[00:02] SpamapS: Yeah maybe I can do a Juju talk at PuppetConf this year :) [00:02] bkerensa: actually we're looking at doing just that [00:02] SpamapS: They had one last year [00:03] Marc Cluet and Adam Gandelman and some other Canonical guys I dont know [00:04] heya guys just popin in a sec then be back in after bit, but SpamapS , bkerensa knows a few people on the inside of CloudFlare and got the ceo dude to give us a lil tour ( and personally i wanna talk nginx magic with em as its their backbone to 4 billion views a month ) anyhow i thought you may want to try and sneek away at lunch too and go [00:04] SpamapS: bkerensa ^^ [00:04] yea billion not mill [00:05] with a b [00:06] SpamapS: http://imgur.com/9GFjR <-- I know Adam G and Marc C but no idea who the rest are [00:06] any how i;ll let you to talk it over more i got to run for a bit , but nothing set in stone yet cept we're definatly invites [00:06] invited* [00:07] back in a while , btw heya bk and anyone else lurkin [00:08] bkerensa: some of our IS guys, I don't know most of them well either ;) [00:09] SpamapS: Btw... What does Paul Oh do? he was walking around OSCON like a big shot and came to our booth and gave me his card and said if we needed anything to e-mail him ;p [00:09] he did have a pretty cool Canonical shirt [00:19] bkerensa: business something ;) [00:20] bkerensa: at 500 people.. its become impossible to know everybody :P [00:20] No wonder his shoes were so well shined [00:20] SpamapS: there needs to be a Public Employee Directory :D [00:20] * bkerensa cannot even figure out exactly how many live in Portland [00:21] bkerensa: thats by design. Canonical is the most serious about employee privacy that I've ever seen a company.. which is funny because we all work so publicly :) [00:22] SpamapS: Yeah... I usually have to ping someone I know and then have them e-mail our local canonical mailing list for events [00:39] Hi guys, I have a problem that appears to be synchronization between the postgresql db-relation-changed hook and the service that is using the db interface [00:40] mars: sync in what way? [00:40] When I use debug-hooks in my custom charm and run the same commands as db-relation-changed does, then it connects to the database fine [00:40] But when running the db-relation-changed hook script from 'juju add-relation', then the command bombs out with a database connection error [00:42] mars: the environment might be slightly different [00:42] ah [00:42] mars: but it should be mostly the same... [00:42] that does not sound like a fun thing to debug [00:42] mars: is it possible you are not checking to make sure the values are already set in the changed hook [00:43] mars: relation-changed is run once no matter what, at the start, sometimes that is before the other side has set the values [00:43] SpamapS, the values passed in by the database relation? [00:43] mars: basically, you need to assume that things returned by 'relation-get' are actually valid, as they may not have ever been set. [00:44] mars: you can gracefully exit a hook if the values you require are not set yet... your hook will get called again when they are set. [00:45] SpamapS, ok, I have these lines in my hook: user=`relation-get user` [00:45] [ -z "$user" ] && exit 0 [00:45] Straight out of the charm tutorial [00:46] Does that check for the user, gracefully exiting if it isn't set? [00:50] mars: yeah [00:50] mars: ok so in that case, perhaps the pgsql charm is broken in some way [00:50] mars: because you should be able to assume the user/pass/dbname you were given is open for connecting [00:51] SpamapS, could be, I already had to patch it once. The DB wasn't starting because of a bogus pg_hba.conf entry. [00:53] mars: did we get that fix into the main charm yet? [00:54] SpamapS, got to run, here's what I had to fix: http://pastebin.ubuntu.com/948657/ [00:54] later! [00:55] btw, that's the precise charm === flaviamissi_ is now known as flaviamissi [00:56] mars: thanks! === Furao__ is now known as Furao [03:30] Wow, I knew they did this with Xcode 4.2+ but never knew why, SpamapS still around ? ready this ( fairly short blog post ) and tell me Apple cant be cool in their own way, yea its not Ubuntu open-ness but its beats the hell outa any other propritary vendor including Novel/Mono , yea I'm looking at you charging 200$ for the mobile iOS api , meh [03:31] http://kennethreitz.com/xcode-gcc-and-homebrew.html [03:31] anyhow that just put a very big smile on my face [03:56] imbrandon: ping [03:56] :D [03:56] heya [03:57] imbrandon: so what do you use for irc? [03:57] irssi mostly, just cuz i'm too lazy too properly evaluate subway :) [03:57] * bkerensa uses znc but their push notification module stopped working and I need push notification and something that sits on a vps or instance since I dont leave me laptop on 24/7 [03:57] imbrandon: yeah well I dont know how to deploy my charm to EC2 [03:57] :P [03:58] only lxc [03:58] irssi plus some php perl and bash glue for all kinda stuf :) [03:58] hahah [03:58] but then again I like gui client or cli not browser based [03:58] ok one sec [03:58] ^ [03:58] nvm [03:58] maybe weechat [03:58] subway is nice because its persistant [03:59] thats really all i care about cuz i may actually close my irssi term like 50x a day [03:59] or get to it from 5 diff computers [03:59] thats it, feautres i can always hack something in if i need it in another [03:59] :) [04:00] and apparently according to jcastro i can do that with subwayfine [04:00] :) [04:05] imbrandon: ok run me through it how do I deploy charms on EC2 instead of lxc? [04:07] ahh I got it [04:10] i was gonna say its the same [04:10] just a diff env.y === bulatshakirzyano is now known as avalanche123|h === bkerensa is now known as Tester === Tester is now known as TesterBlah123 [04:51] getting a Error: Permission denied - /tmp/homebrew-juju-0.5.531-PkyV/juju_0.5+bzr531.orig/build/bdist.macosx-10.7-intel when installing juju on OSX using homebrew [04:51] avalanche123|h: ahh you had it installed prior [04:51] imbrandon had juju installed? [04:52] this is the first time I'm trying to do this [04:52] i need to fix that but yea that tar [04:52] hrm ok [04:52] that last bit is easy [04:52] imbrandon anything I can do to fix it? [04:52] um for not the best thing to do ( this si how i got to updatye the forumla to do correctly ) === TesterBlah123 is now known as bkerensa [04:52] is download that same tar from LP [04:53] open it and run [04:53] well open it cd to the dir and run [04:53] sudo python setup.py install [04:53] and you'll be good [04:53] one sec let me get you the tar url [04:54] thank you! [04:54] np, it did the hard part for you already [04:54] getting teh deps intalled [04:54] thats was the last part [04:54] gotcha [04:55] * imbrandon maintains that brew formula btw, got lucky in the timing :) [04:55] indeed, fastest response I've received on irc :) [04:56] 14 seconds :) [04:56] https://launchpad.net/ubuntu/quantal/+source/juju/0.5+bzr531-0ubuntu1/+files/juju_0.5+bzr531.orig.tar.gz [04:56] hehe yea me and bk was just talkin [04:56] nice, thanks again [04:56] np [04:56] if that still dont work let em know [04:56] i;ll be here all morn [04:56] kk [04:59] imbrandon got it installed, thanks! [04:59] np [04:59] now when its time to update [04:59] you MAY have to manualy uninstall [05:00] because brew wont know its already there [05:00] but just do unstalll instead of install [05:00] or pop back in here [05:00] but there proaably wont be an update for a week or so i;m guessing ( dunno tho ) [05:01] imbrandon cool, sounds good [05:01] np [05:02] * imbrandon goes to fix the formula [05:04] imbrandon ping [05:04] yup [05:04] getting this now: [05:04] import zookeeper [05:04] ImportError: No module named zookeeper [05:04] when running juju bootstrap [05:04] brew install zookeeper [05:04] brew install bzr too [05:04] gotcha [05:04] if you dont have bzr [05:04] hrm it should have dine that for you tho [05:04] mind if I add it to formula dependencies? [05:04] lemme check [05:04] no,please do [05:05] brew list | grep zookeeper [05:05] zookeeper [05:05] I have it [05:05] any help is welcomed, i'm a ubuntu dude and run osx too, but a brew newb [05:05] heh [05:05] it is indeed already listed as a dep [05:05] ok hrm hal sex [05:05] yea i thought so [05:05] maybe python bindings are not included? [05:06] do you have brew python or just sys python [05:06] sys python [05:06] well the forum should install those too [05:06] iirc [05:06] zkpython [05:06] or something [05:06] txzookeeper==0.9.5 [05:06] got this in pip freeze [05:07] k one moment [05:07] k [05:08] system "sudo","/usr/bin/easy_install","txzookeeper","PyYAML", "txaws", "pydot","oauth" [05:08] yea no zkpython , let me see if its actually needed or somethign else [05:08] * imbrandon looks on lp [05:08] sorry it does work local and on a few others but i guess i'm missing a dep somewhere where [05:08] * imbrandon looks [05:09] i wish there was a pbuilder for brew [05:09] lol [05:09] heh [05:09] I'll try diggin [05:09] might end up with a pull request [05:09] sure thing [05:10] like i said i'll be here as well [05:10] i'm gussing thats what it is tho is the pythong zk glue isnt installed and i had it already or soemthign [05:10] if LP would ever load it should hvae the deps listed for my on the official linux deb page [05:11] btw undocumented but brew install --HEAD juju should pull right from bzr [05:12] tooo [05:12] fyi [05:12] for update [05:15] Build-Depends: debhelper (>= 7.0.50~), python, zookeeper, python-twisted, python-txzookeeper (>= 0.9.5~), python-txaws, python-yaml, python-pydot, python-apt, python-oauth, help2man [05:15] ok no need for help [05:15] oauth loisted dot listed [05:15] no apt needed [05:15] yaml good [05:15] yxaws good [05:16] txzookeeper versiion [05:16] Frak [05:16] oh wait you got 0.9.5 [05:16] k [05:16] twisted is preinstalled in 10.5 + [05:16] zk is there [05:17] hum [05:17] avalanche123|h: can you try to launch it in a fresh shell [05:17] sec [05:17] k [05:17] same thin [05:17] kk [05:18] is zkpython in pip ? i cant rember [05:18] * imbrandon looks to see if i have that installed [05:19] looking at brew zookeeper formula, there is option to install with python and c bindings [05:19] ahh [05:19] by running brew install zookeeper --c --python [05:19] lemme try that [05:19] cruft ok [05:19] kk [05:19] wonder how mines workin then [05:19] lol [05:19] run brew info zookeeper [05:20] what does it say? [05:20] bholtsclaw@ares:~$ brew info zookeeper [05:20] zookeeper 3.4.3 [05:20] http://zookeeper.apache.org/ [05:20] /usr/local/Cellar/zookeeper/3.4.3 (191 files, 11M) * [05:20] https://github.com/mxcl/homebrew/commits/master/Library/Formula/zookeeper.rb [05:20] bholtsclaw@ares:~$ [05:21] i bet your correct i just dont know how i got around it [05:21] but this install is been upgraded manny times from many diff macs [05:21] lol [05:21] i think it started life ina g3 ibook [05:22] i need to setup ci tests on commit , just havent took the time with release "comming" the last few days [05:22] haha, gotcha [05:22] I still can't get python module [05:23] ok so [05:23] but I'm gonna try completely re-installing zookeeper now [05:23] maybe brew is messed up [05:23] this is hokey but now that it has them easy install the others again [05:23] sure thing [05:23] brb one sec, got hit the little boys room quickly [05:24] afk [05:26] btw way i planed on adding charm-tools and charmrunner to the brews too so not just now but anytime pull req are very much welcomed if i goof, still pretty new to brewin, old hat on the other side of the fence, funny i use osx for photoshop and Espresso.app traditionally so the last month with the role reverse its kinda funny ( been doing lots of css stuff in linux for a diff proj ) [05:27] and prog etc etc on my ubuntu boxen :) [05:28] haha [05:28] totally [05:28] I'm just starting with juju [05:28] i dont really like that i cant pre-build it in a "clean" install tho like a chroot or pbuilder [05:28] yeah, no real packages on osx [05:28] yea i've been packaging debs for 6 or more yeas, not much on the osx side till lately [05:29] well fink is almost [05:29] fink actually uses .debs and dpkg and apt [05:29] but fink is a PAIN [05:29] heh [05:29] who knew label printing was easy in Ubuntu [05:30] but its about the closest, macports being like bsdports [05:30] heya bkerensa [05:30] but yea packing for fink is actually simple if your used to deb packages, essentiually the same [05:31] but installing fink for a "end user" sucks ass even for me [05:31] and i've done it 100000000 times [05:31] :D [05:31] thats why i think brew took off so well, its really a "lite" juju [05:32] * bkerensa is burning 250 12.04 cd's and and and.... wait for it... Testing each and every CD [05:32] :D [05:32] where it can do anythging [05:32] bkerensa: thats what the interns are for, like a blonde brunette and re^W^W^W^W ummmm yea [05:34] avalanche123|h: so yea if you like making brews tho you'll love juju , imagin me being able to install a brew from here onto your mac and then a second and have it on bk's mac and then they can talk to each other and config them self based on that talking [05:34] thats the lowdown [05:34] there is lots more to it but easy way to explain it to a brewer [05:35] yeah, I'm looking for a nice configuration management / services orchestration solution that's why exploring juju [05:36] can it run services in lxc's? [05:36] its still young, so expect bumps but nothing terrible [05:36] right, oss [05:36] yea there is one cactch with lxc tho [05:36] they dont persist reboots [05:36] heh [05:36] better off usein MAAS if you need a persistant local [05:36] more than just testing [05:37] avalanche123|h: yea not just oss tho , juju is moving very fast , its a good thing, i love it but yea [05:37] quick iteration and releases are the norm [05:38] right, totally undestood [05:38] like weekly or so it seems , not sure it will continue like that with 12.04 out [05:38] wow [05:38] but likely for a while untill 1.0 [05:38] that's fast [05:39] yea and the guys ( and gals ? ) are religipous aobut ttd and stuff [05:39] so never much breakage, just once in a while backwards compat stuff is iffy [05:40] that sounds awesome [05:40] * imbrandon is a ubuntu core dev but not a juju dev , i only make charms like your starting to do and happend to package the brew app since i seem to be the token Mac guys arround :) [05:40] just fyi [05:40] haha, got it [05:41] same with bkerensa here, infact he just won the last charm contest [05:41] i'm determined to win this one, well one of the 3 [05:41] ;) [05:41] imbrandon: and if I win this one coming up well hot damn I will be the new juju evangelist everday [05:41] :D [05:42] heh , psudo am anyhow, just not a great evanglist [05:42] lol [05:44] :) [05:46] bkerensa: you ready to join ~charmers yet btw? [05:47] bkerensa: I would normally say if you have two charms promulgated, you're pretty much ready. [05:50] let's wait until UDS and make up some silly ceremony [05:50] involving beer [05:50] (just throwing that out there) [05:52] SpamapS: I need to push openphoto maybe monday or tues for u took peek at [05:53] jcastro: saki bombs for all new ~charmers! [05:53] bkerensa: cool. I would be seriously grateful for another pair of eyes to look at charms as they come in. [05:53] * SpamapS notes that he has *4* submitted for review. [05:53] and ELB as a subordinate is about to drop as well. :) [05:54] saki bombs? [05:54] jcastro: Q: can we byob to UDS? [05:54] :D [05:54] hmm, we need to whine about reviews tomorrow [05:54] * bkerensa noticed photos showing bottles of liquor at previous UDSes [05:54] :d [05:54] like at tables [05:54] :D [05:55] bkerensa: after sessions, if you don't go find beer, beer will usually find you [05:55] and frankly... in Oak town.. you want to do your own beer finding. :) [05:55] bkerensa: you need a strong liver for uds [05:56] you have to porepare all yea [05:56] r [05:56] SpamapS: hahaha [05:57] SpamapS: I lived in Bay Area for a few years [05:57] ;p [05:57] SpamapS: i was thinking about trying to find that damn kereoke bar with mrs two teeth in it and dragging all to it, there was some kiler sushi accross the screet too [05:57] Oak Town is not my fave :P [05:58] but it is on the other side of the bay [05:58] imbrandon: dude in chinatown sf they have asian karoke [05:58] :D [05:58] funny as hell [05:58] :D [05:58] hahah [05:58] there is a story bhind the other and it ended with jono dancing with a bottle of beer [05:58] on camera [05:58] no less [05:58] :) [05:59] SpamapS: http://i.imgur.com/3SdUp.jpg [05:59] ;p [05:59] http://tu-349002179.us-west-2.elb.amazonaws.com/ [05:59] fairly certain jcastro was there too, i know kenethwimher was and oh hell simon and all of us [05:59] ELB controlled thinkup [05:59] (haven't related mysql yet ;) [05:59] what? no spdy? [06:00] jcastro: we doing elb so we have 5x the 37mill numbers from nginx [06:00] :) [06:00] jcastro: you sir, are eeevil ;) [06:00] thirty seven frakin milllion [06:00] it's just one more relation! [06:00] If I win XPS I will donate my Dell 14z to Partimus [06:00] :D [06:01] which is only a few months old itself [06:01] Partmius ? [06:01] partimus.org they setup nix boxes in the bay area for non profits and schools etc [06:01] pleia2 and a few Ubuntu CA folk are on the board [06:01] :D [06:02] ahhh nice, in that case i'll bring the older of my two macbooks and make the same deal, if i win one i'll donate it [06:02] :) [06:02] sounds like a great cause [06:02] and its not a terrible macbook, its the white one before the unibodys [06:02] intell core 2 duo 2gb ram [06:03] not a speed demon but good to ssh and surf [06:03] :D [06:03] jcastro: we should instantly win ^ for our charitable natures [06:03] :P [06:03] lol [06:03] nah [06:03] heh [06:03] I would actually trade a XPS for a old macbook LOL [06:03] :D [06:04] ok time to get cracking on charms [06:04] energy drinks ready [06:04] :D [06:04] well if i DONT win i'll trade ya, honestly that one is my beater i take on things like this incse it dont make the trip [06:04] * bkerensa is going to work till 2am (its 11:04pm now) [06:04] imbrandon: which one is it? [06:04] i mean its not tore up or anythign but you know what i mean, stickers everywhere , 4 gens old [06:04] the brushed metal one [06:04] etc [06:04] a1181 [06:04] or the ugly white ones [06:05] is the model, white [06:05] hmm [06:05] yeah I want a brush metal one :D [06:05] http://www.everymac.com/systems/apple/macbook/specs/macbook-core-2-duo-1.83-13-specs.html [06:05] its that one [06:06] i have a 17inch mbp i use for work/contract and then my mini [06:06] and i use that white one for camping trips and uds [06:06] and stuff like that [06:06] where it might not make it home [06:06] ;) [06:07] but yea aside from being not the uber latest and i'm spiled its honestly a great machine for someone [06:07] spoiled* [06:08] and really if it had a display port and not mini vga [06:08] i would use it more [06:08] but i need screens [06:08] when sitting at a desk, even with a lappy [06:09] mbp has thunderbot == 3 dvi off one port :) [06:09] more if i was not so $$ cheap [06:10] ok SpamapS and jcastro and bkerensa yall are whitness to me saying that , hold me to it, but i'm not gonna bring it up again cuz i dont want anyone thinking i am trying to bribe a win [06:10] :) [06:11] :D [06:11] imbrandon: is it possible to piggy back on another charm? [06:11] like deploy apache ? [06:12] you mean a sub [06:12] sure [06:12] ots got to be designed that way but there isnt much to the "design" [06:12] SpamapS: is the current expert on it but i can fumble threw one atm [06:13] jcastro: whom has control over the charms import into github nightly [06:13] like where is the script that the cron runs [06:13] heh, it has the old url in the descriptions [06:13] http://charms.kapilt.com/charms/oneiric/haproxy [06:13] etc etc [06:14] m_3 runs the github import stuff [06:15] kk [06:15] i'll prod him next time i think about it [06:15] bkerensa: you probably want the new 'subordinate' charm type. [06:15] >.< [06:15] k back to charming for me for a lil while [06:15] screw it I will just write it all [06:15] :P [06:28] hrm.. so this kind of sucks.. if I make ELB a subordinate.. we never get the 'stop' hook called so there's no simple way to know that the other nodes have left. [06:28] Perhaps a peer relation.. hm [06:33] yea a peer-relation so ip can be both added and removed [06:33] i'd like that better than start stop anyhow [06:34] well the issue is that the subordinates run on the same box as the "remote" unit [06:34] so.. its a bit tricky [06:34] right , so check the service type too ? heh [06:34] lol [06:35] too bad subs got to be all at the same level, i can imaging some nice trees that would mimic inheritance [06:35] visually if dran [06:35] drawn [06:35] or cacade maybe is better word [06:35] cascade [06:35] yeah [06:35] I think we'll have to make that a reality at some point [06:35] subs of subs [06:35] yea [06:35] we can call that middle-management [06:35] :) [06:36] imbrandon any idea what ERROR Could not find any .pem files. means? [06:36] I get this when running juju boostrap [06:36] hrm [06:36] its not finding your ssh key [06:36] .pem? [06:36] so its not likely setup int env y file to point to your auth_keys [06:37] SpamapS: yea osx mixes em [06:37] hmm, I must be missing some init steps, where can I read about this? [06:37] I could put it in the formula for the future I guess [06:38] ava please do if its part of the install [06:38] but i would not addd bootstrap stuf in there [06:38] can you run [06:39] whatever juju cmd you was doing | tee pbcopy [06:39] then pastebinti [06:39] err pastbinit [06:40] don't have pastbinit [06:40] also might open a second term tab ( cmd t ) and keep juju debug-log , going for now [06:40] ava thats what pbcopy is [06:40] its built into osx [06:40] gotcha [06:40] it will copy the ourtput to you clipboard [06:41] also might open a second term tab ( cmd t ) and keep juju debug-log , going for now [06:41] and watch as you run things ( it will lag a few behind and seem "stuck" for 30ish sec or more sometimes in the debug log [06:41] just ewait for it [06:42] https://gist.github.com/c607887c23c8b9742cdd [06:42] when I rub debug-log [06:42] now as far as whatever those output , SpamapS can proibably help ya better then [06:42] * imbrandon looks [06:42] *run [06:43] hrm [06:43] oh [06:43] same issue [06:43] yeah [06:43] it cant find your priv ssh key [06:43] eg your env.y is wrong [06:43] likely [06:43] which is just ~/.ssh.id_rsa [06:44] ~/.ssh/id_rsa [06:44] right and is that listed in ~/.ssh/authorized_keys [06:44] ? [06:44] heh, I don't have that set up [06:44] ~/.ssh/authorized_keys doesn't exist [06:45] heh ok make sure its listed there and then in the env,y make sure authorized-keys-path: ~/.ssh/authorized_keys [06:45] then bootstrap the env [06:46] and it will have your key copied and work at that point [06:46] kk [06:47] SpamapS: every time ssh complains on mac about it wanting a .pem file its the priv key, and some priv keys are .pems for vertain programs like cyberduck needs it like that [06:47] the keygennon osx makes both most of the time default [06:47] iirc [06:47] but the error only ever says one [06:47] and its confusing as shit [06:47] weird [06:48] SpamapS: http://paste.ubuntu.com/949112/ [06:49] see i only got one [06:49] and man i need to clean those up [06:49] ugh [06:49] and no idea what those .keystore ones are but likel;y another mac thing [06:50] oh no there is two .pem's in there id_rsa.pem and me@bradnonholtsclaw.com.pem [06:50] I run cat ~/.ssh/id_rsa.pub > ~/.ssh/authorized_keys [06:50] just missed one [06:50] still same issue [06:51] did you re bootstrap [06:51] the env after [06:51] ah [06:51] it needs to ahve the pub key in the bootstrapped env [06:51] so you can ssh in and run juju stuff [06:51] or nadda works [06:51] :) [06:51] would opening a new shell do that? [06:52] and with that i'll be right back [06:52] no [06:52] juju bootstrap [06:52] , it makes the first "control" nde [06:52] that fails because it can;t find key [06:52] mode* [06:52] that one dont need a key [06:52] well it does [06:52] but it needs the auth [06:52] you just setup [06:53] the pubkey [06:53] make sure authorized-keys-path: ~/.ssh/authorized_keys [06:53] is in your env.y before you bootstrap [06:55] ok i got to be afk a few, SpamapS or bkerensa should be arround if you still need more, at this point you should not have any more specific osx problems [06:55] your past all that [06:55] brb [06:55] thanks a lot [06:55] np [06:56] * SpamapS doesn't like helping other dudes w/ their osx-y problems, but if he must.. he will take one for the team. :) [06:56] * bkerensa dances [06:56] SpamapS: promis, as long as bootstrap there works , then its golden, just a matter of knowing what order to run stuff in [06:56] https://code.launchpad.net/~bkerensa/charms/precise/locker/trunk <- jcastro Y CAN HAZ [06:57] * bkerensa goes on to pinry [06:57] cuz i used it on like 10 omg dep[loyds [06:57] deploys* [06:57] and near every command [07:01] http://tu-349002179.us-west-2.elb.amazonaws.com/ ... [07:01] two instances [07:02] 10.252.25.165 - - [27/Apr/2012:07:02:30 +0000] "GET /assets/img/favicon_googleplus.png HTTP/1.1" 304 186 "http://tu-349002179.us-west-2.elb.amazonaws.com/" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_7_3) AppleWebKit/536.9 (KHTML, like Gecko) Chrome/20.0.1115.1 Safari/536.9" [07:02] mac users [07:06] hey teh modern internets was made on that OS [07:06] maybe fore it was called Mac OS X but still :) [07:07] ok, so I'm sieging that url.. both instances are CPU bound [07:07] was still NeXT then on Tim Burns Lee or showever its spelled [07:07] nice [07:07] adding another one [07:08] here ina bit i'll have ya try ti on 2 or 3 of my nginx boxen [07:08] exploding wax [07:08] and see how many thousand we can all toss at it [07:09] heh.. sucks to use the amazon EC2 tools.. because they need java :-P [07:09] I wonder if boto has grown elb yet [07:09] threre is a rest api [07:09] curl it [07:10] I'm eh-scared :) [07:10] heh [07:10] also wouldn't that mean I'd need a CERT, precluding use of IAM ? [07:10] or is that the SOAP? [07:10] dont be scared lil boy there nice shaggy carpet back there in the back .... of the van [07:10] I can never remember [07:11] well not precluding the use of it [07:11] but yea [07:11] you can use the cert or 509 or keypair or pass or [07:11] ummm one more [07:11] oh mfa [07:11] with iam [07:12] the keypair requires all that HMAC magic tho [07:12] and yea soap needs 509, rest uses header with the aws keys [07:12] nah the example code uses wget and the does -D for data [07:12] and plain txt keys [07:12] etc [07:13] or maybe curl and -d but still [07:13] one of the two [07:13] and, hooray, 3rd node joins, boxes drop to load of 0.5 [07:13] takes a while tho [07:13] stickey sessions [07:13] I'm not using those [07:13] oh [07:13] because I think they're *silly* [07:13] hrm [07:14] no I mean from add-unit to unit being ready.. [07:14] they may be silly but saved my asss more than once without any new code [07:14] took about 1 minute for amazon to do its thing [07:14] oh yea [07:14] that always takes a long time to me [07:14] then another 3 to get everything installed [07:14] i thought it was just part of juju [07:15] heh i kick off the scripts from bootstrap to expose as one long chain cmd [07:15] and it comes bacl to promt [07:15] but then its 40ish min [07:15] beofre they are all ready [07:15] on omg [07:15] well thats because there's a 20 minute DB restore [07:15] because mysqldump *SUCKS* [07:15] nah i watched thats like 30 sec [07:16] no, what? [07:16] the db is only 74mb now [07:16] we trimed it [07:16] Oh haha was it like, every user from back when they did comments? [07:16] there was tons of extra crap in there [07:16] yera and 40k play drafs [07:16] and more [07:16] plus* [07:16] oh [07:17] not just back when they did em [07:17] fksin discus was [07:17] doing it in js [07:17] then posting it to the wo [07:17] wo* [07:17] wp* [07:17] so it would clear the cahe on us [07:17] i was like gah [07:17] turn that crap off [07:18] yea so double of ALL comments old and new [07:18] imbrandon: you falling asleep? [07:18] and each time it deployed discus would "sync " and more than double the comments [07:18] missing quite a few letters :) [07:19] yea and i really need to get this other fished so i'ma ignore chat a bit ;) [07:19] but yea its 74mb now, heh [07:19] quite nice [07:19] SpamapS: you want to review :D Locker? :P [07:21] well the good news is the departed hook is called when subordinate units are removed [07:21] so ELB can scale up/down properly [07:22] nice, cant wait to try it here in a bit [07:23] only real trouble is that its hard to know when it is ok to delete the ELB fully [07:23] its also really hard to know when its ok to add/remove availability zones [07:23] but I err on the side of enabling them [07:23] so dont, only do it on config chang and make an option [07:24] imbrandon: yeah perhaps we don't need to be super automatic on that [07:24] yea its just cleanup [07:24] imbrandon: problem is juju is kind of random about AZ's unless you specify it as a constraint. [07:24] juju leaves aton of that crap now anyhow [07:24] which I think we need to work out [07:25] I should be able to set a constraint on a service like 'ec2-zone: a,b,c,d' and have it balance the service accross them [07:25] oh yea i think thats a bug imho [07:25] erm well yea but multi zone elb is mucho $$ per hr [07:26] is it? [07:26] like how much more? [07:26] yea like 4x [07:26] oh suck [07:26] Ok I'll just register it in the first AZ and print in debug log where it ended up [07:26] gotta see a man about a horse, afk [07:26] and encourage use of ec2-zone in the README [07:27] yea sounds sane [07:27] for noe [07:27] now* [07:27] i can be iterated through too if we thing about a nother way once it s in real use somewhrere too [07:27] a few places [07:28] etc [07:28] hell thats how the others get so good [07:28] is actual use [07:28] imho [07:31] yay yet another e-mail address :P === almaisan-away is now known as al-maisan === al-maisan is now known as almaisan-away [09:43] Good friday world! [09:45] Hello #juju! Does anyone have a Juju Charm for MongoDB 2.0+ ? [10:01] dpkingma: http://jujucharms.com/charms/precise/mongodb [10:01] is 2.0+ [10:02] bkerensa: thanks! [10:40] Hi again! Can anyone link to information on how to deploy with juju on my own OpenStack cloud? Can't find much info on that... [12:05] Hello there all again I was just looking over open week and am not seeing anything for juju. are you all not doing anything for it ? [12:13] When an add-unit is called, is the -relation-changed run on every unit in the given relation? or only on the new node? [13:09] should relation-list fail on a node launched by juju? [13:15] phschwartz, https://juju.ubuntu.com/docs/charm.html#hooks has a good explanation of which relation hooks fire when [13:15] and relation-list only works in the context of a relation hook [13:16] hmm, from my changed hook, it fails to output anything. But I am seeing from the joined that the relation was set. [13:16] Any change if I do a paste of the log you can take a look for me and see if you can point me in the right direction? [13:18] phschwartz, sure - can I see your charm as well? that would help [13:19] All of it or just the changed and joined? the charm works to start a single node fine. But fails when there is a join. [13:19] Here is the log snip. http://paste2.org/p/1997476 [13:21] And probally easier for you to see the base charm which I have been trying to figure out why it fails on add-unit. [13:21] https://code.launchpad.net/~negronjl/+junk/hpcc [13:21] That is a link to it in Juan Negron's repo on LP. === carif_ is now known as carif [13:33] phschwartz, ugh - so you are seeing issue trying to scale out? [13:34] I suspect that the bit of code in hpcc-cluster-relation-joined which determines the netAddress does not work that well [13:35] phschwartz, that charm needs updating to use the private-address stuff provided by juju rather than trying to figure it out itself. [13:36] i.e. relation-get private-address $MEMBER === rog is now known as Guest17887 [13:38] phschwartz, when add-unit is called, every other unit will get joined & changed event for the new unit. the new unit added will get a joined/changed for all the other units.. assuming a peer relation. [13:39] I was thinking that, but if you look at the pastebi, it looks like the relation-set is being called correctly. [13:39] relation-set name=ec2-23-20-114-5.compute-1.amazonaws.com netAddress=10.80.215.254 [13:39] relation-set name=ec2-23-20-152-190.compute-1.amazonaws.com netAddress=10.116.202.39 [13:40] the other problem with that charm is that it never removes entries from that list === asavu_ is now known as asavu [13:40] the ip temp stuff [13:40] negronjl, ping ^ hpcc [13:40] it is far from complete. Only a start of the charm I have been working towards with Juan. [13:40] jamespage, it is using the private-address stuff just resolving it as well to ip from dns entry [13:41] hazmat, thats not terrible reliable tho [13:41] Is the Flush of values for the hook correct? All of the settings changed are saying (was unset). [13:42] phschwartz, the values are automatically flushed if the hook exists successfully (exitcode 0) [13:42] ok, so that is fine. Now just not sure why it seems like that changed hook is not seeing them. [13:46] phschwartz, the change hook isn't seeing what? [13:47] returns from relation-list [13:47] phschwartz, the change hook may be executed multiple times for a single unit.. ie if the unit changes its setting [13:48] phschwartz, we have this alias expansion going on to make things a bit easier.. ie. a join always is accompanied by a change hook.. but if the unit does actually change its settings, a subsequent change hook will be invoked for it on the related units === fjlacoste is now known as flacoste [13:49] In the logs I see each run of the changed hook and I did a juju-log "IP: ${UNIT_ADDRESS}" for the address I should get with UNIT_ADDRESS=`relation-get netAddress ${MEMBER}` [13:49] But I see an error that ${MEMBER} is undefined and because of that I never get the address. [13:50] This is done for UNIT in `relation-list`; do [13:58] phschwartz, that's what i'm saying though.. the change hook will be executed subsequently with the value, because of the alias expansion it doesn't nesc have that value because its join hook hasn't executed [13:59] The errors for it are when the changed runs after the join in the logs. [13:59] I doubt that it is logging out of order. [14:05] imbrandon: github.com/charms all have precise links now... need to decide how to handle multiple series there, but at least they're correct for the moment [14:12] jamespage, you can now use relation-list outside of a relation hook if you specify -r RELATION_ID [14:13] so you just need to know the relation ids. $JUJU_RELATION_ID is specified for relation hooks; and you can use relation-ids RELATION_NAME to get a list of relation ids for that name [14:24] * jamespage stands corrected [14:24] I'd not realised that had landed [14:27] jimbaker, nice [14:29] jamespage, i still need to update the docs - right now, it's just in the specs [14:29] jimbaker, right-oh [14:30] phschwartz, re-looking at the hook - MEMBER should be UNIT or vica verse [14:33] I will try swapping them around. That and I am going to go on the nodes and try the relation-list with the -r ID to see if I can figure out what the return is and why it is failing [14:33] phschwartz, they need to be one or the other - not both :-) [14:33] oh. That might be the issue. Which is the prefered UNIT or MEMBER? [14:34] phschwartz, for UNIT in `relation-list`places each unit name in UNIT in turn [14:34] but MEMBER is then used to get the netAddress [14:35] so its up to the hook but I would use UNIT [14:35] so I should be using for MEMBER in `relation-list` and then place MEMBER in netAddress [14:38] phschwartz, something like http://paste.ubuntu.com/949906/ [14:38] which deals with hazmats point that netAddress might not be set at the point that the changed hook runs === Guest17887 is now known as rogpeppe [14:39] hmm, that would be an issue for me. I need it set or the program run below will not work correctly. [14:39] phschwartz, note that I can't type ADDRESS [14:39] lol [14:39] phschwartz, it will be called again when the address is set [14:39] ok, let me try that and see what happens [14:39] and lines 15/16 could be put inside of the conditional block [14:40] that would be a good idea [14:40] phschwartz, the point is that it will eventually get the right information -changed will fire multiple times [14:41] Would that loop only return 1 ip or will it loop more then once and place all ip's from the relation in the file. [14:41] hazmat, thats why if possible I normally use relation-get private-address $UNIT - I believe that is implicitly set [14:41] The reason I ask is I need the full list of ip's, not just one. [14:42] jamespage, yup.. that is always set [14:42] phschwartz, that loop will populate the list of ip's with all the related units that have set a 'netAddress' in their -joined hook [14:42] ok, good. That is what I want. [14:42] jamespage: m_3 marcoceppi nijaba lynxman- : clint has like 4 charms he's submitted that need reviews [14:42] https://bugs.launchpad.net/charm/+bugs?field.tag=new-charm [14:43] jcastro, good - I need a break from bug triage [14:43] I think clint is a subordinates fan! [14:43] https://bugs.launchpad.net/charms/+bug/977552 [14:43] <_mup_> Bug #977552: Charm Needed: Terraria server < https://launchpad.net/bugs/977552 > [14:43] I'll take munini-node [14:43] if someone wants to do that one he's been waiting all month ^^ [14:44] hmmm, we'll need to figure out a better github workflow at UDS [14:45] jcastro, I also want a 'review-charm BUGNUMBER' tool [14:47] jcastro, I'll take a look at that one as well [14:49] I'm starting to really appreciate subordinates [14:49] * jamespage wonders if m_3 is going todo ganglia as one soon (or whether I should jump in :-)) [14:51] jcastro: will catch up on reviews when testing is updated for precise and ec2 [14:52] ooh, I missed shelr.tv being promulgated [14:52] jamespage: by all means... the primary ganglia charm needs to use the packages in the repo, not the ppa BTW [14:52] http://jujucharms.com/charms/precise/shelr.tv <--- really awesome service [14:52] m_3, yeah - I spotted that - I'll take a sweep through ganglia next week then [14:53] there's a whole slew of subs that can be written now [14:53] jamespage: oh hey, jenkins needs a readme too, I'd like to ask the QA team to start using it [14:53] m_3, I want to try some stuff with the hadoop charm and subordinates so that both ganglia and hadoop send information to the ganglia master.... [14:53] jcastro, so does zookeeper [14:53] they are currently not using juju for jenkins and I'd like to change that [14:54] jcastro, historical.... [14:54] jcastro: I'll start with the terraria review [14:54] yeah, they're keen on doing it [14:54] they were just waiting for you know, us to ship, heh [14:54] jamespage: just did an add-unit so will let you know if that change worked. [14:55] jamespage: yeah, hadoop's a tough call on that one... there's builtin ganglia integration... but we still need a ganglia-node sub in general tho [14:56] m_3, yeah - but I'd like to faciliate that through either the subordinate OR directly with the ganglia master [14:56] yes, great idea.. config options [14:56] its working direct in the current charm with your generic hook and some reconfig of hadoop [14:57] I guess it's really a config option to enable the builtin integration.... the sub would be ignorant of that... perhaps they'd even need to work if _both_ were accidentally enabled [14:57] m_3: gonna scratch my head about that next week [14:57] 1600 on a Friday is not a good time todo that kinda thinking :-) [14:58] jamespage: the current hook is not using the hadoop builtin integration though... I'd imagine that'd potentially have more instrumented stuff... dunno though [14:58] ha! yes [14:58] m_3, the new one for precise does [14:58] cool [14:58] its a bit nasty cause you have to restart the hadoop daemons to get it to start sending metric.... [14:59] dang [14:59] * m_3 still don't have a good feel for ramifications of service restarts on data integrity [15:00] seems like there'd have to be some serious sequencing going on if you're using a lot of different bigtop components [15:01] gmb, ping [15:01] hazmat: Hi [15:01] m_3: yeah - I need some sort of super-orchestrator [15:01] hi gmb i just wanted to followup on the lp installs taking a while.. i was trying to understand what the slow points are [15:02] gmb, gary_poster had mentioned you had tried to stuff some of the bits into s3, but it was too slow to download? [15:02] hazmat, Yes. So, LP dev environments are slow to install. [15:02] We can't change that any time soon. [15:03] ~30 minutes is the norm, ~40 on a small instance. More if we're installing lxc and creating containers... [15:03] gmb, that's probably the best solution, so i'm trying to understand what's slwo about them? [15:03] hazmat, Lots of stuff. We have to: [15:03] Install build deps. [15:03] Grab the source [15:04] Build all the sdists (there are a lot) [15:04] Build mailman (takes ages, don't know why) [15:04] Build WADL for the API (also takes ages) [15:04] Create the sample DB (shorter than it used to be, so maybe not a big issue any more) [15:05] checking out the branch alone takes awhile :-/ [15:05] gmb, could you see a src tarball onto s3? [15:05] s/seed [15:06] gmb, gary_poster: If you're happy with a snapshot, you may as well snapshot that procedure by uploading the final state onto S3 [15:07] gmb, didn't you determine that S3 itself was not as snappy as you liked? [15:07] hazmat, niemeyer: Trouble is, the tarball would be in the order of 1.5G once everything is built. And as gary_poster said, s3 was pretty damn sluggish. [15:07] There are huge pipes between S3 and EC2 [15:07] Maybe they were clogged. [15:07] heh [15:07] gmb, you should be parallel download from s3 [15:08] er. able to [15:08] gmb: I was transferring 8GB just the other day.. it doesn't feel bad at all [15:09] niemeyer, Oh. In that case my experience was not representative. However, I don't have time to try it all over again today. I'll give it another shot when I'm in Oakland, since I'm going to have to re-do things on us-west-1 anyway. [15:09] gmb: Cool, please ping me if you'd like someone to be brainstorming this with [15:10] gmb: Will be there as well [15:10] niemeyer, Will do, thanks. [15:10] Excellent. There may be questions, then :) [15:11] niemeyer, gmb, we'd like speed reliability, so while a single experience of slowness might not be representative, it is worrisome: if people are unable to participate in the hack session after 10 or 15 minutes, the value is gone IIUC [15:12] Yes, exactly. But the only way to really decide is to stress-test it, I guess :) [15:16] gary_poster: There's no such thing as speed reliability when talking about shared ip networks.. that said, there are huge pipes between S3 and EC2.. if these pipes are not enough, I'm not even sure that image snapshots would help.. there's still transference going on to get image snapshots into place. [15:18] gary_poster: There are other possibilities too, like EBS, etc [15:21] ack about no complete speed reliability, niemeyer, and about transference to get image snapshots into place. I believe the image transference is cached in a way that S3 is not, but I could very well be wrong there, since that's merely an impression. EBS: yeah, I looked at that briefly for this, but simultaneous sharing seemed to not be supported AFAICT. I'm sure you know much more about this than I; thank you for your off [15:21] er to help with brainstorming with gmb. I bet that would be a big help. [15:23] niemeyer, gary_poster: My "I don't have time" turned into "This is bugging me..." much, much, _much_ faster this time, so it might actually be workable after all. [15:23] gmb, which, S3? [15:23] gary_poster, Yep. [15:23] cool gmb [15:23] took seconds instead of several minutes. I must have just hit a bad pipe :) [15:23] heh [15:24] If the max for getting a new LP instance to hack on is < 10 minutes that's pretty good, I'd guess. If the median is < 5 that would be awesome. You'll still need to run make schema I suspect [15:25] Yes. Plus all the install steps. But if I can have mailman and wadl built a lot of the problems diminish relatively quickly. [15:25] you could make the dev do this as a "learning experience" to put it in their hands sooner :-) [15:25] Heh :) [15:25] and the eggs [15:25] And those, yes. [15:27] juju is basically an appliction that controls your devops deployments. In my case I'm testing the set up of a hadoop cluster (just starting) so I need to set up my keys in my yaml config and that's it ? [15:27] I'm wanting to deploy this on ec2 [15:29] gary_poster: I'm really curious as to why you are building mailman instead of putting it into a .deb [15:30] gary_poster: all thoe sdists seem ripe for .debs [15:30] senior75151: right, just setup your EC2 access creds and then bootstrap/deploy away [15:31] SpamapS, mailman: it is a hacked version of mailman done by barry because mailman 2 was not designed to be used as a library. mailman 3 has changed that, fwiw. [15:32] SpamapS: deam that's easy!! just found the tutorial here https://juju.ubuntu.com/Documentation Thanks a lot. I honsetly spent like 2 days setting up hadoop on my local computer, ha! I'm not a devops I'm just a developer.. soo crossing fingers this works well :) [15:33] senior75151: so are we :) [15:34] gary_poster: but ultimately, it sticks a bunch of files on the filesystem, right? .debs are good for that sort of thing. :) [15:34] SpamapS: can one tell the charm which computer type to use. e.g.: which amazon ec2 instance type. Like Mid, Large, etc. While keeping the same config… Say for staging we only need like a Mid but for prod we need Large [15:36] senior75151: juju deploy hadoop --constraints instance-type=m2.xlarge [15:36] senior75151: or, more generically juju deploy hadoop --constraints cpu=16,mem=10G [15:37] SpamapS, I'm struggling with jitsu gource - does it work on precise? all that happens to me is gource whinges about the log format and everything dies... [15:37] senior75151: btw, you need to be on Ubuntu 12.04, or have the latest juju from the PPA or Brew (OS X) for that to work. [15:37] jamespage: I think you might need to try the latest version [15:37] jamespage: it was quite broken in 0.4 [15:38] senior75151, the README for hadoop has some guidance on sizing for the different roles within hadoop - you will need to tweak the config of each to make sure hadoop uses it all [15:38] SpamapS, ack - I'll give it a spin [15:38] SpamapS, sdists vs. debs: this is a years-old argument circling around what is the proper packaging for consuming Python packages for a non-system non-commodity application under development. There is absolutely no need to rehash it now for LP's sake, for many reasons. In Canonical now, fighting the deb route would be a challenge for future projects, but some might still be willing to take up that challenge. I'm on the no [15:38] n-deb side actually for this, but not enough to fight the Canonical tide. Debs are moderately alright with me for development, especially for a project that does not have >150 Python egg dependencies like LP. [15:39] SpampS: thanks for the tips will do… I'm getting an error on MACOSX goign to try to install on ubuntu. [15:40] jamespage: thanks for the heads up. WIll keep in mind. [15:42] gary_poster: but can't you basically just change the setup.py args to build a deb? [15:42] More important SpamapS, IMO it's worth noting from juju's perspective that this *is* an argument. "You're doing it wrong, package everything in debs" is not a great answer for all potential adopters IMO. [15:42] SpamapS, thanks for that pointer - works great from trunk [15:42] jamespage: probably worth making a release, though I want to talk to jimbaker about renaming 'do' [15:42] ack [15:43] jimbaker: hey, the name 'do' is too confusing. Can we rename that command to be more clear about what it actually "does" [15:43] SpamapS, certainly. "do" is very generic [15:44] SpamapS, maybe exec-in-hook-context ? [15:46] jimbaker: that seems long winded.. but maybe it needs to be :) [15:46] SpamapS, it does sort of convey the backdoor aspect of what's being done here [15:46] almost want to call it "jitsu poke" ;) [15:48] jimbaker: All the other sub commands are very obvious as to what they're for. open a port, run gource, get unit info. [15:49] jimbaker: exec-in-hook-context doesn't tell me what it is for [15:49] It requires deep juju knowledge to know what its for, rather [15:50] jimbaker: run-hook-cli ? [15:50] SpamapS, or maybe run-as-hook? [15:50] *perfect* [15:50] cool, i will make that change [15:50] jimbaker: if you can mege in a rename, I'll release as 0.6 [15:54] SpamapS, change the setup.py args: you mean of LP? No, it's not nearly that easy, I'm pretty sure. Of the dependencies: building debs from reasonably well packaged Python code is impressively easy these days from what I've seen, but it's still not clearly documented (we recently found the wrong way on the wiki and carefully tried to clean it up, for instance, only to be pointed elsewhere once we asked an expert to review i [15:54] t) and still has plenty of potholes for the uninitiated (for instance, getting deb numbering right in PPAs has tripped us up several times lately). I'm not saying that these are a problem for experts, but I am saying that it is not trivial and it is not without issues. It won't happen for LP given the current focus, I predict. For other projects, IMO the cost of packaging is not negligible and not as clearly a win to ever [15:54] yone that the juju project will encounter, which is the only message I really feel like conveying. As I said, rehashing the LP argument really doesn't have a lot of value, I think. [15:54] * gary_poster likes walls of text. So pretty. [15:55] gary_poster: setuptools has an option to automatically build a .deb instead of a .tar.gz [15:55] yes it does, SpamapS. [15:55] We also could build eggs [15:55] yeah [15:55] Which have the same characteritics [15:55] that seems better :) [15:56] but these have caused problems in the past across architectures [15:56] :) yeah [15:56] we have devs on different architectures, and as of a few months ago even different architectures in production unfortunately [15:58] gary_poster: well given all of that, I think the idea of just tarring up all your needed stuff and sticking it in S3 is the best option. [15:59] gary_poster: the EBS idea is sound too... but juju won't help you much there unfortunately [16:00] hm, unless.. I make an EBS subordinate. :) [16:00] * SpamapS cannot stop coming up with ideas for using subordinates to take over the world [16:06] SpamapS, S3: certainly in the short term, yeah. I think that's the approach that gmb is working on now. My only remaining interest in this topic TBH is that I think there's a real-world scenario that juju would ideally support easily, rather than demanding a world view of their potential adopters. "Don't do it that way" responses may be appropriate and necessary sometimes, but should be challenged and probably even docume [16:06] nted (in part to make sure there are not too many of them). I understand from other conversations that this use case isn't necessarily all that easy to handle, particularly within the juju vision. But "the real world is messy" may be a reasonable response to some of this. [16:07] (and so juju should consider handling at least some of the messiness) [16:07] but this has gone well past the immediate needs of my squad :-) [16:08] so...yeah! yay, subordinate charms! ;-) [16:08] gary_poster: LP is not alone in the long-build-time camp [16:09] Yeah, that's really all I'm saying [16:09] gary_poster: in fact I see *huge* value in being able to do things like run your full test suite on the first deploy of something, and then snapshotting that so its a known good build. [16:09] All that I care about communicating, at least ;-) [16:09] +1 yeah [16:31] SpamapS, i renamed "jitsu do" to "jitsu run-as-hook", so it's ready for release now [16:31] jimbaker: nice... that makes sense [16:37] jimbaker: thanks! [16:37] also just added --version to jitsu [16:38] anybody want to squeeze anything else into juju-jitsu 0.6? [16:38] SpamapS: yes, but it's not going to happen this week... so perhaps 0.7 [16:43] * SpamapS pulls the release lever then [16:44] \0/ [16:46] I made a little into video to juju for youtube today. once I start to understand more about It I am going to make many many more :) So flippen cool after I watched the classroom video I understand the power that it could have. [16:46] bobweaver: awesome!... links? [16:47] http://www.youtube.com/watch?v=495V7FokwBU&webm=1 [16:47] It started out as how to file a bug [16:47] but I just could not help but talk about juju [16:51] :) [17:02] Hi everyone, I saw there was a fix-committed bug for a Cloud Foundry charm. Has anyone tried it out? [17:09] mars: negronjl wrote those.. and there are some issues with them.. namely that they use their own "special" mysql/mongodb/etc. charms. [17:09] negronjl: ^^ whats happening with the cloudfoundry charms? anything? [17:09] SpamapS: i haven't done anything with them. [17:09] SpamapS: What's the main issue ? [17:11] I just wanted to try them out, to see if I could use them to get a Django app onto a juju/cloudfoundry environment. [17:13] I see the instructions are in the bug, but I was wondering if anyone else had tried them out. [17:13] negronjl: that they don't use the stock mysql/mongodb charms IIRC [17:14] SpamapS: CF adds a lot of "glue" code to their db's. Enough to justify having their own charms [17:14] negronjl: we discussed this in the past, that this glue would still benefit from living in the main mysql charm. [17:15] negronjl: so that you can, say, tune the mysql with the same 'juju set' commands [17:15] SpamapS: What we didnt realize is the amount of glue that is required .... There are several daemons that need to "live" where the mysql/mongodb daemon is at as well. [17:16] SpamapS: We would have to add substantial code to the mysql/mongodb charms to make them "play nice" with CF [17:16] SpamapS: doable yes ... advisable ... not so sure [17:16] daemons?? [17:16] like, mysql proxy? [17:16] negronjl: those daemons can be started when the relationships are established. [17:17] negronjl: I'm mostly concerned about fracturing the space.. if you do something awesome for CF's mysql that is a generic mysql improvement.. it should benefit all juju mysql users if possible. [17:17] Isn't making things live on the same unit as a service what subordinate services were designed to do? [17:18] mars: in this case.. not really. If its directly related to the service working or not, it belongs in the main charm, as part of the relations. [17:18] ah [17:20] negronjl: if its a ton of work, no big.. but it should still be a goal. I am not convinced that anything "special" that CF does to the mysql server is special enough to keep it out of the main mysql charm. [17:20] The idea is to group devops knowledge around the best way to run a service in one place. [17:20] SpamapS: Let me finish a call and I'll talk to you more about it [17:21] SpamapS: Give me a few minutes [17:21] and fork only when necessary. :) [17:21] negronjl: sure no problem. Just triaging bugs. :) [17:21] I just changed my reviewboard charm from postgresql to mysql: three lines changed in the hooks, plus the juju commands to upgrade and deploy the new services. Wow, very cool. [17:24] mars: you can leave both in there if you want. "requires" is a bit misleading.. admins can choose one or the other. [17:47] <_mup_> Bug #989967 was filed: juju should ignore dotfiles in local charm repositories < https://launchpad.net/bugs/989967 > [17:56] how to set a default on juju [17:56] i get this error [17:57] There are multiple environments and no explicit default (set one explicitly?) [17:59] SpamapS: There is significant work that would be needed to add the CF bindings into the existing mysql/mongodb charms. [17:59] SpamapS: It can be done but, I am not sure we should do it [18:00] SpamapS: Another way would be to use the cf-specific charms to launch the initial DB ( ie: mysql master ) and have the regular charms do the slave ones [18:00] SpamapS: For mongodb, It would work in a similar mannger ( cf-mongodb as master and regular mongodb charms as slaves ) [18:16] boolean config params are _much_ harder to use than they were before typechecking was enabled [18:21] negronjl: I'll take a look.. I think I need to understand the problem space better before I expect you to do more work. :) [18:22] nathwill: SpamapS and jcastro are the chiefs here [18:22] :D [18:22] yeah.. i looked at the cloud foundry stuff its non trivial for generic usage [18:22] thx bkerensa [18:23] anyone got tips on how i can troubleshoot issues with locally deployed services ending up with agent-state: pending, public-address: null? [18:23] senior75151: you can add default: in your environments.yaml for that [18:23] senior75151: mine is "defaults: ec2" [18:24] senior75151, you can specify a default: env_name in your environments.yaml.. or specify one directly on the command line with the -e flag.. or specify one with the JUJU_ENV environment variable [18:24] nathwill: its possible the first machine is still downloading/booting [18:24] nathwill, the first provisioning on local service can take a little while [18:25] bobweaver: I'd be more than happy to link to any juju videos you make from juju.ubuntu.com! [18:25] we need a better log watcher for that [18:25] debug-log doesn't show anything IIRC [18:25] nathwill, ps aux | grep lxc [18:25] spamaps, hazmat: this is the bootstrap? [18:25] kk [18:25] nathwill, post bootstrap.. adding the first unit to a local env, has to download a bunch of packages.. [18:25] effectively a minimal distro [18:25] jcastro: thanks [18:25] ok how does one search for like juju instances [18:25] ala apt-cache search [18:26] I tried juju deploy hadoop [18:26] didn't work [18:26] nathwill, if your on fast conn it can take like 10m [18:26] I'm sure there has to be a site or a CL command to search [18:26] senior75151, http://jujucharms.com [18:26] hazmat: thank [18:26] err… Thank you. [18:26] hazmat: ok.. thanks for the tip. i see a lot of lxc activity in ps, so i'll give it some time. [18:29] sooo if one does precise on the configurations on amz it works.. ? has anyone tried? [18:29] that is for the default-series: precise [18:29] hazmat, what do you mean by 'non-trivial for generic usage'? [18:30] yep, it does [18:30] mars, just the cloud foundry stuff has some specific bits in each of their services which are not nesc. appropriate for generic usage [18:30] jcastro:thanks [18:30] senior75151: I do believe that just not setting a default-series defaults to precise, but not sure (hazmat or spamaps?) [18:31] hmm [18:31] jcastro, it does when generating it, but that key/value needs to be present in the file [18:31] hazmat, but it still works fine for end-users? [18:31] ah ok [18:31] mars, you mean cloudfoundry will? or the cloud-foundry services used outside of cloud-foundry? [18:31] hazmat, just to play around with [18:31] im assuming im doing something wrong here using relation-ids + relation-list + relation-get, any ideas? http://paste.ubuntu.com/950464/ [18:32] mars, i haven't tried the cloud-foundry charm in a while, but it works afaik [18:32] mars, negronjl is the man to ask when it comes to juju+cf [18:32] jimbaker: ^^ (adam) [18:32] adam_g, woah [18:33] adam_g, you should have something in the unit agent log as well [18:33] adam_g, could you paste that also [18:33] mars: hello [18:33] hazmat: sure, one sec. the charm log is spewing: juju.state.errors.RelationStateNotFound: Relation not found [18:34] adam_g, could you paste one of the tracebacks [18:34] pastebin that is [18:34] hazmat: unit agent log == charm.log ? [18:34] adam_g, yup [18:34] negronjl, unping, just finishing off that earlier conversation about cf :) [18:34] mars: np [18:35] hazmat: http://paste.ubuntu.com/950474/ [18:35] jcastro or SpamapS can you help nathwill with a juju issue he is having :D [18:35] senior75151: to search the charms, install charm-tools, and use 'charm search' [18:35] senior75151: its *very* limited [18:35] senior75151: basically just greps the list [18:36] nathwill: what's the problem? [18:36] senior75151: there are plans to ad full search capabilities into juju though [18:36] adam_g, so no relation state for that relation id [18:36] jcastro, not much right now, trying to wait for the 15 min recommended by hazmat... [18:36] thanks. i'll hit you up if it's not working after that [18:36] jimbaker, but he has the that rel-id immediately proceeding as the output of relation-list [18:37] er. relation-ids [18:39] hazmat, yes, that definitely looks like a problem - it's with respect to that service unit [18:39] SpamapS: [18:39] thanks [18:39] swift-proxy/1 [18:39] How does one know how many instances a charm will launch, Like I did juju deploy hadoop-master and it deployed 2 small instances [18:39] does the charm-tools give you that info [18:39] say I do charm search hadoop-master —info [18:39] or smth [18:40] like that [18:40] dunno [18:40] senior75151: charm will only ever launch one per deploy [18:40] ohh ok [18:40] senior75151: unless you pass '-n #' [18:40] the other thing here is that we need to register RelationStateNotFound with amp so it doesn't default to the unknown error [18:40] senior75151: add-unit/remove-unit to control the size of the service itself [18:40] soo weird, the first deployed failed… (as in the error message) but succeeded on ec2 [18:40] senior75151: bootstrap launches one as well [18:40] ohhh [18:40] ok [18:40] jimbaker, yes on relstate not found.. but what do you mean respect to that unit [18:40] so that must be it [18:41] SpamapS: soo one more question. Now that it's deployed I just write apps against those services and that's it ? no way! that's awesome === koolhead11 is now known as koolhead11|afk [18:42] hazmat, i would expect that if relation-ids returns an id, it's there for subsequent use in the hook (such as this debug hook session) [18:42] Also I did $man juju; juju help deploy … how do I know the switches for deploy [18:43] i don't explicitly test this scenario with respect to debug hooks, but it's not what i would expect. anyway... i will take a look at this issue [18:43] say I want to give a name to the instance i deployed where to look for this info [18:44] jimbaker: i can try executing during normal hook execution (outside of debug-hooks) if you'd like [18:44] adam_g, sounds good [18:44] jimbaker, aren't you assuming its not there anymore? [18:45] adam_g, could you pastebin your status? [18:45] adam_g, you didn't modify the rels/units while you where in debug hooks? [18:45] senior75151: juju deploy --help [18:46] hazmat: no, i havent touched anything. im just trying to query the private-address for each of those relations. [18:46] jimbaker, so there's a bug of a different sort it sounds like [18:46] hazmat, indeed [18:46] adam_g, thanks [18:47] juju status -> http://paste.ubuntu.com/950511/ [18:49] SpamapS: sweet thanks… [18:51] SpamapS, how are you release juju-jitsu? [18:51] SpamapS, into the distro? [18:51] SpamapS, ppa .. or pypi? [18:52] hazmat: no change :( http://paste.ubuntu.com/950525/ [18:52] hazmat: jimbaker i should mention that only fails when im executing those against relations other than the current, if that makes sense.. [18:52] nathwill, what version is the host? [18:53] nathwill, is the host precise as well? [18:53] hazmat precise upgraded from beta 2 [18:53] yessir [18:53] or ma'am... [18:56] :-) [18:56] nathwill, any other files in /home/nathan/.juju-data/nathan-sample/units/mysql-0/ .. ie. is there an agent log there? [18:57] SpamapS: charm policy question ... your spdy charm installs something beta. Do we care about the state of an upstream whenn considering putting something in the charm store? [18:57] hazmat: container.log, output.log, unit.log [18:59] hazmat: unit.log has some zookeeper handle_socket_error socket connection timeout msgs [19:00] nathwill, could you pastebin the unit.log [19:01] hazmat: sure thing [19:01] that's from the unit agent which is what's setting values reported by the status cmd [19:02] hazmat: http://paste.ubuntu.com/950553/ [19:03] nathwill, do you have any firewalls/iptables on that machine? [19:04] hazmat, running ufw in default deny mode [19:04] nathwill, bingo.. that's the problem [19:04] heyo... alright, i'll disable, destroy-environment and try this again :) [19:05] nathwill, you might not even need to do that if your up for modifying the firewall rulz live.. the container agent will continually try to poll [19:05] to connect to zk [19:05] but its probably better to destroy-env and start again [19:06] nathwill, the host has a zookeeper instance binding to the bridge address 192.168.122.1 .. and the containers which are connected on the subnet off the bridge attempt to connect back to the zk instance [19:06] sounds like the firewall rules stop that connectivity, which is the problem [19:06] hazmat, ok, that makes sense. [19:11] I just saw my db-relation-changed hook get executed twice in a row. Is that normal? [19:12] hazmat, i think that got it, the units have public-addresses now, and tailing the unit.log, i can see them pulling in packages [19:12] hazmat: thanks so much for your help :) [19:18] how come http://pastebin.com/zmGqs8Qc I get an install error [19:18] even though the deployment was 'successful' [19:18] so I was thinking of a way to make a video for juju and would like too interview one of you. Is there any way that I could get some one on one time with one of you ? [19:18] juju status says agent-state: install-error [19:20] I was thinking like Saturday ? If I could set up interview ? ping me if you like the idea [19:23] adam_g, that's what we call the "implied relation". were you able to run this in normal hook execution? [19:25] jimbaker: working on that now. strange, im actually using the same logic elsewhere in the charm with another relation, and it seems to have worked fine until i hit that problem, then it became broken. [19:26] hazmat: definitely working now :D time to take this puppy for a test drive! [19:27] nathwill, sweet! [19:28] senior75151, so basically that means the machine is up and running, but the charm install hook had an error [19:28] senior75151, and the unit transitioned to an error state [19:28] ohh hazmat: thanks… so how to fix [19:29] senior75151, you can use debug-log in conjunction with resolved --retry to watch what goes wrong.. or you can log into the machine via juju ssh hadoop-master/0 and look at the charm.log at /var/lib/units/hadoop-master-0/charm.log [19:29] redeploy I guess ? [19:29] senior75151, the juju resolved --help [19:29] ahh thanks [19:29] excellent will do [19:30] jcastro: re spdy being beta.. no. We only care (for now) that it is verifiably the source you intended. [19:30] hazmat I think that might be the problem juju ssh: error: unrecognized arguments: hadoop-master/0 [19:30] jcastro: once we have tagging and default search.. we will start being more picky. :) [19:31] it prolly couldn't do it… I'm going to change from precise to oneric and see if that helps [19:31] senior75151, hm.. that seems like a bug in juju ssh [19:31] jimbaker: oh, disregard that. it just happened to be working because of the ordering of the relations (it was only ever being executed from an implied relation). it looks like the same issue happens during non-debug-hooks execution [19:31] senior75151, you can juju ssh 1 [19:32] to login directly to the machine the hadoop-master unit is running on [19:32] $ juju ssh 1 [19:32] usage: juju ssh [-h] [-e ENV] unit_or_machine [command] [19:32] juju ssh: error: unrecognized arguments: 1 [19:32] how can that not work but juju status work? [19:32] weird [19:32] huh [19:32] were it me, I' use juju debug-hooks hadoop-master/0 [19:33] SpamapS, debug-hooks takes some explaining.. [19:33] juju ssh works fine for me.. something is fishy [19:33] soo that worked [19:33] I think what happened [19:34] was that I didn't accept the public key [19:34] cuz when I did debug [19:34] it asked me if I would accept the public key [19:34] err…. fingerprint [19:34] for ssh [19:34] and then i'm in... [19:35] debug-hooks puts you in the charm root [19:35] and you get the right context with the right env vars [19:35] got you…. [19:35] The only confusion is that you have to exit the shell when you want juju to run the next hook [19:35] i think it automatically starts screen or smth [19:35] senior75151: right [19:35] tmux, but yeah [19:36] Its weird, sometimes I get tmux+byobu, sometimes just tmux [19:36] anoying cuz i'm using tmux on my host [19:36] adam_g, thanks for the feedback [19:36] so now is double tmux [19:36] errrr... [19:38] soo I think I found the error but not sure why… Started service unit hadoop-master/0 [19:38] 2012-04-27 18:33:55,073:4640(0x7f7d675b6700):ZOO_WARN@zookeeper_interest@1461: Exceeded deadline by 57ms [19:38] that's from /var/lib/log/machine-agent.log [19:38] there is a bunch of timeouts [19:38] any ideas ? [19:40] I don't think is the precise vs oneric [19:40] doing an lsb_release -a shows 12.04 [19:40] soo i think that part is not it… wondering. [19:41] senior75151: no thats just noise [19:41] senior75151: the install error will be very clearly shown in the charm log [19:41] senior75151: /var/lib/juju/units/hadoop-slave-0/charm.log [19:44] jimbaker: np, ill keep the environment up for a while. let me know if there is anything you'd like me to test, or i can give you ssh access [19:45] adam_g: you have a knack for finding all the weird bugs in juju :) [19:46] SpampS: http://paste.ubuntu.com/950654/ [19:46] Couldn't find any package by regex 'hadoop-0.20-namenode' [19:46] basically [19:46] didn't isntall [19:46] :( [19:47] senior75151: aha! [19:47] i gess that's a bug ? [19:47] perhaps, looking at the charm now [19:47] k [19:47] apt-add-repository ppa:mark-mims/hadoop [19:48] m_3: hey! that doesn't look right [19:48] m_3: don't we have a better place to get hadoop than your PPA ? [19:50] adam_g, sounds good, thanks [19:51] SpamapS: do you know if the oneric works ? [19:52] senior75151: probably, but I think we can fix precise fairly easily [19:52] I think we just need to point it at https://launchpad.net/~hadoop-ubuntu/+archive/stable [19:52] command line argument ? [19:52] or smth [19:54] I'm deploying with that new PPA right now [19:54] SpampS: wow! awesome… [19:55] :) [19:55] senior75151: the power of debug-hooks :) [19:56] man…. this IRC is awesome! hehe...how to reinstall on the same host? [19:57] i'm hoping i figure juju out well enough to port our shitty datacenter to ec2 soon :) [19:57] senior75151: simplest thing is to just deploy again and throw that host away [19:57] sounds good. [19:57] soo on the host that has juju [19:57] sudo apt-get update ? [19:57] first [19:57] or is all dynamic [19:57] how to blow away the cache [19:57] so it doesn't use the local copy if any [19:57] oh you want the new charm? [19:57] I'm still testing that :) [19:57] ok [19:57] I'll wait [19:58] ok I terminated the instances [19:58] adam_g, i was able to reproduce your error, so i will take a look at what's going on here [19:59] adam_g, feel free to destroy that env [20:01] senior75151: initial test looks good [20:01] ugh, except the part where we have to download 30MB of hadoop package from ppa.launchpad.net ;) [20:01] SpamapS: sweet… [20:01] haha [20:01] * SpamapS wishes it were in the archive. :-P [20:02] senior75151: let me run a terasort just to be sure it works [20:02] sure thing... [20:04] * senior75151 goes to HN for a second while SpamapS's test pass [20:06] yeah just waiting for slave (also needing updates) to deploy [20:15] senior75151: looking better, just running my terasort now [20:18] SpamapS: excellent [20:18] Good evening gentlemen [20:21] 12/04/27 20:21:41 INFO mapred.JobClient: map 24% reduce 0% [20:21] woot [20:22] :) [20:23] senior75151: ok, so it will take a bit for that to publish into the charm store. [20:24] excellent [20:24] how long ? [20:24] senior75151: good question! [20:25] niemeyer: I just made commits to two branches in launchpad. How long would you expect it to take for the charm store to see them? [20:25] senior75151: you can pull them locally if you want [20:26] senior75151: mkdir -p ~/charms/precise && cd ~/charms/precise && charm get hadoop-master && charm get hadoop-slave [20:26] senior75151: then 'juju deploy --repository ~/charms local:hadoop-master' [20:26] senior75151: not nearly as easy or automatic, but .. it gets you the latest crack. :) [20:27] ahh k [20:27] let me give that a go [20:27] thanks soo much [20:27] going to try it and let you know [20:30] fetching charm === benji___ is now known as benji [20:32] deploying... [20:37] SpamapS: state: pending [20:37] what does that mean ? [20:37] i suppose is still configuring [20:37] ? [20:37] or smth [20:37] or do I have to 'publish' the charm [20:37] is not enough to 'deploy'? [20:42] soo juju ssh doesn't work [20:44] SpamapS: is that what juju expose is ? [20:46] senior75151: instance-state means you are waiting on Amazon [20:46] senior75151: agent-state means you are waiting on your instance to start the agent [20:47] I have seen a few times where agent-state says 'not-started' but it is stuck.. and a 'service juju-machine-agent restart' fixed it [20:47] senior75151: juju expose opens the ports that the charm thinks should be exposed. [20:48] ohh ok [20:48] so I should def… do that for hadoop [20:48] cool [20:48] thanks a lot SpamapS [20:49] So I assume is typical to set up a 'dev' or staging environment for a developer then destroy it at the end [20:50] senior75151: saves money that way :) [20:54] that's pretty awesome... [20:55] soo say I have hadoop, mongodb, mysql, nginx… etc.. I can build my own script and that basically runs juju deploy X or can you save that config ? [20:55] once you have it running [20:55] does 'destroy' terminate the machines or…. turns them off ? [20:55] ec2 speak [21:13] senior75151, terminate [21:15] well [21:15] destroy-environment terminates [21:15] destroy-service only de-allocates [21:15] which, btw, I think we should re-think [21:15] it makes sense if we clean up what we did [21:15] but right now we can't do that, so we should just terminate the machine when we remove-unit and destroy-service [21:22] ahh got you [21:23] thanks for being such an awesome source of info [21:23] :) [21:24] SpamapS does your juju ssh [21:25] work ? [21:26] Alright! I have a reviewboard charm up and running. SpamapS, I used the Drupal charm from the tutorial as a template; can you suggest and example I could look at next to make this new charm cleaner? [21:27] senior75151: not sure I understand what you are asking. [21:27] so i do juju ssh hadoop-master/0 [21:27] doesn't work [21:27] then juju ssh 1 [21:27] doesn't either [21:27] mars: got it in a branch somewhere? [21:27] but I do juju status [21:28] and it shows is running with correct tcp ports open [21:28] SpamapS, not yet, I'll push it [21:28] senior75151: juju ssh works fine for me with machine number or unit id, yes. [21:28] hmm weird. thanks [21:34] SpamapS, pushed: lp:~mars/charms/precise/reviewboard/trunk [21:34] SpamapS, one thing I would like to figure out is how to unroll running the application setup script after the XXX-relation-changed hooks. [21:34] The Drupal example does the site setup once the DB settings are available [21:35] I copied that pattern, but I have /two/ settings that I need available before running the site setup: db, and memcache [21:36] mars: what I do in that case is just write settings to a file, and then run a shared script that checks for both, and once they are satisfied, runs the setup === almaisan-away is now known as al-maisan [21:37] mars: pip installs, btw, are not suitable for charm store inclusion btw. :-/ [21:37] mars: they do nothing to cryptographically verify the sources [21:38] * flepied found that juju-create has the gateway ip hardcoded and that was making my setup fail... [21:39] mars: looks good other than that. :) [21:39] mars: ideally with this case, you'd be able to configure the cache/db independent of rb-site [21:40] mars: otherwise it will be hard to do things like migrate to a new database service [21:42] SpamapS, +1 for saving the settings. That is a good idea. [21:43] SpamapS, as for pip: reviewboard recommends easy_install for installation: pip is better, obviously, but neither does any verification. [21:43] * mars wonders if the .tar.gz on PyPI has an md5 [21:43] mars: I would imagine that rb-site just writes to settings_local.py .. so what you really need is to just regenerate that file in total every time relations change [21:44] Yep, that's what I was thinking [21:44] mars: pypi has no SSL [21:44] 10 subordinates, SpamapS has like 7 of them [21:44] mars: so you have to do all the MD5's yourself anyway [21:44] Save a copy as a template, then rewrite it, and run rb-site once everything is available [21:44] hazmat: heh.. I have 2 more in the works [21:47] Ah darn, PyPI doesn't have a signature, but reviewboard.org does have sha256 for each release. [21:49] hazmat, out of curiosity, what is the list? I read the subordinate charm spec, but it doesn't really give a good idea of the types of subordinates I may consider writing [21:51] It's a proposal, not a tutorial or design guide :) [21:51] mars its kinda of raw.. but here it is .. http://pastebin.ubuntu.com/950950/ [21:54] hazmat, ran it through ipython: http://pastebin.ubuntu.com/950963/ [21:57] * hazmat adds a bug for charm browser to view just subordinates [21:58] SpamapS: we need to remove the hadoop-xxx charms from precise... they're only good for oneiric. precise is just lp:charms/hadoop [21:59] which should be pointing to the correct packages... in universe. I think [21:59] SpamapS: the readme in lp:charms/hadoop is the place to go [21:59] m_3: sooo I should be juju deploying to oneric only? [22:00] senior75151: unfortunately, it's different charm names for oneiric than it is for precise. They _might_ have made lp:charms/hadoop backwards compatible with oneiric, but I don't know offhand [22:02] SpamapS: dang, wish I would've caught this before you pushed changes to the lp:charms/hadoop-xxx branches... those need to be deprecated, but I don't know how to do that in lp [22:02] ok, good to know. [22:02] sooo I'll destroy the thing and use oneric for now. [22:02] thanks [22:03] senior75151: here's the doc: http://bazaar.launchpad.net/~charmers/charms/precise/hadoop/trunk/view/head:/README.rst [22:03] senior75151: no, actually, I recommend using precise, but using the precise charm [22:03] i.e., lp:charms/hadoop [22:04] I'm sorry, I know that's not clear from the list of charms in the store.. we need to figure out how to remove lp:charms aliases from deprecated branches [22:04] m_3: HAH ok yeah those should have been deprecated yes [22:05] m_3: hence the need for 'unpromulgate' [22:05] SpamapS: they should be just lp:~charmers/charms/oneiric/hadoop-xxx/trunk [22:05] m_3: should just be the reverse of the calls that promulgate makes [22:05] b/c they're still good for oneiric [22:05] oh, really? I was poking around but couldn't figure out how to remove the alias without deleting the branch [22:06] SpamapS: btw, we still have to fix several of adam_g's branches too... they're hosed [22:06] m_3: sooo I did it wrong then, cuz all i did was juju deploy hadoop. Where does one look to know how to install a charm precisely [22:06] m_3: its doable, I just don't know how [22:06] like this readme for example [22:06] senior75151: http://bazaar.launchpad.net/~charmers/charms/precise/hadoop/trunk/view/head:/README.rst covers the latest info for precise [22:07] that covers for hadoop [22:07] i'm saying in general [22:07] oh [22:07] say I want to deploy mysql and nginx [22:07] and mongodb [22:07] and .... [22:07] in general, most of the precise charms are working... and deployable from the store [22:08] hmmm well.. great, but how can I know for each one wether it does or doesn't. It's important for production to know if I've done it right [22:08] senior75151: thought you were just asking about hadoop (and naturally getting confused because of the stale links in the repo) [22:08] thanks.. [22:09] SpamapS, could you point me to the charm that you used that 'save the relation settings for later' recipe in? [22:09] senior75151: working on charm tests right now as a matter of fact... it will look something like charmtests.markmims.com when done [22:09] senior75151: but those are still oneiric... precise should be up this weekend [22:10] m_3: thanks… will they be there on this same site [22:10] or…. [22:10] senior75151: we'll be testing only the charms in the charm store... so to have them autotested here, you should submit them for review and inclusion in the store [22:11] charmtests.markmims.com is temporary... we'll roll them up to jenkins.qa.ubuntu.com next week [22:11] ahh good to know... [22:11] I'm not developing charms, just using them now.. I tihnk I just started using this a couple hrs ago and still confused about somethings [22:11] like why ssh doesn't work and that sorta thing [22:12] meaning juju ssh [22:13] senior75151, i'm confused why it doesn't work for you either [22:13] senior75151: understand... we're still working on putting the overall "experience" together for new users... with the charm store open, we should be good once all the links have been fixed for the precise charms [22:13] does juju run as root [22:13] senior75151: yeah, juju ssh is one of the parts that usually works pretty well [22:13] like I need to add the mongodb-hadoop library [22:13] and wondering if I need to install as root [22:14] err.. easy to to fix [22:14] look at permissions on folder [22:14] senior75151, the agents on container run as root, and charms execute as root [22:14] ah, gotcha [22:14] yeah.. dunno.. I assume juju is installed correctly since it deploys… but juju ssh gives me errors. [22:14] is fine cuz I can do debug-hooks [22:15] and it works [22:15] wow, that's strange [22:17] senior75151: can you paste the whole error you see? [22:21] SpamapS: I see that james built oneiric versions of the hadoop-ubuntu ppas too, so your changes should stay when we revert those packages to lp:~charmers/charms/oneiric/... the charms have never been tested against those packages though, so I'll make a note to do that [22:22] s/revert those packages/revert those charms/ [22:24] SpamapS: sure [22:24] thing [22:25] hold on installing something on the charm… weird that hadoop depends on java but didn't install java [22:27] $ juju ssh hadoop-master/0 [22:27] usage: juju ssh [-h] [-e ENV] unit_or_machine [command] [22:27] juju ssh: error: unrecognized arguments: hadoop-master/0 [22:27] m_3: its not reverting.. its porting [22:27] SpamapS: that's the error [22:27] I tried 0 and 1 as well as haddop-master [22:27] with the same error [22:27] though I think that 0 is the host pc that has juju as an app running === al-maisan is now known as almaisan-away [22:28] let me know if you need anything else [22:30] ok sooo it says that the instance is running and the ports are open and the machine is started whyen I do juju status [22:30] but i sshe'd into the machine [22:30] and do a ps aux | grep hadoop [22:30] and nothing comes up [22:30] sooo it didn't start [22:30] that's weird [22:30] prolly because lack of the java sdk on the server [22:30] any ideas why it didn't install the requirements [22:31] m_3: perhaps until we can remove hadoop-* we should put a big warning in the description? [22:31] yes, that's a great idea [22:31] SpamapS: well... now that I think about it [22:31] we should think about ways to express "replaces" like dpkg does [22:32] any idea how to restart the service [22:32] without reinstalling ? [22:32] like ideally you'd say 'juju upgrade-charm my-hadoop-master' and it would say "this charm has been replaced by 'hadoop', see README for migration instructions" [22:33] SpamapS: the big reason we were going to keep them for oneiric and the new one for precise is the packages... the precise and oneiric packages were in different ppas [22:33] SpamapS: but if the new ppa works for both releases, then that simplifies things [22:34] SpamapS: http://www.youtube.com/watch?v=5-SmNPjMcRQ [22:34] SpamapS: I can ssh into the machine later and test anything else you need me tooo... [22:34] thanks for the support [22:34] juju looks great [22:34] I'll try oneric see if that works better [22:35] I'll drop by here when I get a minute [22:35] thanks again [22:35] m_3 too [22:35] later [22:35] * senior75151 has to rush to dinner [22:35] m_3: we also can't be removing charms from an existing release :) [22:36] have a good weekend folks [22:36] hazmat: u2 [22:38] SpamapS: understood [22:52] m_3: which one is "right", rabbitmq, or rabbitmq-server ? [22:53] SpamapS: rabbitmq-server [22:53] bzr: ERROR: Not a branch: "bzr+ssh://bazaar.launchpad.net/~charmers/charms/precise/rabbitmq-server/precise/". [22:55] SpamapS: so it's really confusing... we _had_ a lp:charms/rabbitmq-server that pointed to an oneiric branche [22:55] adam also had a lp:~charmers/charms/precise/rabbitmq-server/trunk [22:56] as well as several branches stacked on top of the latter [22:56] m_3: so promulgating his branch should have worked.. no? [22:57] then when we pulled the crank last weekend (previous weekend? who knows)... we changed the alias to lp:charms/rabbitmq-server and it promoted the oneiric branch over the other [22:59] so lp:charms/rabbitmq-server has been placed at the base of a stack of branches... it can't be removed unless the branches stacked above it are removed [23:00] m_3: but we never stack on the alias [23:01] SpamapS: additional complication is that I merged that branch to bring it up to the same version as the previous precise branch... thinking this would be enough to get it working [23:01] m_3: I think we only need to promulgate the one we want as lp:charms/rabbitmq-server .. the others don't matter [23:01] there're merge proposals and additional branches associated with the _other_ precise branch [23:01] so we'd be dumping those [23:02] it's really not that big of a deal here b/c it's adams... he can delete and/or fix things the way he'd like. We need to make sure we don't do this again [23:03] so the series promotion needs to do something different when there are existing branches for that charm on that new series [23:17] m_3: I love how Ebay's hadoop is RHEL for the namenodes, but CentOS for the data nodes.. perfect example of how RHEL doesn't scale out :-P [23:18] SpamapS: $$ [23:42] <_mup_> juju/debug-relation-hook-context r533 committed by jim.baker@canonical.com [23:42] <_mup_> Log actual caching of relation hook contexts === flaviamissi_ is now known as flaviamissi [23:56] marcoceppi: welcome