[00:06] hazmat: *sweet* wasn't sure if it was done yet :) [00:37] jcastro: http://news.ycombinator.com/news .. I see you at #6 [00:38] Yeah, way to make front page :) [00:39] <_mup_> juju/relation-hook-context r522 committed by jim.baker@canonical.com [00:39] <_mup_> Tests re visibility of changes in relations from an invoker [00:41] SpamapS jcastro it also made the ycombinator twitter feed: https://twitter.com/#!/newsycombinator/status/187314809887395840 [00:43] <_mup_> txzookeeper/managed-watch-and-ephemeral r53 committed by kapil.foss@gmail.com [00:43] <_mup_> connection err trapping for managed conn [00:44] <_mup_> txzookeeper/managed-watch-and-ephemeral r54 committed by kapil.foss@gmail.com [00:44] <_mup_> add .bzrignore [00:54] <_mup_> txzookeeper/managed-watch-and-ephemeral r55 committed by kapil.foss@gmail.com [00:54] <_mup_> remove unesc. expire handler from retry client [00:55] <_mup_> juju/relation-hook-context r523 committed by jim.baker@canonical.com [00:55] <_mup_> Fix typo and old comment [01:16] <_mup_> juju/relation-id r501 committed by jim.baker@canonical.com [01:16] <_mup_> Merged trunk (to make lbox handle old prereq; this branch is already merged) [01:37] SpamapS: marcoceppi: thanks for the coverage while I was at taco night. [01:37] yes, there is such a thing as taco night. [01:37] hazmat: did you say constraints just landed? [01:55] np jcastro [02:07] jcastro, yeah.. last friday [02:08] hazmat: woo, tequila time! [02:08] jcastro, you can set constraints at bootstrap or when deploying a service [02:15] <_mup_> juju/trunk r509 committed by kapil.thangavelu@canonical.com [02:15] <_mup_> [trivial] change deploy output when using charm store [r=niemeyer] === shazzner-away is now known as shazzner === poolie_ is now known as poolie [03:40] marcoceppi: Marco, I'm checking out that sourceforge php download script. When it's set to https it doesn't work, gives ssl error. Any ideas? [03:41] specifically this: http://paste.ubuntu.com/914043/ [03:59] <_mup_> juju/relation-hook-context r524 committed by jim.baker@canonical.com [03:59] <_mup_> Display parent context relation ident, if available, when logging relation settings [04:04] <_mup_> juju/managed-zk-client r510 committed by kapil.thangavelu@canonical.com [04:04] <_mup_> agents use managed conn [04:04] <_mup_> juju/relation-hook-context r525 committed by jim.baker@canonical.com [04:04] <_mup_> Merged trunk [04:45] * benonsoftware wonders if the topic should change === almaisan-away is now known as al-maisan === al-maisan is now known as almaisan-away === almaisan-away is now known as al-maisan === shazzner is now known as shazzner-asleep === al-maisan is now known as almaisan-away [10:13] I have a half-formed question [10:14] I can see how juju is useful for deploying services, and (for me at least), especially so for deploying dev instances of things I'm working on [10:14] well, at least once I'm allowed to target precise. [10:15] but is there some way to use it, or the charms or something to make setting up development environments easier? [10:15] sure [10:15] <_mup_> Bug #973260 was filed: Unable to deploy charm that uses symbolic links from the charm store < https://launchpad.net/bugs/973260 > [10:15] its not really streight forward atm but yea, just create the charm to install all the dev env tools [10:15] and maybe a current copy of the code [10:16] and deploy or even deploy local to a lcx [10:16] lxc [10:25] hmm [10:26] imbrandon: some differences that I can think of: [10:26] k [10:26] - maybe want to have fewer units than on prod [10:26] sure we do that on omg [10:26] - want to restart services way more frequently (as part of a dev cycle) [10:26] stage has far fewer and even runs on micros vs larges [10:26] - want to have more verbose logging and some possibly insecure debug options [10:27] config option [10:27] - would want to have dependencies for running tests installed [10:27] (and I guess for compiled languages whatever you need to do a build) [10:27] sure all of thats doable now, the last bit i would religgate to a ci server [10:28] but yea [10:28] well, I can't write code without running tests. [10:29] sure i just wouldent have it on the same box, e.g let jenkins run the tests [10:29] imho [10:29] i mean you CAN, that is just prefrnce [10:30] imbrandon: that's generally not effective. Lots of projects deploy have long-running test suites, so while I'm TDDing I need to run a subset of the tests. [10:31] projects *I* deploy, that should read. [10:31] thats a bug in the system imho , if you have long runnign test etc then sure i have lots of sites that do that but they run dedicated testing instances in addition to dev that use the ci server to mamnage the test runnign failures metrics etc [10:31] e.g you end up with local dev, test env, dev env, qa, prod [10:32] ci server manages the test env and long running cont test there [10:32] spell it out for me, you're in your dev env and you want to fix up bug. You write a test that reproduces the bug. How do you run that test? [10:32] you keep dev and qa and prod as close to the same as possible just more/less instances [10:33] you check in the code to vcs for the test [10:33] the ci server picks ut up and runs the test on the test evn that mirrors dev [10:33] ever used travis ci ? [10:34] and then you watch that test fail and then write the fix and then commit that? [10:34] no, I haven't. [10:34] exactly, or you do the test and the code for the fix at the same time [10:34] and check into vcs [10:34] but then you aren't doing TDD [10:34] sure you are [10:34] you;d have to of run the test local [10:34] If you haven't run the test to see if it failed, then how do you know if it's a good test [10:35] OK, now we're getting somewhere :) [10:35] For some projects, it's really complicated to set things up locally to be able to run the tests. [10:35] i'm assuming here that you already have a process and adapting it to thus [10:35] this* [10:35] sure in those cases you do like i said first [10:36] if you dont have a local instance [10:36] and watch it fail or pass etc [10:36] depending on test [10:36] it sound like more work, but its only more work to setup, really its the exact same as you do now [10:37] and once setup its setup for the whole team / multi prohject [10:37] so a one off, its like setting up soyuz the first time, a PITA but its a one off :) [10:38] if done right that is [10:38] and if not its a charm, iterate adn redeploy :) [10:39] wish irc had a white board, would be easier to explain a bit [10:39] lol [10:40] Hmm. Maybe. [10:40] but yea , local dev w/ test ( if possible ) --> check into vcs --> ci server watches vcs and runs test long and short and repeats etc --> passing code goes to dev --> then nightly to qa or whatever the schedule is --> pass qa ( automated qa or live ppl ) --> migrate to prod ( hopefully automaicly for XP ) [10:41] there is a bit more there, but thats the jest [10:41] I've already got CI set up, but any system which requires committing and pushing to run tests is unacceptable, and my goal here is to find a way to make it easier for people to set up environments to run tests locally [10:42] well here is my deal, if it takes more than the developer to commit code ONE TIME to have it automaicly migrate to prod then something in the system is broken [10:42] badly [10:43] thats from many years of iterating this process [10:43] ever read joel on software ? [10:44] Yes, of course. [10:44] yea then bascily how he decribes tdd xp setups [10:44] Reference? [10:44] is how i would/do have it [10:45] sure, let me dig it up, been a while since i linked it :) [10:46] and btw you'd be suprised at the ppl in our field that have never read that :) [10:46] so i had to ask heh [10:52] jml: here is most of it, there is a more specific one i'm looking for though [10:52] http://www.joelonsoftware.com/articles/fog0000000023.html [11:02] bah i cant find it right now [11:02] i need a better bookmark system :) [11:06] short awsner is yes though, no matter what your evn or workflow juju can probably make it easier [11:12] jml: just curious but why "but any system which requires committing and pushing to run tests is unacceptable" i mean thats the goal of a continious intergration server, to be continious this no lagtime thus it dont matter if its local or a spun up juju instace or a vm template in vsphere [11:12] or i'm missing somehting else that you dislike about it [11:13] but yea, in your case i would setup a "prod" juju charm, then have the devs use it localy in lxc containers [11:13] as the test env [11:13] imbrandon: oh, that's easy to explain. [11:14] imbrandon: requiring commit and push to run tests increases the time and effort it takes to run tests, making it more difficult than a system where the tests can be run locally. [11:15] i think thats where we differ, i dont see the extra effort, its miliseconds diffrence [11:15] imbrandon: since I want to run tests very frequently, often after changing just a single line, that is unacceptable. [11:16] ahh see i dont see it that way, but i often commit aftter just one line and then rollup commits for later [11:16] imbrandon: no, it's not. coming up with a commit message takes over a second. [11:16] but thats whats good about a vcs like git that lets you do that whewre as hg or bzr wont [11:16] jml: only on a finaly rollup commit, savr commits and test commits are automated [11:16] imbrandon: well, I guess to each their own, but I'm not adding extra steps to my TDD cycle what amounts to no gain. [11:16] with automated messages [11:17] jml: sure sure i was just reading back and thinking [11:17] not trying to sell ya on another way [11:17] :) [11:17] i do think even with your flow though with minimal change juju could help via the local lxc containers [11:18] but it would take some inital setup of the charm etc [11:19] in that way , they would be more like vm snashots of prod that could have extra things in meta.yaml for when running in dev [11:19] mode [11:19] metadata.yaml [11:19] like the verbose logging etc [11:20] the one hiccup that i think will bite you right away is juju lxc dont persist past reboots [11:20] so you reboot, redeploy local but in reality thats a good thing becuse it makes sure the charm is "ture" [11:20] true* [11:23] have you used like vmware and made a vm then shapshoted it right after everythign was installed, then rolled back to that snapshot regularly ? it would be like that but more "live" as in the charm when "reverted" or deployed is the better word would be like a snapshot as in its clean env, but it would be the latest and greatest version of the code and env [11:23] jml: if that makes more sense i think [11:24] and exactly match prod as far as setup since its the same charm, possibly with other options ticked on or off for logging more etc [11:24] when deployed [11:25] anyhow i may be missing a big chunk of your goal too, i've still not had enough coffee yet this morn [11:25] :) [11:26] that and i think another thing we differed on , is in my initial tdd xp setup i had acceptance tdd mixed in with developer tdd [11:26] that assumption can make a big diff too [11:37] jml: btw if you need/want help with a skeleton charm to get you a proof of concept or something going i'd be glad to help. not the best charmer in the world but yea :) [11:41] imbrandon: thanks. [11:41] imbrandon: I think, realistically, I'm going to wait until our production DC is running precise. [11:41] i hear ya, probably wont be terribly long [11:41] might be worth playing with other simple charms before then [11:41] it had better not be [11:42] just to get feel, took me ages to grasp their usefulness [11:42] mor than like chef or something [11:42] I've already had a bit of a play around. [11:42] :) [11:46] shazzner-asleep: That's weird, it should work [11:46] moins marcoceppi [11:47] imbrandon: o/ [11:48] think i could bribe you into import-id for me on stage.omg at some point this morning ? [11:49] Yeah, in an hour or so when I'm back in the office, I don't think I have the same ssh key that's currently on omg here on this machine [11:50] no worries, just whenever [11:50] ty [11:53] jml: also i'm still very new to the juju stuff too, so Spa*mapS and marco*ceppi etc might have even more detailed insight when they are active ( put the * so irc wouldnt ding them un-needed ) [11:55] been round webdev many years, juju still wrapping my head round, although its clearing up mostly === Guest31547 is now known as tobin === imbrandon is now known as bholtsclaw [13:00] <_mup_> juju/trunk r512 committed by kapil.thangavelu@canonical.com [13:00] <_mup_> [trivial] tweak cli help options for consistency/formatting === spladug` is now known as spladug === almaisan-away is now known as al-maisan [14:12] imbrandon: so, what crack were you smoking that you thought joel on software recommends that you not run test suites on dev machines during development? ;) [14:12] * SpamapS_ is reading backscroll === SpamapS_ is now known as SpamapS [14:12] SpamapS: I have to say the same thought crossed my mind. [14:13] no [14:13] thats not what i said [14:13] i said run it via a ci server in a test env [14:13] imbrandon: its what you said. :) [14:14] imbrandon: jml was talking about setting up a server *to write the code on* not to test the code. [14:14] hazmat, ok I pushed up some doc updates so at least they're not totally wrong, how often does the doc cron generator run? [14:14] huh? that in no way says that [14:14] imbrandon: what you're describing is the stage where the developer believes he has triumphed over the bug/feature/etc. [14:14] <.< >.> [14:15] jml: what you're describing has two good answers in juju right now [14:15] sure after a test has been written and sent to the ci server to fail [14:15] imbrandon: jml wants to write a single test... and then run it and see it fail. Then write code to make it pass. *then* commit and run the whole suite. [14:15] jcastro, i don't see that info in the rt... not sure [14:16] hazmat, ok I guess I'll just find out when it happens, heh [14:16] imbrandon: you don't need a CI server to run *just that one test* [14:16] SpamapS sure and i said thats wrong , no you dont but you should have [14:16] SpamapS, let's finish up the contest today [14:17] "just that one test" and things like it are what make inconsistencies happen [14:17] jcastro: yes, sorry, I was sort of useless yesterday.. should be ready to wrap it up in a couple of hours [14:17] cool [14:18] the only one left to propagate for the contest is znc [14:18] imbrandon: also bzr's whole design is based around rollup commits. :) [14:18] i can write tests and have them pass all day long but it dont mean in a clean env that the ci box ensures it will pass [14:18] lynxman, we just need a thumbs up from you on that one if you're reviewing it [14:18] SpamapS, oh, and openerp of course [14:18] imbrandon: so the idea of commit commit commit commit then make a single merge commit is quite common in bzr land. You can even rebase if you'd like to erase all your crackful iterating commit messages. :) [14:19] SpamapS, ok so let's plan on hanging out this afternoon? [14:19] i've never seen bzr used like that, i'd need to look over the docs its not the norm for people [14:19] imbrandon: to be crystal clear, I'm not saying I don't need or want CI. [14:19] jcastro: 10:00 my time, 13:00 yours? [14:19] jml: i know i;m trying to keep SpamapS_ from putting words in my mouth out of context [14:20] SpamapS, actually let's go post-lunch your time, I have physical therapy today and the wife works late so I can be flexible there. [14:20] SpamapS i dont think you understood what i was getting at [14:20] well obvuously now [14:20] not* [14:21] jcastro: I have to jump on a train at 14:00 my time, so my afternoon will be spent on 3G. :-P [14:21] but no i did not say joel wants you to do it local, infact i said i would have it set up like he describes the tdd xp process [14:21] imbrandon: no I didn't understand at all. :) I read it as "you must commit to run a test" [14:22] sure but if you look at the contect that dosent mean its the final commit to the prod branch [14:22] sure, I didn't think that either. [14:23] jcastro: I wasn't asked to review that one, send it my way if you fancy :) [14:23] jcastro: I only had sbuild [14:23] then yes, you must commit to run a test unless you have a full abstracted local env [14:23] oh [14:23] or your doing it wrong [14:23] lynxman, https://bugs.launchpad.net/charms/+bug/956259 [14:23] <_mup_> Bug #956259: Charm needed: znc < https://launchpad.net/bugs/956259 > [14:23] it should be pretty easy [14:23] imbrandon: I support the idea of a CI server sitting around waiting for a commit to run tests on commit. However, if the full suite takes 5 minutes, it takes 5 minutes for me to find out that I mispelled "websrever"... [14:23] just tell me when and we'll promulgate it [14:24] jcastro: awwight [14:24] bah thats just a bad setup then, it shoudl run on triggers [14:24] imbrandon: whereas if I'm running just the tests I have just written, I can at least have a chance at getting the dumb things right before waiting 5 minutes for the full suite to fail. [14:24] if you cannont garentee that the env is IDENTICAL to the variables in prod then the test is useless local [14:25] imbrandon: we disagree fundamentally there then [14:25] you have never had something work local and fail in production then [14:25] I have [14:25] many times [14:26] not saying you shouldn't have a real test env between dev and production [14:26] then your a gluten for pushishment , fix the porobme [14:26] SpamapS, imbrandon: hey so, go to pm so we can work, thanks! [14:26] I'm saying that a lightweight local test suite *should* in fact prevent you from wasting time. [14:26] imbrandon, fix drupal, SpamapS, go promulgate. That will be all. [14:26] jcastro: eat me, we're answering jml's question about how to create test environments w/ juju [14:26] jcastro: ... this pertains to juju kthx [14:27] heh [14:27] mmmmhh, gluten, *drool* [14:27] lol [14:27] actually, no don't eat me. I am very high in fat. ;) [14:27] * jml goes to get coffee and some kind of pastry with gluten in it [14:28] jml: I have two options for you for setting up dev environments w/o compromising the production charm. buzz me when you're back. [14:28] SpamapS: will do. [14:28] SpamapS: and i'm saying that that *way way* less than 5 minutes waiting for the test to fail is not counter productive, infact i spend more time opening the browser than it would take for the ci to grab it test it and notify me [14:28] and garentees its a clean kill [14:29] imbrandon: launchpad's test suite takes *hours* to run, even with the recent efforts to parallelize it. [14:29] ok then in that case your not talking 5 minutes [14:29] lets keep the same varibles here [14:29] well I'm saying, 5 minutes is too long.. but many large systems would want to run the full many hours long test suite. [14:30] if your running a test suite that takes hours then 5 muntes dont mater and you should NOT be doing it local [14:30] exactly why you have a dedicated test evn [14:30] as well as local [14:30] imbrandon: when I do TDD.. I write a test, run to make sure it fails the way I expect it to, then write the code to make it pass. Its important that I not start writing the code until I see that fail. [14:31] imbrandon: No, I just want to run *one test* ocally. [14:31] locally even [14:31] imbrandon: just for the iterative "prove to me that I have solved this one tiny problem" effect [14:31] sure, but if you run the test and it fails local thats just a CURSORY test, it must fail in the env it was ment for to mean anything [14:32] jcastro: charm looks incomplete, giving some more feedback [14:32] * jcastro nods [14:32] jcastro: jamespage was already reviewing it btw [14:33] imbrandon: well, this gets into an unsettled theoretical area where we decide between unit tests, functional tests, and integration tests. :) [14:33] SpamapS that and i gave the middle ground, i sugested deploy the charm local , that takes care of both really, run you one off i cant wait 5 minutes because i wrote a bad test or typos then push it to the ci server [14:33] imbrandon: if its a unit test, then in theory, it is its own universe, isolated from all the integratino points of the environment. To be fair, for that one, you shouldn't need any special dependencies as jml was asking for. :) [14:34] sure [14:34] but thast not .... ugh you know we're talking about 5 diffrent things here [14:34] :) [14:34] imbrandon: anyway, I like where you're going with this actually, and juju can make this *super* easy. [14:35] right [14:35] and ensure that it matches [14:35] :) [14:36] SpamapS: back [14:36] so, w/ what jml was describing, we have a *great* use case for subordinates. You can have a subordinate "CI" charm which you deploy along w/ the code/app servers, which then gets down to the business of running the tests on commit from your dev server (config option for branch to watch I suppose) [14:37] yea, i was thinking a config oiption since sub dont exist yet [14:37] jml: ^^ option 1 [14:37] and making it all complicated [14:38] SpamapS: I don't quite know what subordinates are, but sure, that sounds good. [14:39] jml: they are charms which don't serve a primary service purpose, and are always deployed along with primary charms. The key is they can relate to things via the filesystem. [14:39] jml: charms have to each be their own "instance" right now, sub;'s let you have 2 or 3 or 6 charms on the same machine, like a nginx charm and a rails charm [14:39] jml: the last branch or two for them is under heavy review right now [14:39] imbrandon: to be clear though, you can't have more than 1 non-subordinate charm on one machine. [14:40] I guess these are for things like plugins, monitoring systems and the like? [14:40] or, in juju-future-ese, on one "container" [14:40] jml: thats the primary use case yes [14:40] jml: also the idea that you could have a node.js charm and then the apps would be subordinates [14:40] SpamapS_: ugh, what happens when i need apache + rails + myapp ? that suks [14:40] imbrandon: rails and myapp would both be subs [14:40] SpamapS: oh, interesting. [14:41] right , but you just said they cant [14:41] subs have their own relations and everything, its really quite elegant [14:41] or i misunderstood [14:41] imbrandon: no, apache would be the primary [14:41] SpamapS: what's the other option? [14:41] kk [14:41] jml: the other option is, as imbrandon suggests, a config option, something like "dev=1" [14:42] jml: that just brings in the things you need, [14:42] and then something in the charm that responds to the config change? [14:42] err, that was a . [14:42] jml: precisely [14:42] yup, a good explination of that is on jcastro lastest post on cloud.ubuntu.com [14:42] where phpmyadmin from upstream or the deb [14:42] etc [14:43] config-changed would do something like "if dev==1 then install testing tools vcs tools etc etc" [14:43] https://juju.ubuntu.com/docs/drafts/service-config.html [14:44] cool [14:45] jml: the idea with juju is you would have the dev deploy as similar to the prod deploy as possible, just with less resources available to it. [14:45] but local to lxc or maybe to a micro [14:46] yeah, local so you can destroy your machine's performance, or t1.micro so you can wait a long long time. :) [14:46] haha [14:47] we do need to figure out if lxc-start-ephemeral can be used.. the shared VFS cache of overlayfs is a huge win. [14:48] * SpamapS goes afk for a bit [14:48] i was using it yesterday on the HPcloud boxes [14:48] i provisioned a 16GB server and then made baby lx boxed [14:50] SpamapS: so when you get back, what did you mean about the multi sub thing, i'm not quite sure i followed unless you just mean one needs to be the "master" [14:51] if thats the case then i'll cheat and make a ubuntu jeos server charm and everything is a sub of that [14:52] lynxman, you are correct - I had provided feedback but I don't think it had been implemented [14:53] * jamespage reads [14:53] oh - it has - there was just a query over 'install' [14:54] jcastro: btw drupal is done and it rocks, now just looking to add some of the cool options to it like marcoceppi demo'd at POSCON [14:54] :) [14:54] imbrandon: it's got a few blockers in it, I'm in the middle of a review right now [14:54] nice, let me know, i've very keen to add them to the drupal charm [14:55] for like drush archive / upstream, drupal 6 or 7 archive, upstream or pressflow etc etc etc [14:55] will make it very nice [14:56] I think phpMyAdmin is the only charm that uses an upstream switch, and it's a bit convoluted in that charm. [14:56] It'd be nice to get this hammered out, then just open a bug against the promulgated charm to add version switching support [14:56] yea there are so many flavors of official drupal that it could benefit, have a sane default then the options [14:56] and merge in updates [14:57] yup [14:57] well not sure mine is prom yet, havent checked today [14:57] not even sure if its been reviewd [14:57] * imbrandon looks [14:57] jamespage: yeah, he's missing some other stuff though :) [14:58] imbrandon: it hasn't I'm reviewing it now [14:58] lynxman, I'll leave it in your capable hands then! [14:58] oh nice ok === al-maisan is now known as almaisan-away [14:58] * imbrandon goes to grab a mt dew then [15:22] mmm caffeine :) [15:23] marcoceppi: were we ( read: you ) gonna try and make the aws switch this evening still ? if so i'll make sure i'm around [15:26] I'm not sure, lets shake over to the other room [15:39] * hazmat reads haxor news comments [16:27] imbrandon: I *love* that idea ... the only thing I don't like about it is that I didn't come up with it. [16:30] heh [16:31] imbrandon: (re a primary "jeos" charm that everything else is a subordinate of) [16:31] like, seriously, its the most brilliantly obvious solution to density that we all missed [16:32] * SpamapS reserves the first gold star on his fresh sheet of gold/silver/green/red star stickers for imbrandon [16:32] :) [16:32] * imbrandon is happy [16:41] lol === shazzner-asleep is now known as shazzner [16:53] SpamapS, this reminds me when i was at pycon about 5 or 6 years ago... and there was a teacher conference overlapping with us. one of the talks: using star stickers to motivate students [16:54] jimbaker: it always worked for me :) [16:54] :) [17:02] marcoceppi: ping [17:02] shazzner: pong [17:03] marcoceppi: hello :) [17:03] hey Marco, can you try running that parse_upstream php file? [17:03] shazzner: Yeah, it worked for me [17:03] huh [17:04] if I remove the https to just http it works swimingly [17:04] I have no idea, is there an import cert or something I have to call :/ [17:04] Shouldn't be [17:04] hmm [17:05] what's odd, it seems to be only sourceforge, google https works fine [17:06] I'm probably going to leave it in for now, and just see if it's some weird configuration on my computer [17:07] <_mup_> juju/relation-ids-command r510 committed by jim.baker@canonical.com [17:07] <_mup_> Merged trunk [17:07] shazzner: http://paste.ubuntu.com/914797/ [17:08] You'll want to make sure you update the project id at least :) [17:10] marcoceppi: I did :) [17:11] shazzner: for testing, if you're only doing local, you could just drop https for http [17:11] Though, it'll need to be https for the charm school [17:11] like this doesn't work for me: http://paste.ubuntu.com/914802/ [17:12] pretend there is a so weird. what does the php info for your setup look like? [17:13] good question, one sec [17:13] I know openssl is running [17:15] man I forget how big these things are: http://paste.ubuntu.com/914809/ [17:16] file_get_contents is off for remote calls by default [17:16] marcoceppi: [17:16] need to use curl [17:17] shazzner: try installing php-curl [17:17] k [17:17] one sec [17:17] Though, that shouldn't affect it, since straight file_get_contents works [17:17] http://paste.ubuntu.com/914812/ [17:18] i dunno some reason i cant see him talk, i'm only getting half the convo [17:18] lol [17:18] but there is a function i use in my classes to "replace" file_get_contents [17:19] right, that's fine, but it's working on my machine for https, not on his for https, both machines work fine for http [17:19] hrm [17:19] yeah [17:19] php-curl, no help :/ [17:21] according to the phpinfo I'm on openssl 1.0.1 [17:21] only thing i can think of is not having the ca's or somthing and it rejecting it, but the log should say that [17:21] maybe see if the curl works anyhow [17:21] shazzner: weird, in the phpinfo for you check if you have this: Registered PHP Streams => https, ftps, compress.zlib, compress.bzip2, php, file, glob, data, http, ftp, phar, zip [17:22] this is all it returns: http://paste.ubuntu.com/914821/ [17:22] and it bugs me i cant see him talking, wtf, jcastro noticed that with the subway client anywhere else ? [17:23] I can see you talking imbrandon [17:23] shazzner: check your phpinfo, it looks a bit was left out. There's a part at the top of the info that dumps out PHP Sterams and Registered Stream Socket Transports [17:24] marcoceppi: http://paste.ubuntu.com/914825/ [17:24] brb gonna get on irssi [17:24] marcoceppi: you're right I forgot to copy it :p [17:26] http://paste.ubuntu.com/914828/ [17:26] imbrandon: can you hear me now? :) [17:26] yea [17:26] yay :) [17:26] finally [17:26] no idea why now before [17:27] i was on that new subway client though [17:27] not* [17:27] huh [17:27] shazzner: So, the final thing I see is this: http://stackoverflow.com/questions/5498497/error-with-fsockopen-and-ssl-failed-to-enable-crypto [17:27] similar error, but it's due to firewall [17:28] huh [17:29] and this: http://gcov.php.net/viewer.php?version=PHP_5_3&func=tests&file=ext%2Fopenssl%2Ftests%2Fbug54992.phpt === bac`` is now known as bac [17:42] marcoceppi: heya [17:42] marcoceppi: Do you happen to know if the wordpress+mysql pair is working in the charm store? [17:42] niemeyer: It is last we deployed, about 3 weeks ago [17:43] marcoceppi: Awesome, thank hou [17:43] you [17:43] Wordpress + HAProxy don't work though [17:43] in fact, I'm willing to wager a few of the applications won't work with HAProxy [17:45] marcoceppi: it works, but not the way we want it to [17:46] marcoceppi: everything I've promulgated works w/ haproxy [17:46] marcoceppi: and wordpress did, at one time, work, but I think something got changed that broke the way it handled the Host: header [17:46] SpamapS: it didn't work when we deployed it, the WordPress charm is setup to listen to hostnames, and only the hostname of the unit. Since originating requests from haproxy are under a different hostname the wordpress box needs to be tweaked [17:47] SpamapS: right, should be fine when we start backporting the nginx configuration [17:47] marcoceppi: there's a default too though, the default should have worked [17:47] yea i'm sooo makin a patch for that [17:47] I'm just excited to see the charms deployable without needing all of openstack or costs for ec2 [17:47] wops wrong window [17:53] haproxy should pass the hostname though [17:53] like we have nginx doing [17:53] err we do now [17:54] imbrandon: there's some logic in the haproxy charm to try and determine whether or not to turn on passing the hostname. [17:54] but i'm still gonna work on a patch for wordpress , thats ludacris the way it does it, every other cms and framework in the world is smarter [17:54] imbrandon: that still would not have mattered in the omg case because the backend services were expecting the external hostname. [17:54] ahh [17:54] imbrandon: It's not a wordpress thing, it's WordPress multi-site from the archives [17:55] marcoceppi: well wp does it too [17:55] with the login [17:55] that's a different issue though, it's more a security thing [17:55] You can only post private data to the site_domain in the configuration [17:55] nah, its misguaded secuirity [17:55] misguided, sure, but it's why it exists [17:55] like how drupal handeles it [17:55] yea [17:56] but yea the normal one doe the host thing too, thast why i had yoy add [17:56] that filter to the settings [17:57] it removes that host header filter === almaisan-away is now known as al-maisan [18:03] I think we need to add an 'endpoint-host' component to the http interface so that chains of servers can configure appropriately for the intended endpoint. === al-maisan is now known as almaisan-away [18:04] like if you have nginx load balancer -> mod_security -> appservers .. need to know that the expected Host: is "mysite.com" [18:04] * SpamapS rings the bell twice [18:04] OpenERP promulgated! [18:04] is there a way to make it an array ? i'm thinking about when i setup sites [18:04] without juju i have 2 endpoints [18:04] as i seperate the admin side [18:04] most of the time [18:05] imbrandon: sure, comma separated would work fine.. just have to think about that. [18:05] hrm [18:05] kk [18:07] cuz its common for me to setup a drupal site but have domain.com/**/add domain.com/**/edit and domain.com/admin 403 via .htaccess and then another admin.domain.com vhost on the same box that restricts it to an ip subnet [18:07] <_mup_> txzookeeper/managed-watch-and-ephemeral r56 committed by kapil.foss@gmail.com [18:07] <_mup_> address review comments [18:09] frak [18:10] marcoceppi: i was looking over your comments and rember how i said i changed install quite a bit and you was like "huh?" for omg ? looking at your review and the file, i think i switched the install hooks for drupal and omg [18:10] hazmat: hey, is there a way we can get access to have 'charm promulgate' make _mup_ talk for us? [18:10] i bet if you look in the omg install hook it has git-core [18:11] on the apt-get line [18:11] hazmat: would be nice to have the bot do it :) === shazzner is now known as shazzner-away [18:11] imbrandon: it doesn't [18:11] i was working on both charms at the same time so .... [18:11] ok [18:11] good [18:11] the omg install hook is still all omg specific [18:12] and it works, because we re-deployed it recently [18:12] ok good, then i just fubared one [18:12] cuz i kjnow i changed all thois [18:12] but no biggie [18:12] there are other things you mentioned i needed to change anyghow [18:12] ok , just wanted to make sure i dident screw the pooch really bad [18:13] yea it would have worked [18:13] but just had extra stuff [18:13] like git-core installed etc [18:13] imbrandon: I review all the diffs for merges very carefully, that would have caught my eye [18:14] rockin :) [18:14] yea see i have the admin.* stuff in the drupal one, frak [18:15] i def screwed something up here , /me checks git history [18:17] btw did you ever add my key ? [18:19] ugh yea and i had all the backup stuff replaced with backup-and-migrate .... [18:19] * imbrandon rolls back git [18:27] * Advisory ID: DRUPAL-SA-CONTRIB-2012-056 [18:28] erm wrong win [18:29] SpamapS: oh yea i posted some more unholly acts of bash on my blog this morning, dunno if you read planet or not [18:29] SpamapS, yeah.. its trivial afaik the bzrcommit hook uses a fairly simple protocol.. but its a client side hook [18:30] hazmat: thats fine, I want to bake it into charm-tools so that when somebody successfully runs 'charm promulgate' the bot informs #juju :) [18:30] which needs the creds distributed with it.. its basically just a message relay/send to mup [18:30] ah, so the bot doesn't support openid? ;) [19:12] SpamapS: heh i wasent the only one who said it "like osx brew on crack" comment on jcastro's latest blogpost :) [19:17] marcoceppi: ok so i pushed the fixes for your review, do i put the bug back to needs-review or ? *me is excited to get in the store* [19:18] imbrandon: put the bug as fix committed [19:18] k [19:24] jcastro: HN! congrats! [19:25] thanks! [19:25] just catching up and saw that... awsesome [19:26] you made HackerNews ? [19:26] :) [19:26] nice [19:27] ... and clint just promulgated openerp. [19:27] another fine patrick hetu charm. o/ [19:28] m_3_: SpamapS: marcoceppi you guys need to pick a winner [19:28] SpamapS: realizticly how long before subs hit ? before uds ? [19:28] jcastro: so many good ones [19:29] hazmat: does the doc generator sphinx thing only do RST? Or will it generate Markdown as well? [19:29] jcastro, rst only [19:29] generate from markdown I meant [19:29] dang. [19:30] jcastro, i'm mean its extensible.. and it can be done.. but its not the default, and likely requires an RT to do. http://stackoverflow.com/questions/2471804/using-sphinx-with-markdown-instead-of-rst [19:30] in this case it looks like it needs some dev [19:30] in other words. "deal with it Jorge" :) [19:30] * jcastro goes to fix the docs he broke [19:33] imbrandon: instead of rejecting it again, can you please add git-core to the install hook [19:33] imbrandon: instead of reading what I wrote, ignore me [19:34] heh [19:34] i was gonna say i did [19:34] when I did a pull it conflicted, didn't read the output carefully enough [19:34] if there is anything else minor like that though lemme know :) [19:34] np [19:36] and ty for re-looking at it so fast :) [19:37] sabdfl always has such well thought out blogposts, but i guess when you have as many eyes on you as he does you kinda have to [19:38] it ended up being something like 6k pageviews. [19:38] 'nice [19:38] so still small potatoes, but at least people know what it is now [19:42] imbrandon: so what other hardcore stuff do you have other than drupal? [19:47] jcastro: hrm nginx drupal advanced js and css server side optimzations [19:48] build and deploy schemes [19:48] hrm [19:48] i really like build and deploy schemes [19:48] hmm so for all the nginx stuff we're likely waiting on subordinates right? [19:48] yup [19:48] ok [19:48] i was actauly thinking about just a website build server though [19:48] not just jenkins [19:48] etc [19:49] one that runs thfough like the h5pb ant scripts and [19:49] optipng [19:49] and less [19:49] scss, closure compilers [19:49] etc [19:49] imbrandon: your upgrade-charm hook is broken [19:49] doh. me looks [19:50] it fails in two places, if an upgrade charm is run prior to a db-relation being joined it fails, and even after a db-relation is joined if fails because it's looking for files/drupal instead of just drupal [19:50] It needs to make sure /var/www/drupal exists before it rsyncs, and also correct the source path [19:50] hrm kk, qwick fix [19:50] yea [19:50] hrm [19:50] then git will fail [19:51] wtf [19:51] Just do if [ -d "/var/www/drupal" ]; then [19:51] in the upgrade charm hook [19:51] k [19:51] yea then next round [19:51] it will come back if ... [19:51] ok ok [19:53] pushed [19:53] err git pushed [19:53] one sec [19:53] k pushed [19:54] well fuck , dient change the path [19:54] i mean frak [19:54] wait a minute, files drupal is right isnt it [19:54] not just drupal [19:55] oh no its not [19:55] i have it on the outside [19:56] take your time [19:56] i know, was just anoying [19:57] ok i moved drupal into files/ that was the intent anyhow and its refrenced other places as in files/drupal [19:57] and pushed [20:02] oh nice , did not know a storm was comming, heard thunder and look outside and it s pitch black out [20:02] WINTER IS COMMING [20:08] hazmat: it looks like the code generator thing IS set up for the docs is smart. I've been fixing the docs and then it just regens them, so I get the feeling it's more than just a daily cron job. [20:09] imbrandon: subs are nearly here I think [20:09] SpamapS: sweet [20:10] jcastro: maybe its */15 instead of daily [20:10] yeah so anyway, docs being generated quickly, so none of us have excuses. :) [20:10] jcastro: to better awnser what you was fishing for earlier , any part of the xAMP stack [20:12] * imbrandon wants to see cgi perl wwwboards come back into fashion [20:14] jcastro: i was thinking about a mysql ga 7.2 ndb cluster next , with those speedy joins [20:14] dunno tho [20:19] hrm MAAS isnt new, i was doing that 8 years ago when i worked in a datacenter, infact we had dedicated ports on each switch in the racks that would be on the pxe boot + reimage server + add configs badd ass network, i thought it was going to be more than that from Marks blog post [20:19] jcastro, cool might be hrly [20:19] * imbrandon feels jipd [20:20] imbrandon, it basically wants to be that with an api.. but its also the foundation for bare metal deploys of juju [20:20] seriously did i miss something major there [20:21] imbrandon: agree, the concept isn't new [20:21] ahh k [20:21] but it's more than just pxe boot [20:21] well yea base kickstart config [20:21] or ubuntu preseed [20:22] the api's will be intresting [20:22] i dint see that part [20:22] but at fisrt glance i was like ummm [20:22] its also pretty early for maas (really early actually), it will be interesting to see what it can evolve to [20:23] definatly , esp since it will now have alot of attn [20:23] that kinda thing has always been datacenter wizrdry [20:23] imbrandon: so you know the charms we've been writing. Doing a deploy would just have MAAS firing up bare metal boxes instead of EC2. [20:23] or in addition to, or both. [20:24] right [20:25] yea , but thats a natural evolution , i mean i'm even kinda doing that now on my esx box heere at home minus a few shell scripts, the reall coolness will be to orcrstrate that via a charm [20:25] with the api [20:25] yeah but you know how to do that [20:25] and bought ESX [20:25] esx is free :) [20:25] we're talking about ootb part of the OS. [20:25] yea true [20:26] esx would make a fat charm on bare metal too [20:26] since its free cuz it has not gui [20:26] :_ [20:26] :) [20:27] vsphere is what you pay for though jcastro , the gui part with all the goodies, the under pinning esx is free if you know what your doing [20:27] like xenserver from citrix [20:27] xen is free but not the othert citrix goodies [20:28] and yea i forget sometimes that even tech ppl are as tech as us [20:28] for the most part [20:28] imbrandon: let me put it this way as an example. [20:29] if I was at my old job and someone would have said "quick, we want to set up 250 blogs, one for each faculty member." I would have been doomed. [20:29] with all this it's a juju for loop [20:29] arent* [20:29] heh true, or one s/wordpress/drupal :) [20:29] heck, this is so easy I am wondering if I should get into the wordpress hosting business. :) [20:30] but thats nother story :) [20:30] hehe [20:30] ahhaah [20:30] you dont want to , wp upstream is a bitch [20:30] start a saas host for drupal like getpantheon.com [20:30] :) [20:31] hmmm so, with this load balancing nginx thing + subordinates [20:31] that's going to be pretty awesome [20:31] * SpamapS has been happy with drupalgardens which drizzle.org switche dto recently [20:32] my goal after uds is to convince getpanthon to use juju instead of their home grown drop scripts, they already are on rackspace, with like 350 instances or so [20:32] thats 350 charms [20:32] :) [20:32] jcastro: yes, its going to be sexy.. every web app server will be a able to be a load balancer unto itself. :) [20:32] SpamapS: should there be a drizzle charm? [20:32] marcoceppi: yes I've always wanted to write one [20:32] SpamapS: yea smae exact think except david from pressflow runs pantheon and dries / acquia runs drupal gardens [20:33] SpamapS: after seeing your g+ post about the next release, I though it would a be a sweet charm to have [20:33] exact same setup though, infact they share a codebase [20:33] marcoceppi: totally, so many ports to expose/consume [20:33] pantheon / drupal gardens [20:33] marcoceppi: the HTTP+JSON thing is especially interesting I think [20:33] the main diffrence between them is gardens is on AWS and Pantheon is on Rackspace [20:33] thats about it [20:34] marcoceppi: also the 0mq transaction poub/sub should be interesting. [20:34] jcastro: yea like omg does now ;) [20:35] jcastro: and the drupal charm :) [20:35] mmmm node charm [20:36] http + json ? /me perks up, whada i miss , reads backscrolll [20:37] SpamapS: it be interesting to see the charm dynamically open/close ports as relations to specific things are made [20:37] From my initial investigation there's a lot to be done to make the charm right by the Drizzle devs [20:38] ohhhhhh pantheon isnt invite only now, gogo, get an account jcastro SpamapS marcoceppi , that is the most perfect workflow for small teams using drupal [20:38] * imbrandon ran outa invites the other day to send [20:38] imbrandon: Drizzle has a plugin now that listens for HTTP requests and responds w/ JSON of rows read. [20:39] nice [20:39] wow [20:39] really nice [20:39] * imbrandon wants [20:39] man this suck tho [20:40] mysql proper added memcache api , the other dudes added socket handler [20:40] and now this [20:40] wtf [20:40] if they would all 3 combine and omg [20:40] imbrandon: I believe the memcache thing is in drizzle too [20:40] sweet [20:40] thats the two i would use, not so much socket handler [20:41] imbrandon: drizzle's model of having an actual, working, efficient plugin architecture is really the main reason to use it. [20:41] imbrandon: I maintain handlersocket in Debian. Its not the most awesome thing. ;) [20:41] has to be built inside the mysql source tree [20:41] i wonder if they have the upstream enhacements from 7.2 for the joins [20:41] imbrandon: thats NBD only [20:41] doh [20:42] imbrandon: 5.5 has some join optimizer stuff.. but I don't know if Drizzle has those. [20:42] thats drupals bigest downfall, have tyou seen some of the crazy joins it does [20:42] Drizzle was forked from "MySQL 6.0", so it might have had it already. (6.0 became 5.5, minus some things like Falcon) [20:42] ahh [20:43] i breefly looked at it , kinda like mariadb then brished it off [20:43] percona was all i really looked at lately [20:43] i should probably look em all over again [20:44] imbrandon: Percona is MySQL + stuff. MariaDB is MySQL + way more stuff. Drizzle is a true fork. [20:44] oh and you have to deal with handeler socket / mysql et al packing, my hats off to you, its like the apache guyes [20:44] heh [20:44] Drizzle was the result of the frustrations of quite a few people with long standing crapiness in MySQL. [20:44] yea with the orig maintainer etc [20:45] It was "we're not going to make money with this for many years to come, so lets rip out all the old broken code" [20:45] No, Montye Widenius was not involved with Drizzle [20:45] percona is mostly xtradb engine, maria is stolen xtradb + configs [20:45] he started MariaDB [20:45] :) [20:45] ahh i had him backwards then [20:46] thought he was with drizzel not maria [20:46] dident that curt dude from canonical go work for maria [20:46] the canook [20:46] heh [20:47] curt vonfink or something like that, he worked as helpdesk i think, cant ever rember but he was funny as hell [20:48] wow, i'm so installing this now [20:48] i was actually looking at nodejs servers that did just that [20:48] grabbed rows and spit json out [20:48] saves a nodejs server running [20:49] imbrandon: yeah he did [20:49] ok, afk for a bit [20:49] k [20:52] i want to recreate the pantheon service in a charm ( probably with lots of subs ) [20:52] seriously like a clone of it, i bet it can be done as they use the same thing in house written scripts on rackspace instnces [20:53] only free and in the store [21:05] imbrandon: so now that you've sort of talked me into juju, I should look at how I can use it for work :) [21:06] * ajmitch can at least set up things with fabric at the moment [21:06] yea [21:06] dude fab is nince [21:07] juju makes fab projects talk to each other though [21:07] ajmitch: i thouth you did php at work, and u use fab ? [21:08] imbrandon: doesn't mean I like PHP :) [21:08] jcastro: btw look who i dragged into the jujubee jamboree wont be long and we'll have all the oldskoolers :) [21:08] ajmitch: hahaha true [21:08] imbrandon: I've been lurking in here for some time [21:09] i knw i knw i seen ya the first day i came, but i got a bit ego :) [21:09] big* [21:09] seriously though ajmitch i just noticed pantheon opened up drops to everyone [21:10] should grab one if nothing else to check out the workflow [21:10] the charm store is interesting, in a way it's an admission that packaging a lot of the bits that people want to use is a bit of a failure :) [21:11] kinda, but some will still want the stability of pkgs, its like having cake left when ya eat it [21:11] ajmitch: not so much that the packaging is a failure, more that it takes more than just a package to configure a full service [21:11] heh you get the stability of a base system and the package you care about are a rolling release [21:12] marcoceppi: when it comes to fast-moving things like node.js it's hard to keep up on the distro side [21:12] i think of it like selective rolling releases kinda [21:12] imbrandon: right, which is a useful model to have [21:12] ajmitch: oh absolutely, in that case the nodejs charm makes use of the ppa or the upstream repo [21:13] as long as it is selective too and not everyting is installed that way [21:13] though in saying that, I see that someone uploaded node.js 0.6 to precise a couple of days ago, I'd thought it was far too late for a change of that size :) [21:13] i mean its mnore to it, but you know, the package side thats how i think of it [21:15] ajmitch: do you really use fab at work though and not like phing ? [21:15] I really use fab at work, honest [21:16] nice, i could have never talked my last employer into that [21:16] infact my whole job was to rid the palce of alll python starting with django apps [21:16] heh [21:16] how does juju determine which amazon AMI to use if no default-image-id has been specified in environments.yaml? [21:17] it likely defaults to aws default then of a m1small [21:17] i would guess [21:17] with a amazon distro [21:18] imbrandon: juju does some magic to find an appropriate ubuntu cloud image to launch given the ec2 region [21:19] ahh === shazzner-away is now known as shazzner [22:25] adam_g: https://cloud-images.ubuntu.com/query [22:30] * SpamapS is so far behind on email, don't know if hazmat +1'd the latest iteration on the cert checking code [22:34] marcoceppi: hey Marco, I tried the same file_get_contents on my laptop and it worked fine [22:34] shazzner: must be something on your local then [22:34] it's running 11.10, I'm on precise beta [22:34] oh, fascinating. Let me try that on my precise laptop [22:34] yeah could be a bug [22:35] I disabled the ufw, which didn't help :p [22:36] shazzner: looks like it, file_get_contents to any https address from my laptop fail [22:36] ah ha! [22:36] finally, I'm not the cause of a bug :p :p === samkottle is now known as samkottler [22:38] shazzner: actually, it works for some https addresses and not others [22:40] yeah I noticed that, works with google and some others [22:40] weird [22:40] it must be a cert issue [22:40] huh [22:40] I would put that as a big "todo" for when the charms move to precise. A better method of rss parsing will need to be put in place [22:41] maybe using pything instead of php [22:42] using urllib? [22:42] sure, or whatever [22:42] * marcoceppi not much of a python guy [22:42] * shazzner me either! [22:43] also php and bash :p [22:43] yes! take that SpamapS [22:44] m_3 on juju: http://www.youtube.com/watch?v=0iFVvOPcm8g&html5=True [22:44] we gotta get that dude a razor [22:44] haha, yeah. It's gotten unruly over the last 6 months [22:44] marcoceppi: python is the one true way, embrace it while you can [22:45] ajmitch: slowly I've been seeing the light with python [22:48] yep, python works like a charm [22:48] w00t [22:50] I'm going to _try_ to rewrite that parse_upstream in python [22:51] feel free! if you get it to work let me know and I'll port it in to the phpmyadmin charm [22:56] wow that php2python site is a godsend [22:57] I'm down with php.. :) [22:57] yeah you know me [22:57] who's down with PHP? [22:58] openssl broke for php [22:58] jcastro: razors are for sissies [22:58] not sure why or how, but python seems to work [22:59] I triaged a pretty nasty bug where servers using some implementation would fool php into thinking they were SSL3 but they weren't so fail would spew forth in full glory [22:59] http://www.youtube.com/watch?v=3QGDQ2RTLMU [22:59] jcastro: the devopsdays ones are recorded too, but you have to sorta search through quite a bit to get to the panel [22:59] hah, before and after [22:59] speaking of which, I should probably file a bug for that issue [22:59] even though it's probably a fault with sourceforge [23:00] * shazzner hates sourceforge [23:00] https://bugs.php.net/bug.php?id=52106 [23:00] jcastro: what can I say... I'm a looker [23:01] shazzner: perhaps that one? [23:01] SpammapS: yep that's the one [23:01] huh