[00:17] nites people [00:18] what does exactly expose do when used with lxc? [00:18] I mean, is there anything I should implement in my charm to support it? === CyberJacob is now known as CyberJacob|Away [00:30] perrito666: it's sort of theother way around [00:30] your charm should call open-port when it wants to expose a port to consumers of the charm [00:32] davecheney: I see, Ill rtfm myself with that info :D tx [00:33] perrito666: the goal, when everything works right, is as the charm author you shouldn't have to worry about the differences between the various providers [00:34] davecheney: I see, I was trying to learn juju trough creating a charm, I found a grey area on that particular feature (expose) [00:35] perrito666: ok, so there are two parts [00:35] one, in the charm, you say open-port or close-port to indicate to juju that the charm will talk on said ports [00:35] on the client side, you call juju expose $SERVICE [00:36] which takes the list of ports indicated by open-port and does firewall magic to make those visible on the public ip [00:36] davecheney: tx, that is very informative === erkules_ is now known as erkules [06:19] hi all === CyberJacob|Away is now known as CyberJacob [09:35] How to install juju manually by this files: [09:35] ubuntu-12.04-server-cloudimg-amd64 [09:35] and [09:35] ubuntu-12.04-server-cloudimg-amd64-root === CyberJacob is now known as CyberJacob|Away [10:02] Oops! How to install juju via cloudimg's? [12:57] so, I use a charm that bundles charm-tools (for reasons currently unknown) [12:58] bloodearnest: wat? [12:59] marcoceppi: I've no idea why [13:00] bloodearnest: is this a charm that you downloaded or one you're writing? [13:01] marcoceppi: it's one some one on my team wrote who has now left, it's an nginx-rate-limiter charm [13:01] bloodearnest: yeahh, there's really no need to embed charm-tools UNLESS they were trying to use the old charm-helpers [13:01] marcoceppi: last commit was 06/2013 [13:01] in which case, they should use the PPA that only builds charm-helper-sh/charm-helper-python [13:02] bloodearnest: well, that's not much help [13:02] right, so I need to re-write the charm, I'm good with that, I just don't have time to do it today, as I have some urgent deployments === dannf` is now known as dannf [13:04] the only reason this is blocking me is I can't deploy the charm as is any more with 1.17, I get an error: [13:04] charm-tools has a number of [13:04] ERROR error uploading charm: cannot read extracted archive: charm "$package" using a duplicated relation name: "relation-name" [13:04] gah [13:04] which I believe is due to some template metadata.yaml files in charm tools [13:05] bloodearnest: you can fix that by embedded a more recent version of charm-tools [13:05] the new thing is that somehow juju is finding the those metadata.yaml files and complainging [13:05] but whatttt [13:05] marcoceppi: indeed [13:05] yeah, juju really shouldn't dive /that/ deep [13:06] marcoceppi: ack, will bump charm-tools an move forward, and after deployment will back track and file a bug [13:06] bloodearnest: ta [13:11] marcoceppi: updating to r318 of lp:charm-tools gives a different error, but same issue of parsing a template metadata.yaml: [13:11] ERROR error uploading charm: charm URL has invalid charm name: "local:precise/$package-1" [13:11] at least, I assume that's the issue, not 100% sure [13:13] bloodearnest: I'm like 98% sure you only need the helpers directory in charm-tools [13:13] but I don't know where to find the charm to verify that [13:14] marcoceppi: it's lp:~ubuntuone-pqm-team/canonical-is-charms/nginx-rate-limiter [13:16] marcoceppi: I will probably rebase on upstream nginx and add the rate limiting as an MP on that [13:16] just wondered if there was a quickfix to get past it [13:16] bloodearnest: very interesting. I don't see charm-tools being used anywhere in the hooks [13:17] marcoceppi: yeah, it's possible I could just remove it? [13:18] marcoceppi: still, it would be good to figure out why it's failing on 1.17.4/trusty and it's been fine on 1.16.x/precise [13:18] bloodearnest: I would just remove it and deploy for now [13:18] oh wait [13:18] bloodearnest: _pythonpath.py shows that it's just using lib/charm-tools/helpers/python [13:19] bloodearnest: delete everything but the helpers dir in charm-tools [13:19] and you should be good to go for the deployment [13:19] marcoceppi: ack will try [13:24] Hi all, does anyone have any pointers (or working config samples) for a reverse proxy setup, so I can expose my charms from my cluster to the 'real' world? My cluster is a private/seperate network from our schoolnetwork (it is being NATted through the MAAS controller) [13:25] o/ zchander [13:25] marcoceppi: is this something that can be achieved with iptables on the controller node as a gateway? [13:27] Hi lazyPower, for what I have tried for now, is a apache2 reverse proxy setup, but that doesn't seem to proxy [13:27] lazyPower: Did the presentation run well? [13:28] zchander: it did. My internet was flakey and dropped while deploying a hadoop cluster [13:28] so i wasn't able to give the full technical rundown with magic without a minor bit of failure [13:28] but my showmanship was pretty high, so it went over pretty well [13:29] brb: going to get some caffein [13:38] back again... [13:40] re [13:40] zchander: when you say the apache reverse proxy doesnt work, how so? is the region controller not reaching the internal instances? or... [13:42] It seems so.... I have just reinstalled my MAAS environment from scratch this morning. === marcoceppi changed the topic of #juju to: Welcome!! Docs: http://juju.ubuntu.com/docs || FAQ: http://goo.gl/MsNu4I || Review Queue: http://goo.gl/9yBZuv || Unanswered Questions: http://goo.gl/dNj8CP || Reviewer: lazyPower [13:54] lazyPower: Is seems the 'reverse proxy' redirects all traffic to a https site, which cannot be found, as no ssl enabled site exists (yet) [13:54] And the log mentions POST requests to my MAAS api [13:59] Currently I get a 502 - Bad Gateway from nginx [14:00] And additionally I cannot connect to my MAAS web interface anymore from my school network === bloodearnest_ is now known as bloodearnest [14:02] hazmat, got a minute to talk deployer? [14:05] Please disregard my last issue: I forgot (with my fat fingers) to include the ServerName directive in my VirtualHost [14:08] zchander: np :) happens to the best of us [14:13] ERROR state/api: websocket.Dial wss://10.0.3.1:17070/: dial tcp 10.0.3.1:17070: connection refused [14:13] I am getting this when trying to teardown the local provider [14:14] jcastro: i typically get that when my local provider has gone awry, such as an interrupted bootstrap [14:14] yeah [14:14] so the "normal" procedures should work? [14:14] just removing .jenv if there are no lxc containers usually resolves that. [14:14] to clean up [14:14] yeah [14:14] if there are leftover lxc containers, follow the AU post [14:16] jcastro: i'm working through porting my writeup on osx/vagrant charm authoring to the docs atm. What subsectionw ould you put this under? [14:16] authors-osx-walkthrough? [14:16] in amulet, can Deployment.setup() be called multiple times in the same test? [14:17] I would do "How To" and then "Vagrant work flow" [14:17] cargill: thats part of post deployment commands. I filed a bug about the behavior you will see if you try it :) [14:17] cargill: it can, but you shouldn't add any relations between setup calls, it's not supported [14:17] jcastro: ack. ty [14:18] cargill: not supported yet* [14:18] cargill: what's your ultimate goal? [14:20] I'm getting 'service "postgresql-sentry" not found' when I deploy my charm and potgresql, and not when I deploy just potgresql, so I'm trying to find out whether it's a timeout from my charm setting up itself (it does take a long time) or something else [14:21] cargill: what's your test file look like? [14:21] and setting up one, then the other does indeed trigger an error suggesting that setup cannot be run multiple times [14:23] marcoceppi: http://pastebin.ubuntu.com/7067852/ [14:23] this is without any "debugging" additions [14:24] cargill: interesting, and so what's the error you're getting? Also, do you have the latest amulet? [14:24] as in, is there a Python traceback? [14:25] jcastro, shoot [14:25] can I make bundles with colocated services in LXC containers? [14:25] lazyPower: can I send you my virtual host config (for a wordpress charm, as per tutorial) [14:25] jcastro, yes [14:25] let's say I want rails and postgres, but I want all the monitoring stuff in one instance on LXC containers [14:25] zchander: sure, pastebin it [14:25] jcastro, http://pythonhosted.org/juju-deployer/config.html#placement [14:26] jcastro, placement is somewhat limited at the moment, its a form of deploy x with service y.. and supports containers [14:27] ok so lxc:wordpress is "whichever machine wordpress ended up on" then? [14:27] jcastro, i've had some discussions last week about allowing machine specification (ie a machines block in addition to a services block) and then allowing for machines to be placed there. [14:27] er.. i mean workloads to be placed there [14:27] lazyPower: http://paste.ubuntu.com/7067872/ [14:27] jcastro, yes [14:28] marcoceppi: yes, the vagrant scripts from lazyPower pick up the latest amulet [14:28] jcastro, i avoided machine spec (minus 0) in the initial pass because its quite fragile unless you can abstract machines a bit from their env specific identity. [14:28] hazmat, perfect, that's exactly what I needed, thanks [14:28] will post the output shortly [14:28] well for this bundle I don't care about machines [14:28] lazyPower: I might be back tonight again, going into a meeting in a few minutes [14:28] I'm making a "rails" one for development and I don't want a ton of instances [14:28] cargill: offtopic from what you're talkign about, but I did a writeup using the precise basebox. it'll reduce that overall footprint of the box to 200mb in your vagrant cache. [14:29] http://blog.dasroot.net/writing-juju-charms-on-osx/ [14:29] jcastro, cool, for most of them you shouldn't.. the call for machines in deployment is more about real world ops scenarios.. where you want to prep a machine a bit before hand [14:29] jcastro, ie.. service co-location is ideal for bundles [14:29] use that vagrantfile with the precisebasebox and you should be g2g. I've tested it locally with no issues. [14:29] that re shared/reused. [14:30] zchander: i dont see anything apparently wrong with this [14:30] zchander: ping me when you get back and we'll do some troubleshooting [14:30] lazyPower: oh, great, right now I'm just running provision everytime, that won't be faster though, will it? but speeding up the bootstrap is definitely a good thing [14:30] cargill: it may slow it down since its cloud-init bootstrapping a juju gui [14:31] the box you're using right now was specifically tailored for amulet testing [14:31] hazmat, yeah I didn't want to be like "here's your rails dev bundle, have fun paying for 9 instances an hour!" [14:32] jcastro, approximately $1 on do ;-) .. if you have a chance i'd love your feedback on that plugin. [14:35] jcastro, not sure you can do better than hulk-smash in public clouds atm, vpc/private networking against cloud apis for containers is still wip. [14:40] marcoceppi: this is the log with the test written as above: http://pastebin.ubuntu.com/7067958/ [14:49] cargill: interesting === wedgwood is now known as Guest24594 [15:00] marcoceppi: this another of the errors I'm getting http://pastebin.ubuntu.com/7068075/ [15:00] cargill: what's weird, is the setup is being killed beacuse of timeout, but it keeps going aftwards [15:00] which is probably why you're seeing that error [15:05] that's because the timeout passed to setup() is 900 and there is an internal one of 600 seconds? [15:09] http://askubuntu.com/questions/432187/how-can-i-deploy-my-local-juju-charm-with-amulet-framework [15:09] just in case anyone missed this [15:14] cargill: it should respect your timeout, however, if you're doing `charm test` command there's timeout per test [15:14] cargill: actually, that's probably where it's coming from [15:14] cargill: you need to modify whatever is running `charm test` and add `charm test --timeout=10m` or something like that [15:14] lazyPower: ^^ [15:15] marcoceppi: can you open a bug on the project? I'll get that update pushed in a minute [15:15] wrapping up my doc edits [15:15] lazyPower: which project/ [15:15] https://github.com/chuckbutler/juju-vagrant [15:16] marcoceppi: will try with 30minute timeout, let's see where it gets me next [15:16] cargill: ack [15:28] https://github.com/juju/docs/pull/27 -- for anyone thats interested in reviewing the proposed merge for Vagrant based workflow on OSX [15:28] also sent an announcement to the list [15:49] hey lazyPower [15:49] wrt. your +1 on the hadoop bundle [15:49] I made a section on how to put them in the store [15:49] https://juju.ubuntu.com/docs/reference-reviewers.html [15:50] oh nm, you're not in ~charmers [16:08] lazyPower: see gh for feedback on pull [16:09] marcoceppi: fixing, addendum to that is we need to do that for the rails howto then, thats what i modeled it after. [16:09] lazyPower: that's fine, we can fix that later [16:09] ack [16:17] https://github.com/juju/docs/pull/28 [16:17] if someone could ack that it's a simple fix. [16:20] marcoceppi: wiped $'s [16:21] lazyPower: re-reviewing === sputnik1_ is now known as sputnik13net [16:30] lazyPower: updated [16:34] re-updated [16:34] marcoceppi: ^ [16:38] lazyPower: okay, finally reviewed everything, only two little things [16:38] I won't be so lazy reviewing next time [16:42] marcoceppi: ack. Fixed and pushed. [16:43] marcoceppi, does bundle proof check that juju-gui is not included in a bundle? [16:43] jcastro: no, but it could/should [16:43] filing [16:43] open a bug and I'll add it to the next release [16:44] lazyPower: hahah, I found one more thing [16:44] you were a little aggressive with the $ removal [16:44] KILL IT WITH FIRE! [16:45] marcoceppi: line 240? [16:45] lazyPower: nooooope [16:45] ;) [16:45] find and replace came back to haunt me :(( [16:46] not sed -i? :) [16:46] sarnold: yuck it up fuzzball :P [16:47] nyuk nyuk nyuk [16:47] <3 [16:47] <3 [16:47] lazyPower: commented [16:48] lazyPower: or just ommit that all togehter [16:48] but, that cuts out my interesting prompt/path [16:48] ack - removing. [16:54] * lazyPower goes back to the review queue [17:07] hey everyone, "State of the Juju Ecosystem" had to be moved to thursday as they added a plenary slot [17:08] so Thursday will be charm school, state of the eco, and then Amulet tests all in one day [17:09] womp [17:09] ty for the update jcastro [17:10] did we move that around on the eco calendar yet? [17:11] I didn't know we were putting those on the calendar [17:11] in that case, no [17:11] maybe we arent. I just looked and didn't see anything there [17:11] i figured with it being UDS related, it would go on the calendar, my b. [17:11] it's UDS, we're supposed to be paying attention heh [17:12] but if you register yourself at UDS it will generate a calendar of sessions for you, etc. [17:12] I'll be sending reminders here throughout the week anyway [17:21] lazyPower: noooo [17:21] lazyPower: you only needed to fix that one line, the rest was fine :( [17:22] marcoceppi: no idea what you're talking about [17:22] gh [17:22] damnit [17:22] i'm quit [17:22] s/i'm/i [17:24] You even asked, and I was like "nooo, those are fine" [17:25] marcoceppi: updated [17:26] lazyPower: now, can you squash 'em all :) <3 === Ursinha is now known as Ursinha-afk [17:31] hey lazyPower [17:31] now that you've documented the vagrant workflow [17:31] http://www.vagrantup.com/blog/vagrant-1-5-and-vagrant-cloud.html#features [17:31] hah. [17:31] jcastro: make a card [17:32] and stand in line behind marco [17:33] I think it'll be a bit before we get 1.5 [17:33] utlemming, https://vagrantcloud.com/ [17:33] utlemming, this is just launched, but we should perhaps look at indexing the official boxes there? [17:34] jcastro: https://vagrantcloud.com/search?utf8=%E2%9C%93&q=juju [17:34] boo [17:34] indeed! [17:34] look at all these chef boxes [17:35] boy sure would be nice if JUJU was a featured box [17:35] * lazyPower tongues cheek === Ursinha-afk is now known as Ursinha [17:55] ping lazyPower [17:55] zchander: pong [17:56] I am at home right now, but with VPN to school [17:57] zchander: ok. my thoughts are check connectivity from the gateway node to the internal net. validate its communicating properly [17:57] start at the lowest common denominator and unwind the onion of complexity layer by layer [17:57] I can ssh into the node without any problem using the hostname from the controller [17:58] ok, is port80 of the node accessible from the controller? [17:58] you can use links or similar text based browser to validate [17:58] or curl for that matter [17:59] I am installing lynx right now [18:01] Done: I am getting a bad gateway from nginx [18:02] that would be on the wordpress charm, and you should have gotten that via the apache reverse proxy. but ok we have validated you have port 80 connectivity [18:02] tail your apache logs for the vhost, and attempt to reach the wordpress node. [18:07] snippet from wordpress-access_log: http://paste.ubuntu.com/7069014/ [18:08] btw, maas-ctrl-001.nimeto.edu/MAAS also gives a bad gateway from nginx [18:08] so the reverse proxy is working? [18:08] (this is the fqdn for my MAAS controller) [18:09] Seems so, but too good? ;) [18:09] so, my thoughts are, poke at wordpress and figure out a) why its a bad gateway (probably PHP-FPM failed to start) b) what do you need to add to get the proper hostname on the wordpress unit to respond to the request. [18:12] This a line from the access log from nginx on the wordpress node: 10.0.41.5 - - [10/Mar/2014:18:06:05 +0000] "GET /favicon.ico HTTP/1.1" 404 140 "http://node001.nimeto.edu/" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_9_2) AppleWebKit/537.74.9 (KHTML, like Gecko) Version/7.0.2 Safari/537.74.9" [18:12] i saw that. its not finding a site for that hostname. [18:12] the 404 in the log detail points to that. [18:13] Seems I have to get up to speed with nginx.... [18:13] zchander: http://wiki.nginx.org/ServerBlockExample [18:13] zchander: there's also a configuration option for the wordpress charm to use apache instead of nginx [18:14] zchander: juju set wordpress engine=apache2 [18:14] https://jujucharms.com/precise/wordpress/#readme [18:14] Can I foo that with a running charm? [18:14] yep. so long as the charm is not in an error state, it will reconfigure itself on the fly [18:16] just did it.... Trying again [18:19] getting [agent-state-info: 'hook failed: "config-changed"'], this will resolve itself in time? Or should I 'resolve' the issue? [18:20] brb: getting kid to bed.... [18:22] zchander: hook failures need manual resolution [18:22] zchander: you can either pass the -r to the resolved command, or use debug-hooks to manually investigate and resolve the issue. [18:22] zchander: units in an error state do not magically resolve themselves, and always require manual intervention. [18:33] getting a HOOK W: Failed to fetch http://ppa.launchpad.net/charmers/charm-helpers/ubuntu/dists/precise/main/binary-amd64/Packages 403 Forbidden when switching to apache (??) [18:35] I am going to destroy the service and retry to redeploy [18:43] bbl, got visite... parents... [19:03] hi sinzui [19:03] hi bac [19:07] anyone playing with juju on Digital Ocean? === CyberJacob|Away is now known as CyberJacob [19:28] geekmush: there are a few using hazmat's plugin for alpha-level testing [19:28] geekmush, just saw your issue [19:28] geekmush, so you need the dop library from git atm.. [19:28] geekmush, i'll switch out to just doing a mini digital ocean client lib for the needed bits within the plugin. [19:29] but probably not today [19:29] hazmat: ok, cool … I'll stand by and also try out the git'd version of dop as well [19:30] geekmush, cool, yeah.. things work if you using the git version of dop.. i put a bug in there for them to make a release, since its been over a release since their last [19:30] er.. over a year === lazyPower changed the topic of #juju to: Welcome!! Docs: http://juju.ubuntu.com/docs || FAQ: http://goo.gl/MsNu4I || Review Queue: http://goo.gl/9yBZuv || Unanswered Questions: http://goo.gl/dNj8CP [20:03] hey fellas, so I am working on a django bundle [20:03] but the charm hasn't been touched in like a year [20:03] anyone mind doing it next on the audit? [20:31] geekmush, any luck/progress? [20:32] just started … got pulled into last minute conf call … [22:50] does anyone know the status of juju on debian? [22:51] i'm seeing some articles linking to a juju package in experimental, but it's gone now [23:01] bodie_: I don't think there has been any effort to push the juju packages to debian yet; those packages in experimental were probably the old python version 0.5 or so iirc, which, while neat, weren't anything that anyone wanted to tell users to use :) [23:05] bodie_: juju the client should just work [23:05] bodie_: though you may need to compile it from source, [23:06] I see :) [23:06] thanks