? [17:30:41] Ryan_Lane: ?! [17:30:55] install.php makes you pass the admin password in via the command line [17:30:59] ha [17:31:00] but it'll take the db pass via a file [17:31:23] mediawiki makes me as angry as puppet [17:31:55] I assume that 99% of mw installs are done interactively, probably not a lot of attention on this pathway [17:32:06] true. but we want people to start doing it via the commandline too [17:32:11] well, we wanted to give the option [17:32:19] we also wanted it to be more secure, but alas [17:35:08] Ryan_Lane: So for DNS name it's just ..wmflabs [17:35:12] ? [17:36:54] yeah. I guess we can do it in the template [17:37:05] but, doesn't nova already know this info? [17:38:04] Yeah, it's easy to do it in the notifier. [17:38:11] Just thinking out loud. [17:38:24] * Ryan_Lane nods [17:41:04] * andrewbogott => lunch [17:54:37] Ryan_Lane, thanks for merging. [17:54:39] One thing though, I see only intances are sending logs, not others. [17:54:49] only my* [17:57:14] adminxor: is the port open in your security group? [17:57:18] udp and tcp? [17:58:07] udp 5544 [17:58:29] I think i opened it. Not sure, let me check. [18:01:07] ah..didn't add that port, I opened another port for elasticsearch long back and thought that was it. [18:10:25] did you restart kibana? [18:10:29] I can't connect to it now [18:11:42] I restarted the instance [18:11:48] ah. ok [18:13:27] Why is it that you have to restart instances everytime you add new security rule? [18:14:08] did you reboot it via the interface? [18:14:22] nova has a bug with regards to this, right now [18:15:37] Initially I did reboot from shell. I didnt think the sec rule worked. So rebooted from the mange instance console [18:16:22] Still a "tcpdump -n port 5544" connections from my own instances. [18:16:36] shows* [18:17:15] btw, you may check Kibana now. It's up. [18:18:50] adminxor: so the sec rule isn't working? [18:18:57] the rules take quite a while to apply [18:19:17] I think so, if I didn't do something wrong! [18:19:52] it should be fine [18:20:14] rebooting the instance via labsconsole would cause nova to forcefullly apply its security group rules [18:20:22] which is lame, because it should be doing that quickly [18:22:09] ahh okay [18:22:53] !log wikidata-dev wikidata-dev-2: separated apache log files for English and Hebrew test clients [18:22:55] Logged the message, Master [18:37:06] Ryan_Lane, can I use the bastion host to do a netcat to the logstash instance? [18:39:34] adminxor: for what purpose? [18:41:05] I want to stop the logstash service, open udp 5544 with nc on the instance, then I will do an nc from the bastion host to the instance to check if it works. [19:01:32] Ryan_Lane, I don't think the sec rule has worked, tried to reboot from the console couple of times and Failed to reboot came up. [19:01:45] failed to reboot/ [19:01:48] is the instance down? [19:02:26] Nope, I'm still logged in [19:03:03] i think it's rebooting now [19:03:16] and now up [19:04:27] Awesome!!! Now, it has worked. [19:05:16] Flooded with logs coming from every direction :) [19:13:17] yo yo yo, quick question, is the default group for a labs user 'wikidev'? [19:18:21] Ryan_Lane ^^ [19:22:24] or svn [19:22:27] or something like that [19:22:38] adminxor: heh [19:22:40] adminxor: cool [19:23:05] drdee: gid=500(wikidev) [19:23:10] :) [19:24:03] thx guys [19:39:17] Ryan_Lane, could you please add user 'diederik' to the group 'wikidev' in labs, i am member of svn but all the other folks are member of wikidev and we have certain directories that are group writable by wikidev [19:39:21] i tried sudo addgroup diederik wikidev [19:39:25] but that didn't work [19:39:37] why not use the project group? [19:39:41] and use the sgid bit [19:39:45] and a permissive umask? [19:40:12] yeah that's also a possibility :) [21:22:28] Change on 12mediawiki a page Developer access was modified, changed by Miken32 link https://www.mediawiki.org/w/index.php?diff=591032 edit summary: /* User:miken32 */ [21:24:09] Ryan_Lane, andrewbogott: Do you want a option in install.php for adminpassfile? [21:24:43] Jan_Luca: I don't need it for my purposes but it would generally be an improvement. [21:24:47] Jan_Luca: that would be nice, yes [21:25:00] andrewbogott: well, really, it would make your exec less dirty ;) [21:25:06] true [21:25:06] you wouldn't need to cat in [21:56:10] I have created: https://gerrit.wikimedia.org/r/#/c/26894 [22:03:22] In a labs vm, how can one create a local user with a home directory. Creating e.g. /home/osm gives a "Cannot create directory" [22:19:22] I suspect you'll need to use somewhere else for its home directory [22:19:54] I seem to recall ubuntu for transmission uses /usr/local/transmission or something [22:33:51] apmon: You cannot use /home because it is only for shell users (with labsconsole-account). You have to use a different dir [22:35:05] OK, thanks. I'll just use /tmp/home/osm for the moment until I find a more sensible place to put things [23:08:49] apmon, place it inside /data/project/ ?