[02:42] is there much practical difference between the local provider and using the manual provider and using --to lxc:0 all the time? [02:58] mwhudson: well, manual is for boxes that are not your PC, so you can do what you do with local in an external server :) [02:58] using Juju and isolating each service into an LXC container [03:21] mwhudson: there's not a significant difference really. in the local provider you can do "add-machine" and it creates lxc/kvm containers under the covers, but that's about it as far as differences go I think. [03:22] ok [03:23] * mwhudson is going to try to smash openstack onto a single machine because that's all the arm64 nodes i have to play with right now... [03:38] haha uh, is juju-deployer and the local provider a bad combination? === thumper is now known as thumper-cooking === CyberJacob|Away is now known as CyberJacob === vladk|offline is now known as vladk === CyberJacob is now known as CyberJacob|Away === isviridov|away is now known as isviridov [08:47] jacekn, https://code.launchpad.net/~jacekn/charms/precise/swift-storage/n-e-m-with-concat/+merge/221692 appears to have more than just the nrpe support - lots of changes around how rsync is managed as well? [08:53] jamespage: yes it enabled rsyncd functionality, otherwise swift will wipe rsyncd.conf [08:53] jamespage: so in other words swift charm was not subordinate friendly [08:53] jacekn, ah - so the nrpe sub uses rsync as well? [08:54] jamespage: yes, *external-master part means nagios is not part of the environment [08:54] so it needs to grab configs somehow [08:55] if/when we get cross environment relations we can rethink this approach === Ursinha is now known as Ursinha-afk === vds` is now known as vds === thumper-cooking is now known as thumper [11:46] Does juju supports amazon subnet (vpc) ? [11:48] Not yet. There's an open ticket to work with vpc [11:49] You can use the manual provider to interface with aws vpc deployments in the interim... But the aws provider as of today doesn't support it. === vladk is now known as vladk|offline === Ursinha-afk is now known as Ursinha [14:37] Can anyone tell me if Juju has a restful API? [14:38] marcoceppi, I need a link to your 2nd troubleshooting charm school [14:38] automatemecolema, yeah, looking for the spec, one sec [14:39] jcastro thanks sir [14:40] jcastro: https://www.youtube.com/watch?v=75gKKnv_ze8&list=UUm7OifwnZoMCChidCJZQruQ [14:40] automatemecolema: no, not restufl, it's a websocket API [14:40] can anyone tell me what is the minimum nodes require to setup openstatck with juju ? can i do a all in one with juju ? [14:41] marcoceppi, this spec from kapil is from 2012 [14:41] galebba: the BARE minimum is two nodes. Recommended is seven, and if you want HA it's about 12 [14:41] I am asking for API docs, let's see if they exist [14:41] galebba: I've got a bundle that deploys 9 service nodes in a bare openstack setup [14:41] jcastro: it's a websocket API, not resful, and it's been implemented and the docs exist in the code base (kind of) [14:42] marcoceppi, hah, remember this? https://launchpad.net/~jrapi [14:42] Thanks guys, so the two minimum would be not counting the boot strap node i assume ? [14:43] galebba: it does include. its completely acceptable to deploy Openstack on a single machine to encasing VM's depending on the resources provided by that single machine. [14:43] automatemecolema: https://github.com/juju/juju/blob/master/doc/api.txt and there's a Python library that allows you to connect to the websocket, https://pypi.python.org/pypi/jujuclient/0.0.6 [14:43] lazypower-travel, do you have a link to that CMU OpenStack day? I am making the conference schedule [14:44] jcastro: I don't - i'm still wrangling for a date out of them. [14:44] galebba: it does, what you do is put all the supporting services on the bootstrap node in VMs, then use the second machine as your nova-compute node [14:44] (vms being containers, like LXC or KVM) [14:45] marcoceppi: good follow up, ty for clarifying my generic statement. [14:45] maroceppi thanks for the info, ill review it shortly [14:45] lazypower-travel: yeah, you really don't want to virtualize the compute node, virt in virt is pretty slow [14:45] ok thanks, i would like to use 3 physical nodes to distribute the boot strap/open stack control nodes. Any pointers to how to do that ? [14:45] marcoceppi: i'm aware. re: testing virtualbox inside of KVM [14:46] galebba: you can, there's no real pointers I can give other than how to do it [14:47] Don't use ceph, as it needs it own machine and doesn't work well virtualized, so you'd do everything with juju deploy --to lxc:1 (where 1 is the machine you're using for control nodes), then juju deploy nova-compute and create all the relations [14:48] it's be a very light weight deployment of openstack, keystone, mysql, the image service stuff, compute and compute controller [14:51] automatemecolema, so the core team tells me API docs are a work in progress btw [14:53] jcastro Do they have a roadmap time frame of when these will be GA? We are on the hinge about whether to use juju as our production service orchestration engine or use something more supported [14:55] We really want to create a Continuous delivery environment using Puppet, Jenkins, and Juju. Having the api docs will allow us to integrate Jenkins with Juju [14:55] automatemecolema: actually - you dont need to push directly to the websocket API for that automatemecolema [14:56] automatemecolema: there's a juju run command. if you've got your environment configs on your CI server, you can use the juju CLI to perform your workload management from the CI Server [14:56] automatemecolema, here's the start: http://godoc.org/github.com/juju/juju/state/api [14:56] i may be missing the specific requirements though, and just wanted to be clear that there are a few ways to tackle the integration. [14:58] automatemecolema: you may well be better off scripting against the juju CLI client... the API is not really designed to be consumed by outsiders consumers at the moment. You certainly *can*... there will just be a steep learning curve === natewarr_ is now known as natewarr [14:59] automatemecolema: from a technologically perspective it’s not RESTful, but it uses a websocket connection to transfer requests and responses encoded in JSON [15:01] natefinch I really sounds like maybe developing scripting utilizing the Juju CLI is the best route moving forward at this point. [15:02] automatemecolema: it definitely is where we have put the most development time. The CLI is the primary way we expect people to interact with juju, and things which are simple with the CLI may be quite complicated going directly against the API. [15:03] automatemecolema: there's basically only one thing the CLI can't do that you can do via the API, which is create watchers to get updated when stuff in juju changes (so like, getting a notification that a new machine was added). [15:30] where do i get a list of charms that i can use? https://jujucharms.com/ looks like some kind of demo site [15:31] and for folks on a cli server, should i need a desktop/browser to see a list? === mhall119_ is now known as mhall119 === vladk|offline is now known as vladk [15:40] hazmat, https://code.launchpad.net/~james-page/juju-deployer/pyyaml-fixup/+merge/223776 [15:40] busted in utopic right now [15:41] ack, thanks [15:42] jamespage, pushed to trunk [15:42] hazmat, ta === scuttlem` is now known as scuttlemonkey === seelaman` is now known as seelaman === vladk is now known as vladk|offline === vladk|offline is now known as vladk === vladk is now known as vladk|offline [16:23] hi, working through the nodejs charm tutorial... is it appropriate to hack up the mongo-relationship-changed hook? or should that be considered "don't touch if you can avoid" [16:23] use case is -- i want to run a bunch of data import scripts when mongo comes online [16:24] (my importers are in python, and the data lives in the source tree) [16:24] seems like the logical place to do this is in the mongo hook === vladk|offline is now known as vladk [16:28] all, I am a 100% novice to maas and juju. I am trying to get juju to speak to the maas controller and having trouble. This is my setup. [16:28] I have a maas controller with 2 nodes. Each node is a virsh vm. === vladk is now known as vladk|offline [16:29] I have juju installed on my x86 laptop with a 14.04 ubuntu trusty envoronment === vladk|offline is now known as vladk [16:29] I do not want the maas controller to install ubuntu or anything on the nodes. [16:29] How do I get up and running to installing charms on those nodes ? === vladk is now known as vladk|offline [16:30] I read that 'juju bootstrap' forces the maas to install ubuntu etc on the nodes which is not relevant for me. === vladk|offline is now known as vladk [16:30] What do I do after creating the environments.yaml file with the credential info to get to installing charms ? [16:30] ?? [16:32] how do i get a "unit" out of the *pending* state? [16:32] http://paste.ubuntu.com/7669926/ === vladk is now known as vladk|offline [16:36] anyone ? === vladk|offline is now known as vladk [16:39] purpledog2: the documentation is a little vague on that point :) [16:40] I think you are correct but there has to be someway to get it done .. it seems kinda basic ? [16:40] I think juju deploy is what you're looking for? [16:40] even the documentaion on the juju ubuntu site dicates "juju --sync-tools" precedint a bootstrap which is incorrect [16:41] purpledog2: I just noticed though that it looks like you want juju to use the existing ubuntu os on your nodes? [16:41] Do I not need to do anything else before juju deply charm ? [16:41] How do I dicate which node to deply the charm to ? === vladk is now known as vladk|offline [16:41] I need a way to know if basic setup and comunication works between the juju client and the maas server ? [16:41] purpledog2: (I'm way out of my depth here, but does this help? https://juju.ubuntu.com/docs/charms-deploying.html) [16:41] BTW does the maas server need juju to be installed ? [16:42] purpledog2: assuming you started with https://juju.ubuntu.com/docs/config-maas.html === vladk|offline is now known as vladk [16:44] jcw4: thanks .. Yes I followed the maas documentation. but I think juju deployment documeation needs differentiion between when the nodes work with VMs that are already created (in which case it doesn't need to really install ubuntu) vs nodes that are just physical bare servers with nothing on it [16:45] purpledog2: yeah, I'm not too clear on that, but at the very least jujud will need to be running on the machine [16:45] purpledog2: and I think ubuntu is expected to be the base OS [16:47] jcw4: thanks .. I do not think I have jujud running on the maas server. I will research that now. thanks for the tip.. [16:48] purpledog2: I believe that part of juju bootstrap is getting that installed and running [16:48] jcw4: I really wish I could find some real good tutprial on juju and a hands down on each step.. maybe if I have it figured out I will share it!! [16:48] However if a VM is up already there is no installation required and wish juju doc^n would mention what is expected when the nodes are virsh nodes. [16:50] purpledog2: yeah; usually there are folks in this channel who know enough to help; I'm just jumping in out of interest :) === isviridov is now known as isviridov|away === roadmr is now known as roadmr_afk [17:34] Anyone know about juju and why it would be causing my syslog to be zero length. perhaps related to the local provider? [17:34] setting up juju bootstrap node behind a proxy and getting the follwing curl error. Everything else seems to be downloading correctly with maas passing proxy/dns . Any idea how to get around this ? curl: (7) Failed to connect to streams.canonical.com port 443: Connection timed out === scuttle|afk is now known as scuttlemonkey === CyberJacob|Away is now known as CyberJacob [17:46] jcastro: jamie provided a nice answer to http://askubuntu.com/questions/485401/ but it's currently a vote behind the 'leader' -- mind giving jamie's answer a read and maybe evening out the vote count? :) thanks [17:46] sarnold, it helps if he mentions right up front that he helped spec, design, and implement the whole thing [17:47] jcastro: good idea [17:51] hi, i'm working on a charm, and trying to iterate. is it really the case that i have to remove-service and then add-service to test changes? why doesn't juju deploy Just Work on a failed charm? alternatively, does juju have a concept of "upgrade" ? [17:52] ok, i found upgrade-charm and --force [17:52] i'm going to try that [17:52] yep [17:52] achiang, make sure you read juju help upgrade-charm [17:53] for a failed deploy [17:53] you'll want `juju resolved --retry foo` [17:54] oh [18:16] bac, arosales: you guys need to stop stepping on each other in my PR please [18:16] jcastro: i think where there is conflict it is implementer's choice [18:16] jcastro: /me just adding comments [18:18] ok well, I've been trying to go to lunch for 30 minutes but you guys have turned "add quickstart" to "let's do an entire review of every page this touches" [18:18] and i was just adding clarification. at no time did i call arosales a mutton-head [18:18] which is fine, we can just do that after. [18:19] jcastro: feel free to take lunch now and come back to comments later [18:19] I would just like to commit! [18:19] jcastro: patience :-) [18:19] comments are good :-) [18:19] and that's fine === scuttle` is now known as scuttlemonkey_ === scuttlemonkey_ is now known as scuttle|afk === roadmr_afk is now known as roadmr [18:41] jcastro: my comments are in for your https://github.com/juju/docs/pull/123 === vladk is now known as vladk|offline [18:41] jcastro: thanks for working on that === scuttlemonkey_ is now known as scuttlemonkey [19:15] jcastro: hey, is there a known issue with the nodejs charm? http://pastebin.ubuntu.com/7670829/ [19:15] above occurs when i remove-relation, followed by remove-service [19:16] on the phone === roadmr is now known as roadmr_afk === pmatulis_ is now known as pmatulis === roadmr_afk is now known as roadmr === natewarr_ is now known as natewarr [20:22] setting up juju bootstrap node behind a proxy and getting the follwoing curl error. MAAAS boots up the node and installs ubuntu just fine.Seems to be curl not setup for proxy ? Any idea how to get around this ? curl: (7) Failed to connect to streams.canonical.com port 443: Connection timed out [20:48] Can anyone tell me how to have juju associate an haproxy charm with an Elastic IP on Amazon? [20:48] Does anyone have knowledge about how a juju charm could pick up a specified EIP on EC2? Then for kicks, run these with 2 nodes for HA? [20:48] /delete automate_ === scuttlemonkey is now known as scuttle|afk === scuttle|afk is now known as scuttlemonkey === scuttlemonkey is now known as scuttle|afk === scuttle|afk is now known as scuttlemonkey === scuttlemonkey is now known as scuttle|afk [21:54] Hi ALL.. Any juju openstack expert in this IRC .. Need some help to deploy openstack in 2 nodes using juju [22:02] Hi ALL.. Any juju openstack expert in this IRC .. Need some help to deploy openstack in 2 nodes using juju [22:05] hello Delair -- note IRC works best if you just ask whatever questions you want to ask :) [22:05] sure .. didnt wanted to bug everybody.. [22:06] Can I install openstack using JUJU + MAAS in 2 nodes only .. ONE to use a controller/network and SECOND to use as compute node only.... [22:06] i did setup juju+maas with no issues [22:07] Delair: what you do is put all the supporting services on the bootstrap node in VMs, then use the second machine as your nova-compute node (vms being containers, like LXC or KVM) Don't use ceph, as it needs it own machine and doesn't work well virtualized, so you'd do everything with juju deploy --to lxc:1 (where 1 is the machine you're using for control nodes), then juju deploy nova-compute and create all the relations it's be a very light weight [22:07] deployment of openstack, keystone, mysql, the image service stuff, compute and compute controller [22:07] but whenever i try to deploy services it automatically tries to choose nodes [22:07] hey lazypower-travel :) [22:07] o/ sarnold [22:09] Delair: what you need to do [22:09] is run juju add-machine [22:09] this will allocate the second node to juju for use for the compute node [22:09] run juju deploy nova-compute [22:09] that will place it on it's own node [22:09] for everything else [22:09] you need do juju deploy --to lxc:0 [22:10] Thanks lazypower and marco [22:10] that will put it on the bootstrap node as an LXC container [22:10] marcoceppi: read above ^ i got you covered. Paraphrased from earlier. [22:10] this is exactly what i want and what i tried to do [22:10] lazypower-travel: ah, didn't see your message with the --to [22:10] its all buried :) [22:10] i bootstrapped to one node [22:11] and tried manually deploying services using --to option to one node [22:11] Delair: did you use --to lxc:0 ? [22:12] but as soon i tried to build realtion using sql and keystone it give me error messsage "hook failed: "config-changed"' [22:12] I tried lxc also [22:12] Delair: is the error on mysql? [22:12] Delair: did you read the README on the MySQL charm? [22:12] marco the error was on keystone [22:12] o nvm [22:13] lazypower the issue with lxc is that it assign 10.30.0.x ip to each container which i dont know how to access from outside [22:13] odd, can you run juju status and redirect it to pastebinit? (juju status | pastebinit) [22:13] @lazypower no i didnt ready the README on mysql [22:14] @marco not sure what is pastebinit [22:14] i tried to look online to fix the "hook failed: "config-changed"' and looks like it is a bug [22:15] SO :) [22:16] is there system requirement and step by step document which i can following to install openstack in 2 nodes === hmatlock is now known as heath === CyberJacob is now known as CyberJacob|Away