[00:33] jcastro: any idea why django wouldn't be able to connect to postgres when db-relation-changed is called, but can connect after that? [00:34] I'm getting the right username and password now [00:35] and I can ssh into the django unit and manually run the manage.py commands from there and they connect to postgres just fine [00:35] but running the same commands inside the db-relation-changed hook and they fail [00:37] http://paste.ubuntu.com/1003969/ is my db-relation-changed hook [01:06] mhall119, what's the output? [01:09] SpamapS, ping [01:12] hazmat: output of which? [01:17] hmm, ok, it looks like postgresql's db-relation-joined creates the user account, then passes the login info to summit's db-relation-changed which tries to connect and failed, then postgresql's db-relation-changed actually gives the summit unit access to connect with those credentials [01:18] so I need to run summit's db initialization calls *after* postgresql's db-relation-changed hook, not before, how do I do that? [01:18] hazmat: jcastro ^^ any ideas for me? [01:19] mhall119, output of the db-rel-change hook [01:19] hazmat: see above, I seen what's happening now [01:19] gotcha [01:20] mhall119, i'd fix postgres charm to not put the into into the rel till its ready to be used [01:20] that sounds like the fundamental issue [01:20] its telling other things about accounts to use, that its not ready to accept [01:20] what's the difference between -joined and -changed for hooks? [01:21] mhall119, joined is executed when a remote/related unit is first seen (for each remote unit).. changed is execute after join, and after a remote unit changes its settings [01:22] so really postgres should be giving access from the remote unit during it's -joined hook, not it's -changed [01:23] unless there's something in my summit charm that would be called after -changed on postgresql [01:23] mhall119, well.. have you verified/echo'd the username/password.. they might not be set the first time the summit's changed hook is invoked? [01:23] yeah.. you are.. [01:23] * hazmat read the hook again [01:24] hazmat: yeah, they get into the juju_settings.py [01:24] hmm.. so the problem seems to be [01:24] http://jujucharms.com/charms/precise/postgresql/hooks/db-relation-joined [01:24] but postgres has pg_hda.conf or whatever that acts like a firewall, and until that allows my summit unit access it can't run syncdb or migrate [01:24] creates the user [01:25] but it waits till changed to give net access [01:25] http://jujucharms.com/charms/precise/postgresql/hooks/db-relation-changed [01:25] which is decidedly odd imo [01:25] and a bug [01:27] and it should be reloading instead of restarting sighup style [01:28] alternatively it should set some sentinel value on the rel [01:28] that the creds are really usable [01:28] hmm.. ic [01:28] this wants an ip dance [01:29] oh.. it used to, yeah.. this is just odd [01:29] I'll ask m_3 in the morning, it looks like he is the original author [01:38] mhall119, try using this postgresql charm.. juju deploy cs:~hazmat/precise/postgresql [01:38] i fixed it to give the access in db-relation-joined [01:40] hazmat: running now, I'll let you know [01:40] its at lp:~hazmat/charms/precise/postgresql/trunk is the store hasn't bundled it yet [01:52] 2012-05-23 21:52:31,555 unit:postgresql/0: hook.output ERROR: /var/lib/juju/units/postgresql-0/charm/hooks/db-relation-joined: line 41: syntax error near unexpected token `newline' [01:52] /var/lib/juju/units/postgresql-0/charm/hooks/db-relation-joined: line 41: ` echo "host $(get_database_name) ${user} ${remote_host}/32 md5" >> ' [01:55] hazmat: ^^ [01:56] ugh.. copy n paste error [01:57] mhall119, fixed in new rev [02:35] 2012-05-23 22:34:46,010 unit:postgresql/0: hook.output ERROR: /var/lib/juju/units/postgresql-0/charm/hooks/db-relation-joined: line 57: syntax error: unexpected end of file [02:35] hazmat: ^^ [02:50] Need to find me lucky charms. Also, has anyone found a "user freindly" way of putting config file on the server with juju? [02:51] Or must they be packaged with the content repo and moved around with the charm. [02:55] ohh ya, charm name --> golden-crisp [03:48] i wish we could use juju 4 trystack [04:21] koolhead17: why can' we? [04:21] koolhead17: no S3? [04:22] SpamapS, trystack account is cool. we can easily provision and show demo up and running [04:23] SpamapS, https://trystack.org/ a testbed for openstack [04:23] koolhead17: I know what trystack is [04:24] koolhead17: why can't we use it? [04:24] SpamapS, so we can use it? waoo [04:24] koolhead17: I don't know! [04:24] koolhead17: you said we can't [04:24] I'm asking why you say hat [04:24] that rather [04:24] koolhead17: if they expose S3 and EC2, we can use them today [04:25] SpamapS, well let me ask the guys maintaining it. [04:25] HP folks i suppose [04:26] koolhead17: I believe there are two regions and they are both somewhat different [04:26] anyway, I'm out [04:26] SpamapS, will keep you updated. laters [04:35] mhall119, there's a new version up if you want to give it another whirl... :-( [04:35] SpamapS, trystack has no storage [04:35] at least not of lastweek [04:36] hazmat, your correct its just compute node [04:41] i'm out as well [07:55] aujuju: Creating volume group in nova-volume Juju charm === almaisan-away is now known as al-maisan === koolhead17 is now known as koolhead17_afk [12:08] I've been trying an openstack deployment using MAAS server and juju charms, I've spent the last 3 days trying to get it to work and there is alway something within openstack that doesn't work [12:09] any one had any success with openstack deployed via juju ? [12:18] jcastro, jamespage, trying to transfer a css file to a wordpress instance as part of a charm, but cant find how to do it. Can you point me to somewhere that gives example/how to? [12:19] I would like not to have to upload the css file to a repo or something, but to bundle it in [12:19] the charm [12:36] mhall119: hey... so the db-relation-changed in pg... remove the netmask from the hba config. The real fix is for the charm to make sure it's always using ip... this has been a pain b/c it's switched back and forth between dns name and ip addr w/ netmask === al-maisan is now known as almaisan-away [12:39] m_3: my problem was that the pg_hba settings weren't being set until *after* my summit charm was trying to run syncdb against it [12:40] mhall119: ah... yeah just move it to joined then [12:40] mhall119: the orig thinking was that you exposed the db _after_ it was sufficiently created by the changed cycles [12:41] mhall119: but understand the need [12:43] m_3: I believe that's what hazmat has done, so you'll just need to merge his changes [12:44] mhall119: sorry pg still needs some tlc... it's been on the list but below the pain-threshold :) [12:44] mhall119: needs pg9 replication too [13:10] victorp: if you include it as part of the charm the entire charm gets copied over [13:11] so you can stick it in like /stuff or something [13:11] and then as part of the charm copy it over to the WP dir [13:16] jcastro: hey, if I'm generating an admin password for Summit, how do I let the deployer know what it is? [13:18] mhall119: http://jujucharms.com/charms/precise/limesurvey [13:18] you can make a default [13:18] and then have a config option to change it [13:18] (see the end of that readme) [13:21] jcastro: hmm, default passwords aren't ideal [13:22] I suppose I should learn config.yaml [13:22] can you not set one and then just make the user set one via a config option? [13:22] I don't know, I don't know how the config options work yet [13:22] I'm currently setting a randomly generated password [13:22] oh dude, config options are the best! [13:22] yeah that's fine [13:22] and then just say "set a password via config" [13:22] let me show you an example [13:23] http://jujucharms.com/charms/precise/solr/config [13:24] actually, using limesurvey: http://jujucharms.com/charms/precise/limesurvey/config [13:24] jcastro, thanks [13:45] SpamapS: ping me when you're around [13:45] SpamapS: I'm only on for half a day today (leaving for the weekend) and I'd like a quick catch up with my bro [14:07] Hi there. PLaying with maas and juju to see how it works. I m wondering... How do you guys manage name resolution ? [14:08] like, i have installed maas-dhcp, i add 2 nodes in maas, i juju bootstrap [14:08] it will use one of those node, and juju status needs to be able to resolve th ename of this node. [14:08] i have to guess its name by looking in /var/log/maas (and finding some stuff that looks like ip) untill i found the right ip of the niode [14:09] and then set it in /etc/hosts before i could juju status. I bet there are more convenient way to do that. no ? [14:16] hmmm, the maas server itself can resolve the name... I guess the idea is to use this one as a dns then ? [14:18] melmoth, you have to point juju to the maas server.. [14:18] hmmm. make sense. but does not makes me happy :) [14:18] the maas server is a kvm vm, juju runs on the hypervisor [14:19] so maas is actually using the machine juju runs on to resovle names itself. [14:19] i wonder if there is a way to have specific dns resolution path for a given user... [14:23] melmoth, so your saying juju on the host can't resolve the vm that maas is running on? [14:23] not exactly. [14:24] when i run juju status, juju try to ssh on one of the box i have addedd in maas [14:24] and this is the box it cannot the name of [14:24] juju asks maas for the host names [14:24] well, it should. [14:24] but it does not here [14:24] if i remove my entry i added in /etc/hosts, juju status fail [14:25] 2012-05-24 16:25:19,238 ERROR Invalid host for SSH forwarding: ssh: Could not resolve hostname node-000023560101.local: Name or service not known [14:25] sounds like your network is setup quite correctly [14:26] er.. isn't [14:26] .local is a zeroconf address [14:26] the actual name is localdomain, i guess this is just a "formating cut" [14:27] hmmm, funny. on the maas server the domain name is localdomain [14:27] but frmo my hypervisor, i need to put only local (as it appear in the error message) in /etc/hosts [14:28] is it a problem if i associate ip with nodes myself directly with cobbler ? [14:29] this way i could put stuff in /etc/hosts and be sure it s the same once and for all. [14:29] reboot the juju zookeeper, mans should fix your problems [14:30] then look in /var/lib/misc/dnsmasq {tab} {tab} [14:30] also don't use .local breaks stuff [14:31] ops replace mans with mdns (stupid auto correct) [14:32] or while logged onto your maas server run [14:33] tcpdump -npi eth0 not port 22 host {node ip that doesn't resolve} [14:33] then reboot the node with dns issues [14:33] watch the tcp dump as it comes up [14:33] should see some traffic on port 5353 [14:33] it s not a node that has dns issue, its the main hypervisor. If i reboot this one, all the world reboot :) [14:34] but i ll try to fiddel with cobbler manuall to be sure each node have always the same ip [14:34] this way, it should works as i want, as long as maas does not fiddle with this itself. [14:35] hmm I run maas and juju zookeeper on virtual nodes [14:36] didn't want to blow away a full hardware nodes for their services [14:36] do you run juju on a vm or the hypervisor itself ? [14:36] vm [14:36] ah, ok.. that would be another option [14:36] intact you can run juju from any node [14:37] but when you boot strap it deploys the zookeeper [14:37] without the zookeeper you can't connect to the environment [14:38] there an't much doco on MAAS [14:38] it's been a big pain for me [14:39] dnsmasq + MAAS + cobbler is alittle buggy [14:39] I've seen a launchpad blueprint to take cobbler out of the MAAS server, maybe things will be better then [14:52] aujuju: Juju error: Server refused to accept client === koolhead17_afk is now known as koolhead17 [15:25] So many questions... [15:25] What is the url for the entire charm repo? Fill in the blank --> charm getall _____ [15:26] SpamapS: btw, i dident get my golden-star for the Ubuntu charm idea :) [15:27] Does the above statment mean that my comments are being considerd/noticed? (Not trying to be crass) [15:28] 3-- A clover, haha [15:41] hmm, "charm getall" used to work didn't it? [15:42] I distinctly remember a command to snag them all at once [15:42] imbrandon: Hah, true.. that was your idea. :) [15:42] jcastro: it "works" [15:42] but its so bloody slow [15:42] I think we may need to maintain a tarball somewhere with all of them so people can wget the whole archive [15:42] 'morning all [15:43] I bet we can maintain that in charm-tools as a download file actually [15:43] SpamapS: how ? [15:43] the last bit [15:43] launchpad has an API for uploading files [15:43] negronjl: I'm going to assign you a card for the graphs measuring review queue performance if that's ok with you? [15:43] ahh ok i got ya, i was thinking actually IN charm-tolols [15:43] tools* [15:44] jcastro: Sure ... I have some questions when you get a chance re: patch pilot [15:44] i see what ya mean though, works kinda like a reverse debian watch ? [15:44] negronjl: yeah, mira we should have a quick charmers G+ if we can [15:44] jcastro: ok ... invite me [15:45] SpamapS: imbrandon: you guys got time for a quick pow wow? [15:45] yup let me login real fast [15:46] been going back and watching all the juju sessions i missed and some i dident as refeshers, good good stuff [15:47] handy, having the important sessions recorded. :) [15:48] yea [15:48] would be cool if we could get all of them even if audio only [15:48] as i just have them in a minimized browser anyhow ignoring the vid [15:49] invites sent [15:49] I feel like I've wonderd here. You all going to a hangout? [15:50] surgemcgee has stolen the blue moon! [15:56] jcastro: yes b.r.t [16:01] jcastro: curses.. something requires a reboot. 2 minutes [16:10] surgemcgee, why do you want to do charm getall? [16:58] SpamapS: https://groups.google.com/forum/?fromgroups#!topic/ansible-project/x_0gDQv4co4 [16:58] SpamapS: thats the email chain i spoke of [16:58] right, Ansible [16:59] and thats the one wheere i was saying "hey whats next?" he asked [16:59] negronjl: ^^ [17:00] jesus, i think i used all my energy for the rest of the week on that call tho, i seriously think i need to go back to the doc's ( already had one apt since uds ) [17:10] I will just grab the ones I need, no biggie. Also, the charms are installing puppet. Shouldn't puppet be installed on the dev client and puppetmaster be used in the install script? [17:11] And is that DNS entry nessessary? [17:16] SpamapS: is charm getall smart enough to update the local cs too ? [17:17] imbrandon: it will run 'mr update' if it sees a .mrconfig in the dir already [17:17] kk [17:17] imbrandon: I think it might also run 'charm update' in that case too [17:17] that would be helpfull [17:17] lol [17:18] incase the maint dident inc the rev [17:18] i guess [17:18] no it just fetches the list again [17:18] ahh [17:18] so any new charms will get checked out [17:18] the whole thing is really crap, just use charm get :) [17:19] kk, yea between that and the series , i dont see a need ( well anytime even semi soon but not gonna say never ) for cs instructions [17:19] like i thought [17:20] SpamapS: crappy but exists > nothing [17:20] SpamapS: e.g. my php in enterprise + world theory :) [18:06] alright, txzookeeper is in Debian NEW queue. [18:06] on to txaws [18:27] I have a question for juju+maas. Can you declare a specific target for a juju deployed service? [18:28] For example, say I have a "monitoring services" charm and I want to deploy to a specific class of machine or a specific machine [18:30] jkyle, it sounds constraints are what juju provides for what you are wanting to do [18:30] cool, I'll search the docs :) [18:31] https://juju.ubuntu.com/docs/constraints.html [18:32] jkyle, https://juju.ubuntu.com/docs/constraints.html#provider-constraints [18:34] jcastro: are we getting rid of this or what? https://docs.google.com/spreadsheet/ccc?key=0AoW1nhI7IMt3dFRvSFdkZmNqQ0t3RjZ2QTR2Z19teWc#gid=0 [18:36] james_w: I see "orchestra" referred to. I thought that MaaS + Juju was the successor to Orchestra. can you use cobbler classes directly? [18:36] jkyle, I'm guessing that's just a leftover artefact [18:37] I think using cobbler management classes would be stellar [18:37] as a sort of "pool" specifier [18:45] jkyle: I don't think you can use cobbler classes with MAAS + Juju yet [18:45] to my knowledge, only maas-name can be used by juju constraints as of now [18:56] sweet, more swag ... email subject of "blitz.io: You just got a tshirt!" is always cool :) [18:56] imbrandon, just shared the zombie comic and the pinup albums- let me know what you think. [18:56] nice , i'll check em out in a few [18:57] i think i have enough cough medicine in me to make it till bedtime :) [18:57] well and actually be productive somewhat [18:59] jcastro, right now Fedora is having its elections. Next week is townhalls. I was hoping to get a sentence or two from you as far as how cool it will be to have Juju in the repos [18:59] This will be easier with some board members on it [19:04] No big hurry tho :) [19:16] MarkDude: he said something about flying home today a few hours ago so he is likely mid-ait [19:16] air* [19:16] Oh cool [19:17] I will email him then [19:17] is it jcastro@ubuntu.com? [19:17] MarkDude: according to LP it's jorge@ubuntu.com [19:18] Cool, that was my 2nd guess [19:18] ty marcoceppi === hspencer is now known as hspencer[afk] [20:35] hazmat: your postgresql changes seem to have done the trick, thanks! [20:36] mhall119, awesome, thanks [20:48] http://news.ycombinator.com/item?id=4020455 [20:48] upvotes please! [20:49] jcastro: AWS has ARM hardware now? [20:49] we simulate [20:49] so when this hits the market [20:49] people can be ready [20:49] it's a simulated Calxeda box [20:50] oh, ok [20:50] there's something almost zen-like about watching my instances install and configure themselves in debug-log [20:50] I know [20:51] can you believe people used to launch instances like this by hand? [20:51] mhall119: on tuesday you should link up with m_3, he's been at a conference but worked with chris on the django/summit stuff [20:51] I am pretty sure you guys could sort out your issues in a few minutes of facetime [20:52] jcastro: hazmat sorted out the last of my issues [20:52] ah, rock and roll then [20:54] mhall119, thanks for your patience on those round trips.. [20:55] jcastro, oh.. that's what its for [20:55] i was wondering why we have amis for that stuff [20:55] hit the ground running when they ship! [20:58] jcastro: check it out: http://bazaar.launchpad.net/~mhall119/summit/charm/revision/2 [20:58] that's all I had to change between the initial charm generated, and what summit needed to work [20:58] and most of that is because Summit does stuff in a funny way [21:00] mhall119, do you need to do the init-summit and pullapps against the real db? [21:01] hazmat: no [21:01] cool [21:01] just checking [21:02] those just setup the local disk env [21:05] oh crap, i just learned some new nginx magic SpamapS , 4x the current ludicris numbers avail now ... with just a bit of creative request queuing upstream [21:05] hrm , time to get this on one of my sites and really test it more than local, only bad part is it requires a patch to nginx [21:05] :( === hspencer[afk] is now known as hspencer [21:46] imbrandon: does nginx have some kind of request de-duper for when you have to fetch from the backend? [21:46] imbrandon: have always thought caches should do that [21:47] but I've never pushed optimization far enough where that would have been enough of a win [21:47] SpamapS: not sure, but this is basicly making all the requests pool in nginx then only handing php 1 at a time [21:47] and man, its MUCH better at queueing than even php-fpm [21:48] imbrandon: interesting [21:48] and to be honest i thought thats how it worked internally to begin with but i guess not [21:50] that and i think i figured out why i always peek arround 500 concurrent connections and could never figure out why [21:51] i was doing a bit of math wrong and assumed 8192 connections * 4 worker processes == win [21:51] but it actually == about 500 concurrent due to some wonky math [21:51] GWB style math :) [21:52] sooooo i may have found a few bits that will make some nice additions to the base config :) [21:52] lol [21:52] Fool my servers you can't get fooled again. [21:52] still need to test it all out tho [22:30] What is the best method of moving templets around? I wan't a simple vhost in sites-available. Can I do this nativly? [22:32] Also, is there a way to package two charms together. The one is more of a generic template for the main one and I need people to know this. [22:33] ugh, looks like txaws is incompatible with twisted 12 ... [22:34] surgemcgee: There are two ways to get that done [22:34] surgemcgee: there's a very experimental 'charm splice' command that I developed, but its a bit crackful as it can't really relate the two charms together (well it can, but it emulates this in a really weird way) [22:35] surgemcgee: the other way is to turn the detailed part into a subordinate charm, and have the generic service charm as the primary.. [22:48] hazmat: ugh, lp:txaws is a mess [22:49] hazmat: no tags.. any idea what revision 0.2.3 was taken from? [22:50] hazmat: I *think* it is r139 [22:55] Ahh.. txaws works fine w/ twisted 12 .. missing files in the tarball [23:10] SpamapS, that looks about right [23:10] SpamapS, agreed its inanely difficult to tell, i had to resort to a diff [23:12] hazmat: I just used log to find commits before/after release [23:13] hazmat: but yeah, diff confirms.. sort of (since some files are not in sdist) [23:42] aaaaannd txaws is now in Debian NEW queue [23:42] * SpamapS moves on to juju [23:42] man this is so much easier when I don't need a debian sponsor. :)