=== CyberJacob is now known as CyberJacob|Away [00:14] marcoceppi: hi [00:18] i have a apache2 mod now i deploy apache2 now how i install my mod in it, do i make a charm for my mod and run this command "juju deploy --to=1 myapache_mod_charm" is this the only way to do this? [00:31] themonk: do you have to compile the mod specially? I think there's a config option for apache mods on the charm [00:33] marcoceppi: this is a custome mod :) and it has a conf and load file for mods-available and i made a binary mod.so from src [00:33] themonk: I'd recommend making your apache_mod_charm a subordinate charm then [00:33] that way you can deploy it without having to use the --to command [00:36] marcoceppi: hmm i was reading subordinate charm doc its not clear to me how i will do it. is there any charm like this. a subordinate charm to apache2 charm [00:37] there's no subordinate charm like yours for the apache2 charm, but there are quite a few subordinate charms === Tribaal_ is now known as Tribaal === _mup__ is now known as _mup_ === yo61_ is now known as yo61 === mwhudson is now known as 5EXAAR2RG [02:08] marcoceppi: hi, i am back. :) can you give me a charm name? [02:09] marcoceppi: i was looking Gunicorn [02:09] marcoceppi: it is a subordinate charm. === 5EXAAR2RG is now known as mwhudson_ === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha === FourDollars_ is now known as FourDollars [04:38] i cant understand add-relation. say juju add-relation apache2 mycharm and juju add-relation mycharm apache2, is there any difference? === Kyle____ is now known as Eliz === thumper is now known as thumper-afk === jam1 is now known as jam === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha === thumper-afk is now known as thumper === zchander_ is now known as zchander === Lord_Set2 is now known as Lord_Set [11:10] ping frankban [11:12] zchander: morning [11:18] Good morning frankban. I have recreated my MAAS environment this morning, using 13.10 [11:18] Going to try that [11:18] zchander: sounds good [11:19] I am at the point to import the PXE files. Will start juju when that's done [11:20] Do you have any recommendations regarding the PXE system? precise, quantal, raring, saucy? [11:21] zchander: I don't know maas enough to give recommendation. the only constraint from the GUI perspective is that the GUI node must be precise [11:21] Ah... Going to use that [11:31] frankban: I am going to test the 'reverse proxy route' using 'subfolders' instead of virtualhosts [11:31] zchander: makes sense [12:57] Yep, got to the point I am going to make the apache2 config for juju-gui [13:34] I'm trying to setup my first juju system with the wordpress+mysql charms example, but it gets stuck in a situation where machines "1" and "2" have instance-id: pending [13:34] I have already disabled ufw [13:34] accorading to mailing list this my be the result of template to be downloaded, but is there a way where the download is active and how far the progress is? [13:37] phedny: what version of juju are you running? [13:37] phedny: if your instance id's are pending, the hosts are not spinning up. I assume you're using the local provider? [13:37] agent-version: 1.16.6.1 [13:37] yes, I'm using the local provider [13:38] do you see the machines listed in sudo lxc-ls? [13:38] I only get "mark-local-machine-1" [13:38] did you update juju after you bootstrapped your local environment? [13:39] no, I basically followed https://juju.ubuntu.com/docs/config-LXC.html on a clean 12.04 LTS [13:39] hmm... [13:40] If you've waited a reasonable amount of time (5 minutes should be plenty for getting 2 lxc containers spun up) can you tryd estroyiing the environment and restarting? [13:40] I waited over 30 minutes ;) [13:40] yikes [13:40] let me try to destroy and start again [13:40] yeah, lxc containers are pretty quick, and getting quicker in one of the revisions coming down the pipeline [13:41] phedny: let me know if it continues to hang and give you problems. We'll need to aggregate some facts about the environment. [13:43] okay.. I entered all commands again, in machine-id: pending state, so let's give it some time again [13:44] frankban: When trying the 'non-virtualhost' approach, I get 404 errors. see: http://paste.ubuntu.com/7079340/ [13:45] This is my config for now: http://paste.ubuntu.com/7079344/ [13:45] phedny: can you tail $HOME/.juju/local/logs/machine-0.log and look for errors? [13:46] the last time I ran into this, juju was on the same network segment and i had ip address collisions. [13:49] lazyPower: I get only there error messages: [13:49] 2014-03-12 13:42:13 ERROR juju apiclient.go:116 state/api: websocket.Dial wss://localhost:17070/: dial tcp 127.0.0.1:17070: connection refused [13:49] 2014-03-12 13:42:13 ERROR juju runner.go:211 worker: exited "api": websocket.Dial wss://localhost:17070/: dial tcp 127.0.0.1:17070: connection refused [13:49] the remainder are mostly DEBUG level with some json and I saw some PGP blocks [13:49] but it bootstrapped and returned without error? [13:50] bootstrap didn't return any errors [13:50] * lazyPower if baffled [13:50] s/if/is [13:50] why would your state server be denying requests if it bootstrapped ok? [13:51] netstat shows me a listening socket on [::]:17070 === edu-afk_ is now known as edamato [14:06] phedny: i dont have a good resolution for you. My thought is to open a bug against juju-core with your log output. [14:16] for subordinate charm how do i target apache2 only, if i do not want to fork apache2 ? [14:17] my perpous is to install a custome apache module [14:17] themonk: you don't really, you just tell people in the README to only add it to apache2 [14:20] marcoceppi: i am depending on juju-info here ryt and every charm implicitly has it, so apache2 and say nginx provides interface http and if i have both of them then my subordinate charm will triger both of them ryt? [14:21] themonk: juju-info is iplicit, yes [14:21] If you wanted to make it ONLY available for apache2 you'd have to add a new interface to the apache2 charm [14:21] but there's nothing wrong with it just being a subordinate [14:21] just prefix the charm name with apache2 so people know it's an apache2 only subordinate [14:21] apache2_mod_yourmod [14:30] marcoceppi: i am just thinking deploy --to is much easier :) an then i can use add-unit to replicate but will it trigger if i do config change in my modcharm? [14:31] themonk: if you add-unit to either your charm or apache neither will scale with each other [14:32] --to is not a very good solution [14:32] marcoceppi: hmm i understand :) [14:40] marcoceppi: i want to show you my metadata [14:41] themonk: oka [14:45] marcoceppi: http://pastebin.com/GnHGT36w [14:50] marcoceppi: is it ok ? [15:52] marcoceppi: hello are you there? === niemeyer_ is now known as niemeyer [16:33] themonk: the paste link didnt' work [16:34] marcoceppi: it was set for 1h [16:34] themonk: ah, sorry, I was afk for a bit [16:34] marcoceppi: ok doing again [16:34] marcoceppi: no problem :) [16:36] marcoceppi: http://pastebin.com/xQxs0FGP [16:36] themonk: well, that's one way to do it [16:37] but that implies that it'll work on haproxy, varnish, etc [16:38] marcoceppi: yes yes how i fix it [16:38] themonk: just use juju-info [16:43] juju-info: [16:43] interface: juju-info [16:43] scope: container [16:44] marcoceppi: you mean remove website: interface : http and use it https://gist.github.com/anonymous/c08c66103454c2bfcacd [16:44] themonk: pretty much, though call the relation something like apache2-mod instead of juju-info. but just use teh juju-info interface [16:45] marcoceppi: hmm ok [17:30] marcoceppi, did my bundles not get pushed yet? [17:32] jcastro: npt uet [17:42] marcoceppi, is there a way you could partially pull a PR? [17:42] like say, the manual provisioning page only and ignore the other parts of my PR? [17:42] I'd like to get the mention of null out of the docs asap [17:42] or, is there a trick for me to resubmit only X changes? [17:43] jcastro: it depends [17:43] jcastro: were they done in two commits? [17:43] yes [17:44] jcastro: then yes [17:44] jcastro: I'll cherry pick your commit after lunch [17:45] oh man, awesome [17:46] jut got a mail from a guy using Juju to deploy a wordpress blog, he's migrating from this: http://lquest.org/ [17:46] jcastro: sweet, let him know he can ping me with any questions [17:47] seems like he's deploying to LXC, but wants to access the blog from the outside [17:48] http://askubuntu.com/a/282415 [17:48] is this still our best bet? [17:49] jcastro: you def need to set up a bridge [17:50] but the dhcp settings will change depending on provider [17:54] yeah I recommended manual provider instead [17:54] before realizing it still has the null junk in it [18:09] jcastro: https://github.com/juju/docs/pull/30 [18:10] jcastro: I also created a video of how I cherry picked the commit, it's uploading to youtube atm [18:12] brilliant [18:13] oh I see, you resubmitted it. [18:13] there's no sound, and it's actually pretty low qual [18:13] I suppose it's bad form to approve your PR of my PR? [18:13] jcastro: yeah, it still needs to be approved, I just split it out [18:14] jcastro: I mean, it lgtm, but get like lazyPower or someone to ack it [18:14] jcastro: https://www.youtube.com/watch?v=1VWsmr_HdFU [18:15] I think by low sound you mean no sound [18:15] heh [18:15] I should probably make a better video [18:15] where I narrate it [18:16] jcastro: approved [18:17] since I have your attention(s) [18:17] https://code.launchpad.net/~charmers/juju-core/github-docs [18:18] is the place where you can click the button to import the github branch into launchpad [18:18] you'll want to hit the button in cases where you want the latest changes from GH to run with the next docs regen cron job [18:18] jcastro: cool, also, since this was a cherry pick, if you look at your other PR, the one that had null provider and quickstart, you'll see it only shows the quickstart stuff now [18:19] nifty [18:20] does the button only show up when theres open differences in the branches? I see no such button. [18:20] oh i'm not logged in, hang on [18:20] you'll see it when you're in ~charmers [18:21] ah i see it [18:21] oh right! i need to craft my letter to petition the group for a charmer tag [18:21] so much to do, so little time in the day [18:28] how does Juju deal with outages? [18:33] bodie_: depends on the kind of outage [18:34] maybe I'm misunderstanding it a bit as a state map where it's more a system for choreographing a service network and then deploying it [18:34] so I guess my real question is whether juju is more about a state service which knows the state of its nodes, vs. more oriented to choreography or planning [18:35] provider-level outage, in other words [18:35] let's say a node where 3 of our services reside goes dark [18:37] sure, each agents ping the bootstrap node as a healt check [18:37] so juju knows when agents go missin [18:37] lets say there's a network blip, and your 3 nodes go offline for 10 mins, juju does nothing. It'll queue events and when the agents re-register (and come back online) it'll push the events out again [18:38] bodie_: using the API you can add additional logic to juju, say to replace those nodes, but juju doesn't do anything automatically at the moment [18:39] I see [18:39] marcoceppi, expect a ping from mattgriffin re percona-server charms/support in mysql charm [18:39] but it'll at least realize that something's not normal [18:39] we just discussed in a session [18:39] jamespage: ack, thanks [18:39] bodie_: right, juju status will reflect an outage in the machine listing and the unit [18:40] saying agent-state: down or missing [18:40] righto [18:40] so it actively knows about the topology it's deployed and to some extent it's health [18:40] okay, cool [18:40] so... is there someplace I can find the program flow for what happens when a new charm is deployed? [18:40] I mean, obviously the code [18:40] but I'm still in overview mode [18:41] if not, i'll just get to mapping that [18:41] bodie_: the docs and then asking questions. The docs are still pretty fluffy and don't dive too deep at the moment [18:41] there's also a few blog posts floating around [18:41] (noticed that) lol [18:41] alright :) [18:41] bodie_: let me find a blog post from gustavo that might help [18:41] according to nate the server will bring up a replacement unit if it sees an outage? [18:42] awesome, thanks [18:42] bodie_: http://blog.labix.org/2013/06/25/the-heart-of-juju [18:42] that might have a bit more information than the docs [18:43] bodie_: otherwise, if need more indepth stuff feel free to ping in here [18:44] :D thanks [18:45] right now a stranger in a strange land attempting to worm my way into engaging with core code === edamato is now known as edu-afk [19:09] lazyPower: Reinstalled my MAAS, using 13.10 for the controller. Went the reverse proxy way without virtual host, but that also won't work. (juju-gui now gives me a weird webpage). The side of the story is, that I can see the requests for juju-gui on de node ;) [19:12] ...wat [19:12] zchander: what is de node? [19:13] de == the (sorry, was talking to my wife and instantly wrote some Dutch words.... ;) ) [19:13] oh ok. i was confused about intent [19:14] What I meant: I do see the request coming on the juju-gui node in guiserver.log [19:34] * zchander banging head against the wall.... [19:35] zchander: whats this weird webpage you speak of? [19:37] lazyPower: I'll try to upload it somewhere.... (have to rescue my MAAS interface at the moment, was a bit too fast with apachectl :( ) [19:38] lazyPower: http://postimg.org/image/axpl2ii2r/ [19:39] lazyPower: This is the javascript console: http://postimg.org/image/e4qr3wfe1/ [19:39] that was my next request :) [19:40] Pfew.... got my MAAS web interface back [19:40] ok so its missing the assets - or so it thinks. [19:41] zchander: juju-gui doen't expect to run in a sub-directory [19:41] so if you look at your javascript console - its not loading anything as the juju-gui assumes everything is going to be loaded off of a root path of /juju-ui [19:41] so it uses absolute URLs instead of relative [19:41] rick_h_: ^^ [19:42] zchander: what I'd recommend doing, is instaed of using sub dirs to route requests in the reverseproxy, use subdomains [19:42] lazyPower: marcoceppi yea, don't think things work in non root path atm. It's something we've got to fix coming up [19:42] but not a supported way to service atm [19:42] zchander: ie, juju-gui.mass-ctrl-001.nimeto.edu [19:43] rick_h_: ack, thanks for the heads up [19:43] zchander: then have maas-ctrl-001 reverseproxy route based on host name and not directory [19:43] if that's possible [19:43] zchander: it does look like you're close though! [19:43] marcoceppi: I was doing something like that, and using reverse proxy, but that gave me some other issues (with web sockets) [19:44] oh, damn, really? [19:44] it was what I discovered yesterday [19:44] ah, I'm not sure how to get around that [19:44] But I would like to retry (as I now use 13.10 instead of 12.04.4) [19:44] rick_h_: any instructions on how to run juju-gui from behind a proxy? [19:45] * lazyPower afks [19:46] marcoceppi: Also using the reverse proxy borked my PXE environment, so when I did restart my node(s) yesterday, they didm;t get back up until this morning as I reverted my config to only a default apache environment on my MAAS [19:46] :D [19:48] Maybe I should install another node/server with two nics, one connected to the 'cluster lan', the other to our schoolnetwork. [19:48] And run the proxy on that node === arosales_ is now known as arosales === zchander is now known as zchander_ === mmcc` is now known as mmcc [23:08] hey folks, I have a quick question - where should I file a bug against the juju.ubuntu.com website? It's just a simple broken link... === cjohnston_ is now known as cjohnston [23:10] mmcc: I believe this is the spot: https://bugs.launchpad.net/juju-website/ [23:11] sarnold: great, thanks! I was looking for it among the projects here https://launchpad.net/juju-project - no dice :) [23:17] hey, I think I have a bug, although it might be a feature: [23:17] I am using juju 1.16.3-saucy-amd64 [23:18] and I created a charm which uses pgsql relation [23:19] when the hook for the relation change is triggered, it is triggered twice, once of these times relation-get returns no information on the bash script of the hook [23:19] but, if hooked with debug-hooks to the event, in both oportunities relation-get works correctly [23:20] marcoceppi: ping, you'll be running that charm school at uds tomorrow? [23:20] jcastro: ^ ? [23:36] jose: I'll be there, but jcastro is probably your POC [23:36] perrito666: that's not a bug [23:37] I wanted to see if it could be ran at ubuntuonair, since we have part of the story of the charm schools, to get it completed [23:37] jose: i dont see why not [23:37] perrito666: all hooks need to be idempotent [23:38] mmcc: which link is broken? [23:38] marcoceppi: I know, what I think is a "bug" is that the env in debug-hooks is different from the one the hook is run [23:38] perrito666: in what way? [23:39] marcoceppi: it's the "Overview" link in the footer, I filed this bug: https://bugs.launchpad.net/juju-website/+bug/1291674 [23:39] <_mup_> Bug #1291674: broken "Overview" link in footer of https://juju.ubuntu.com/resources/ [23:39] mmcc: ah! thanks for filing that, I'll make sure it gets updated [23:39] marcoceppi: this is my hook (for context) https://github.com/perrito666/mezzanine_charm/blob/master/precise/mezzanine/hooks/pgsql-relation-changed [23:39] marcoceppi: sure, glad it got to the right person :) [23:39] marcoceppi: now, this is run twice in all my tests [23:40] marcoceppi: one of those, the first, relation-get returns empty values [23:40] but [23:40] if I hook to said event, I can not get said behavior [23:40] perrito666: you're just running in to a race condition [23:40] relation-get returns results always [23:40] marcoceppi: that was my tought [23:41] marcoceppi: I tried a sleep of up to 5 mins to discard that [23:41] (in the hook) [23:41] perrito666: it's still not a bug, it's just that when the first run of relation-changed on mezzanine happens, pgsql hasn't sent it's data yet, so the first relation-changed (which juju guarentess will fire at least once in a relation) goes off [23:41] perrito666: but data available to the hook doesn't change while the hook is running [23:42] marcoceppi: I see :) thank you for your help [23:42] perrito666: put the sleep in your relation-joined [23:42] say, just 45 seconds [23:42] by then psql should be ready, and you should see the first run of relation-changed get fresh data [23:42] marcoceppi: Ill do, just for educational purposes, the charm actually works, I was just curious of this behavior [23:43] perrito666: right, I understand the curiosity, so to guarentee that multiple calls to relation-get work, when the hook is executed it freezes what's available in the relation data wire [23:43] that way relation-get is consistent for the time of that hook, if relation data changes while the hook is running juju just queues another relation-changed hook fire in the event queue [23:44] marcoceppi: makes a lot of sense, thank you again for taking the time to explain :) [23:44] perrito666: no problem! it occurs to me this would be great to put in the docs [23:44] any other questions while I'm still around? [23:44] marcoceppi: nope, that is it, thanks again [23:44] perrito666: np!