[04:31:00] its really fixed this time coren :P last time was all lies :P [04:31:09] no idea why it works though after looking at apache [10:19:56] [bz] (NEW - created by: Krinkle, priority: Normal - enhancement) [Bug 40519] webtools: Setup up webtools.wmflabs.org - https://bugzilla.wikimedia.org/show_bug.cgi?id=40519 [13:09:44] Krinkle: what is difference between tools and webtools [13:15:16] petan: In what context? [13:17:39] Change on 12mediawiki a page Wikimedia Labs/Tool Labs/Needed Toolserver features was modified, changed by MPelletier (WMF) link https://www.mediawiki.org/w/index.php?diff=671139 edit summary: [+228] /* More */ done by default [17:22:45] petan: webtools + bots => tools [17:22:50] petan: It is an old bug [17:22:59] both projects are intended to be merged [17:23:06] does webtools still exist? [17:40:01] Krinkle: web tools does still exist AFAIK [17:41:11] k [17:41:23] Krinkle: btw https://gerrit.wikimedia.org/r/#/c/58281/6/resources/mediawiki/mediawiki.htmlform.js,unified [17:41:35] I got a comment of "Cache $(this) in a variable." [17:41:45] is it poor performance to constantly use $(this)? [17:42:04] my javascript skills are not the best. good to learn :) [17:42:15] Ryan_Lane: $() is basically new jQuery($this) [17:42:28] Ryan_Lane: With a chainable prototype [17:42:39] not caching it and/or not chaining it is wastefull. [17:42:43] so, I should generally be using $() ? [17:42:59] oh sure, never do "new jQuery()" that isn't even valid. [17:43:12] but either make sure everything is in one statement: $().foo().bar().baz(); [17:43:23] or, if you need multi-line, store it in a variable [17:43:27] ahhhhh [17:43:27] ok [17:43:30] makes sense [17:43:41] but don't instantiate a jQuery object for the same selection / elements [17:43:50] $(this).parent().append( $container ); [17:43:50] 103 +» » $(this).remove(); [17:44:02] php has ruined me. [17:44:11] (sorry, didn't mean to paste that) [17:44:15] where chaining is likely to break your app :D [17:45:00] thanks for the info [17:45:15] in javascript it comes naturally, since it used for native stuff as well. [17:45:55] indeed [17:46:00] "foo".toUpperCase().toLowerCase().substr(0, 2).split("").join("1") == "f1o" [17:46:21] even changed type to Array and back to String in there. [17:47:06] scary inline duck typing [17:47:16] No, not really. [17:47:20] not in this case anyway. [17:47:35] ah, right because the array is a string [17:47:50] .split() is a method on the String prototype, it returns an array [17:47:55] * Ryan_Lane nods [17:48:22] I need to do some javascript tutorials [17:48:44] .join() is a method on the Array prototype, returning a string. .join on a string would throw a TypeError. Not because join requires an array but because the .join method doesn't exist on a string. [17:49:15] I see [17:49:20] This is one of those good parts of javascript that are really quite intuitive and efficient. [17:49:27] (once you get them) [17:51:04] indeed [17:51:20] well, I guess I'll push in another change for this [17:51:25] oh. btw... [17:51:29] the jenkins tests are odd [17:51:47] is it really necessary to define all vars at the top of a scope? [17:51:52] in the same statement [17:53:01] because I was getting high priority warnings on "too many vars" [18:28:12] Ryan_Lane: Ignore the priority thing that is a Jenkins hack. If you look at the console output there is no priority there is only warnings we enabled and warnigs we decided to disable. [18:28:29] so I didn't need to shuffle all of my damn variables around? :( [18:28:34] In Checkstyle jshint outputs both enabled and disabled. and calls one high and the other normal priority. [18:28:57] Ryan_Lane: You did have to yes, because our jshint configuration and our coding conventions says so. [18:29:02] and that's a good thing. [18:29:04] ah [18:29:04] ok [18:34:21] Ryan_Lane: I'm going to disable checkstyle in a minute, going back to simple console output for jshint. [18:34:53] * Ryan_Lane nods [18:34:56] the checkstyle plugin is making it harder not easier. It adds 4 pages of useless text and the one page we care about (supposedly) "high" priority doesn't even show the errors immediately [18:35:00] you need to click "Details" first [18:35:03] heh [18:35:13] and then it shows a bunch of "No description, update checkstyle" [18:35:14] yeah, it was a pain [18:35:23] i.e. not helping :) [18:43:55] Coren, andrewbogott: check out the new project filter: https://wikitech-test.wmflabs.org/wiki/Special:NovaInstance [18:44:11] give me some feedback :) [18:44:41] I'd really like the form to be: Project [18:44:46] in a single span [18:45:07] well, horizontally aligned, not in a single span element [18:45:44] Ryan_Lane: I see a big field of buttons and no checkboxes… that's what you're going for? [18:45:58] you do? [18:46:02] I see how to remove things from the filter but not how to add them… when I click the 'x' that item is gone from the filter, seemingly forever [18:46:03] which browser? [18:46:10] ah [18:46:11] yeah [18:46:13] that's correct [18:46:16] click inside of it [18:46:31] Ah, I see! Drop-down [18:46:45] It also flickers a field of checkboxes on loading but then the checkboxes vanish [18:46:55] yeah [18:47:00] that's because there's a fallback [18:47:08] and this javascript is loading at the bottom [18:47:25] I should probably add a hide at the top [18:47:31] ok, makes sense [18:47:48] hm. I don't think I can add a hide at the top [18:48:07] since the content wouldn't be loaded yet [18:48:25] I think that the new interface would me much more useful for me. But it's also sort of unfamiliar and unobvious... [18:48:33] Krinkle|detached: any way of making that happen? [18:48:38] I've been trying to think of how this kind of filtering is done on other sites… usually it's in a sidebar, right? [18:48:50] Ryan_Lane: Are the users gone? [18:48:51] (Not that we have room for a 2nd sidebar) [18:49:19] Krinkle|detached: there's some content I'd like to hide immediately when the page loads, so that it doesn't appear then disappear [18:49:24] Coren: users? [18:49:40] andrewbogott: most only let you choose a single project at a time [18:49:53] so, the good thing is.... [18:50:04] andrewbogott: clear out all of the items [18:50:15] Ryan_Lane: It's not taking my password, and says the account doesn't exist if I try to reset it [18:50:18] the first time someone logs in, they'll see that text [18:50:24] Coren: oh, this isn't the production one [18:50:29] Ryan_Lane: here's an example of a sidebar filter: http://www.sears.com/appliances-dishwashers-built-in-dishwashers/s-1020283 [18:50:32] (which I like) [18:50:35] Ryan_Lane: I know. "TEST Coren" right? [18:50:50] andrewbogott: ah. right [18:51:14] the behavior when I clear everything seems right. [18:51:28] that behavior is what people will see when they first use it [18:51:34] What are the prospects for having the filter actually filter the page in real time rather than via 'submit'? [18:51:43] which means afterwards they should understand it [18:51:47] We already use js for the 'collapse' links, right? [18:51:49] Ryan_Lane: Ah, found it. [18:52:02] andrewbogott: yeah [18:52:06] and this is also javascript [18:52:23] Ryan_Lane: Me like. [18:52:31] we can't load the project interface via javascript yet [18:52:34] Ryan_Lane: Especially since I could select bacon. [18:52:34] err [18:52:37] It would be /way/ better if it filtered in realtime... [18:52:40] yep [18:52:41] agreed [18:52:46] we need to make an API for that [18:53:14] Why an API? Can't we just load everything and then hide/reveal according to the filter actions? [18:53:20] * andrewbogott knows nothing about how JS works [18:53:48] the reason for the filter is partially performance [18:53:48] Ryan_Lane: This interface has about 784 milipretties. [18:53:54] Coren: :D [18:54:04] I switched the default back to vector on the interface [18:54:12] Ah, so you don't want to load everything every time, fair enough. [18:54:17] yep [18:54:20] I mean the filter interface. I liked your skin better. :-) [18:54:23] So… overall, this is much better! [18:54:35] there's one thing we can do.... [18:54:58] when elements are added to this list, it fires an event [18:55:09] we can have that attach to the jquery cookie plugin [18:55:16] change the cookie on the client side [18:55:19] then refresh the page [18:55:25] but that's kind of ugly [18:55:29] and it would be slow [18:55:35] Sometimes I see a text-entry cursor blinking in the filter box, which is weird. But probably outside the scope of your work. [18:55:53] we can make an api module that returns the information back in json and let the javascript create the page [18:56:00] that's the better solution [18:56:04] Might be worth it… usually I only care about one particular project when I change something in the filter. [18:56:10] then projects can be requested individually [18:57:29] so, element added -> event -> change the cookie -> make a request for the new projects via ajax -> add the project to the page [18:58:24] we can start with just views [18:58:34] and later we can add actions, which would be awesome :) [18:59:26] anyway, I pushed in a core change for this style of interface [18:59:55] so it'll be a while till we switch to it [19:05:12] andrewbogott: oh. the text-entry blinking is because you can type to filter the drop down list [19:05:20] ah, makes sense [19:05:53] Ryan_Lane: Speaking of; I have a feature request. Can we _please_ have a "admin-" groops alongside project- ones? :-) [19:05:57] Ryan_Lane: BZ? [19:06:05] eh? [19:06:08] for what purpose? [19:06:42] Access control? It seems clear to me that "this instance should not be logged into by non-project-admins" is a fairly typical use case. [19:07:22] So is "that config file editiable by admin-foo group" [19:07:36] etc. [19:07:44] why are folks editing config files? :) [19:07:52] heh [19:08:03] heya Ryan_Lane, you round? [19:08:04] it's kind of a pain to manage the groups [19:08:07] ottomata: yes [19:08:15] but yeah, it would be nice.... [19:08:36] More typical, perhaps, would be "this logfile readable by admin-foo" :-) [19:08:42] Hm. [19:08:45] I've avoided it because it hasn't been very necessary and it heavily increases the number of groups [19:08:58] You know what? Nevermind. It can be handled with a local-admin service group. :-) [19:09:05] ah [19:09:06] true [19:09:13] I just realized that. [19:09:17] that's easier :) [19:09:22] hiii ja so, just wanted to bounce the puppetmaster + clients thing [19:09:28] wanna see what's involved in getting that to work [19:09:37] i could do it manually I think pretty easily [19:09:48] but it would be nice to puppetize it, and I don't know much about our puppet puppetmaster classes [19:10:01] is puppetmaster::labs used somewhere? [19:10:13] yeah. on virt0 [19:10:25] but it can be reused elsewhere, I'm sure [19:10:37] ahhh i see it, [19:10:37] if $is_labs_puppet_master { [19:10:37] include labs [19:10:37] } [19:10:39] that wa shard to find [19:10:39] haha [19:11:29] hm k, its the same as main puppetmaster except for the puppetsigner.py stuff [19:11:32] did you write that? [19:11:47] I did, yes [19:13:17] hmm, i'm looking at that, I think maybe i don't need that, I"m ok with having to manually sign the client certs [19:14:01] doesn't hurt to use [19:14:05] i *think* all I will need is: puppetmaster, and proper puppet.conf and fileserver.conf for the new puppetmster and the nodes [19:14:13] well, does it need to have ldap creds stored in puppet.conf to use that? [19:14:24] ugh. it uses scriptuser [19:14:29] that's not necessary [19:14:31] it can use proxyagent [19:14:36] we should change that [19:14:56] the proxyagent exists everywhere [19:15:09] if we did that we woudln't need the ldappassword in the puppet.conf file? [19:15:59] Ryan_Lane: Anyway of making what happen? [19:16:19] when a page loads, automatically hide content, rather than letting it pop up [19:16:24] and then disappear [19:16:36] Ryan_Lane: What kind of content? [19:16:59] for htmlform I added an optional css class called chosen [19:17:22] it takes the content in the multiselect input (which is a bunch of checkboxes), and converts it into a multiselect [19:17:33] then it uses chosen to turn it into something that doesn't suck [19:17:49] I'd prefer the checkbox content be hidden on page load [19:17:55] so that it doesn't appear, then disappear [19:20:18] I guess jquery will wait till the content is ready before running, so I could just stick it into the head, right? [19:20:49] hm, also, Ryan_Lane, are all analytics project instances in pmtpa? can I create new instances in that project in eqiad? [19:21:02] we don't have a zone in eqiad yet [19:21:28] I think I'll be making that the top priority for the new person we have coming in next month [19:22:18] ah ok [19:22:18] hm [19:22:19] k [19:27:32] Ryan_Lane, does labs have a specific subnet? if so, are subnets per project? [19:27:37] labs instances* [19:27:42] a specific subnet [19:27:46] and it's not per-project [19:27:52] shared pool of ips [19:28:00] cool, is that in a puppet var somewhere? [19:28:10] subnet is 10.4.0.0/21 [19:28:38] it's not in a var [19:28:44] well, it probably is [19:28:50] but it's not handled by $realm [19:29:06] it could probably be added there [19:29:09] err [19:29:11] realm.pp [19:29:29] oh [19:29:31] I take it back [19:29:37] $site will work [19:29:48] /^10\.[0-4]\./ => "pmtpa", [19:29:52] hmm, there are machines that are not in labs that are on that? [19:29:54] /^10\.6[48]\./ => "eqiad", [19:29:56] oh is it all of pmtpa? [19:30:00] yes [19:30:24] 10.4.0.0/21 is specifically labs [19:30:47] hmm, ahh ok, i'm just grepping puppet , i see [19:30:59] great, cool [19:30:59] ok [19:31:00] but we set labs/production explicitly [19:31:07] in $realm [19:31:12] $site is set automatically [19:31:22] I guess we could also set $realm automatically [19:32:17] yeah, but i'm looking at, for example, fileserver.conf for a puppetmaster with clients in labs [19:32:25] to allow more than just 127.0.0.1 [19:32:35] so i'd need that in a var somewhere…which I guess it is [19:32:46] fixed_range => $realm ? { [19:32:47] "production" => "10.4.0.0/21", [19:32:50] pmtpanovaconfig [19:32:51] ? [19:33:04] $pmtpanovaconfig['fixed_range'] [19:33:06] ja? [19:33:14] oh [19:33:15] no [19:33:21] wazzat [19:33:22] fixed_range => $realm ? { [19:33:22] "production" => "10.4.0.0/21", [19:33:22] "labs" => "192.168.0.0/21", [19:33:22] }, [19:33:56] o.O [19:34:09] production would be the default [19:34:15] and labs would be 10.4.0.0/21 [19:34:29] oh [19:34:30] right [19:34:31] * Ryan_Lane sighs [19:34:43] so. that's confusing [19:34:51] that's if that manifest is applied in labs or in production [19:34:53] right [19:35:04] it's configuring labs, so it's a little meta [19:35:04] and the nova config actually runs in prod [19:35:05] yeah [19:35:06] i get it [19:35:30] dmz_cidr? [19:35:36] heh [19:35:50] that looks more like it, no? [19:35:50] it's best to ignore the openstack config :) [19:35:58] dmz_cidr => $realm ? { [19:35:58] "production" => "208.80.155.0/22,10.0.0.0/8", [19:35:58] "labs" => "10.4.0.0/21", [19:35:58] }, [19:36:11] I wouldn't assign specific IPs to production [19:36:16] I would make production the default [19:36:21] and set IPs for labs [19:36:27] and for anything not production [19:36:47] ? [19:37:06] i didn't write that [19:37:28] i'm just looking for the subnet defined somewhere in avar that i could access from puppetization of a labs instance [19:37:39] instead of hard coding it [19:38:37] what is 192.168.0.0/21 ? [19:51:32] ottomata: I'd hard code it [19:51:40] ottomata: that's a fake range [19:51:47] 192.168.0.0/21 [19:52:01] k [19:52:11] we can use the hardcoded value elsewhere to replace the other uses [19:52:52] petan, coren, I have some morebots instances that are hogging entire instances… should I move them to the tools project and use the grid engine, or is there a different better place for them in bots? [19:53:09] can I make a variable in puppetmaster::labs ? [19:53:20] or meh, no this is more of a labs thing, not a puppet thing [19:53:29] andrewbogott: "hogging"? You mean they are alone 'wasting' the instances? [19:54:11] yeah… right now the bot is made to run as a service. I presume that it should be more properly redesigned to run on a grid instead [19:54:44] andrewbogott: It should be trivial to do, even. If it's a self-contained stript, just doing a 'jstart the_script' will have them run indefinitely. :-) [19:55:48] Coren: OK, sorry, my first question is: tools or bots? [19:56:05] …and then my second question is: is there a design guide for how a bot and/or tool should be managed? [19:56:34] andrewbogott: Well, 'bots' is designated as the experimental/looser standard project; 'tools' is the expected home of stable tools. [19:56:57] andrewbogott: Question 2: https://tools.wmflabs.org/?Help [19:57:01] andrewbogott: I just noticed that we're missing a message: https://wikitech.wikimedia.org/wiki/Special:NovaServiceGroup [19:57:01] ok, Ryan_Lane, I'm starting to think this labs puppetmaster client thing is a bit of work to get right, especially because of the ldap integration [19:57:08] err [19:57:21] <novaservicegroup> [19:57:33] i konw how to do puppetmasters in the usual way, never done much with the ldap stuff [19:57:38] Ryan_Lane: OK… I'll fix that shortly [19:57:41] andrewbogott: Of especial interest to you would be the 'gridengine' section. :-) [19:57:47] Coren, is that help paged linked to from wikitech in some obvious place? [19:58:08] andrewbogott: From the project's page, IIRC [19:58:26] Hah. No it's not. This needs fixin' [20:00:03] fix't [20:01:31] andrewbogott: There are certainly tons of holes in that doc; don't hesitate to point them out. :-) [20:04:08] ottomata: it's the same, roughly [20:04:17] ottomata: except that ldap is being used as the node [20:04:27] and that's configured from wikitech [20:06:37] hmm, riiighhhhht [20:07:24] does puppetmaster::self need a puppetmasterd running? [20:07:26] Ryan_Lane: ? [20:08:03] Coren: What do you think about some screenshots for the docu? [20:09:33] Jan_Luca: ... screenshots? Well, I don't /mind/ them, but do you mean like for putty and such? Because everything else is console based, really. :-) [20:09:57] ottomata: yes [20:10:00] Coren: yes [20:10:24] ottomata: node definitions will override ldap, btw [20:10:38] https://tools.wmflabs.org/ has a landing page now. Useful content will come, eventually. [20:10:46] oh that's good to know [20:10:47] just make sure you add the necessary vars/classes, or your instance will probably break ;) [20:11:50] Coren: How do you generate the tool list? [20:12:20] Coren: it would be good if there was a link to the tool/bot documentation on wikitech from there [20:12:34] yea, doing with ldap is cool too. hm [20:12:45] Jan_Luca: Dumbly; I enumerate service groups and provide a link to its webservice if there is an index present in it. [20:13:01] ah, ok [20:13:37] Ryan_Lane: I was presuming that tool authors would do so from their own tool's index.* [20:13:48] * Ryan_Lane nods [20:14:21] Ryan_Lane: I'm also going to add a column with the sanitized contents of ~tool/.description there though. [20:15:08] So authors can add links to docs, or bot's homepages, or whatever. [20:15:22] [20:15:36] Damianz: Please note sanitized above. :-) [20:15:59] Maybe you where just going to sanitize the ponies [20:16:15] markdown -> html would be cool for that, github readme's for tools :P [20:16:24] sorry... forgot we hate github around here #troll [20:17:10] Coren: something somewhere is automatically resetting the owner of my folder I was using for ~addbot ;p [20:17:23] o_O [20:17:26] On tools? [20:17:36] my cheap crappy hack no longer works and I have subsequently moved my folder :P [20:17:39] and no, on bots :P [20:17:57] it will change it to the correct owner and permissions [20:18:03] it's a thing [20:18:09] what if there is no owner? ;p [20:18:24] there has to be an owner [20:19:08] * addshore nammed a folder 'addbot' in /project/public_html, with the owner 'addshore' (which doesnt work and gets un set) ;p [20:19:22] yeah [20:19:30] addbot isn't a valid user [20:19:51] sad times :P it worked for a week or two :P [20:19:55] andrewbogott: I think there is a bug when tring to add a service group and you are not allowed to do that: I get the error: You do not have permission to manage OpenStack projects and roles, for the following reason: The action you have requested is limited to users in the group: Bureaucrats. [20:20:11] I think the right error should be that you need to be a sysadmin [20:20:24] Jan_Luca: Thanks, I'll have a look. [20:20:38] also Coren I spotted somewhere that dbreplication will only be on tools to start with? :) [20:21:04] addshore: Yeah, tools is going to be the guinea pig before it gets deployed to other projects. [20:21:28] Since it's a hacky solution - the proper one would be to use web space as a service - the only way to remove people removed from the project (ie maybe suspended) is to enforce real users [20:21:32] * addshore needs to start sticking some stuff on tools then ;p [20:21:35] I can add folders to the whitelist (such as wm-bot) though [20:22:28] Damianz, could you please add 'addbot'? :) (I dont imagine there is ever going to be a user created for it :P [20:22:36] also where is the whitelist? ;p [20:22:48] in a secret file [20:23:19] mwhahaha :P [20:24:52] https://github.com/DamianZaremba/labs-bots-vhost-builder/commit/c043a1dd0fe33d3d49f76a44020c61a3004adf49 [20:25:17] <3 [20:25:20] cheers :) [20:25:42] 2013-04-09 20:25:23,417 [INFO] wmib appears to be bad, disabling [20:25:43] 2013-04-09 20:25:23,427 [INFO] testuser appears to be bad, disabling [20:25:43] ^ Anyone placing a claim on those? [20:26:45] Ryan_Lane, where did you see <novaservicegroup> [20:26:47] ? [20:26:58] andrewbogott: Special:SpecialPages [20:27:32] Damianz: you should make it copy the previously set owner to the output before it changes it to root? ;p then you know who to poke ^^ [20:28:07] Coren: I'd like to start the migration of tools-project volume soon. the only disruptive phase is when I do the commit [20:28:11] since you have to be root to write to that directly and I don't grant humanity with the sense of intelligence it's probably root to start with [20:28:15] but I can only run one migrate at a time [20:28:27] and I have no clue how long it'll take ;) [20:29:05] It should be smallish. IIRC, the only disruption is a few seconds of filesystem access loss? [20:29:12] yep [20:29:50] Go, then. The jobs that are currently running are bound the the DB; it's one of our last chances to do it outside of a planned maintenance. [20:29:57] ok [20:30:38] Damianz: good point ^^ [20:31:26] Hmm do I get off my lazy ass and fry steak or order pizza... the former requires more standing, the later more walking... [20:31:37] steak [20:32:15] mmm steak 2 nights in a row sounds like an idea [20:33:14] steak is always the right choice :) [20:34:08] Coren: ok, ready to commit [20:34:23] Ryan_Lane: Make it so! [20:34:38] done [20:34:43] I'm going to do home next [20:36:19] As far as I can tell, it went without a hitch. [20:36:34] * Ryan_Lane nods [20:36:45] it should be mostly unnoticable [20:37:11] files that people are actively writing to probably got errors for a couple seconds [20:41:41] Coren: committing home [20:41:49]