=== kadams54-away is now known as kadams54_ === anthonyf is now known as Guest23073 === natefinch-afk is now known as natefinch === anthonyf is now known as Guest46780 === kadams54_ is now known as kadams54-away === anthonyf is now known as Guest47313 === kadams54 is now known as kadams54-away === kadams54-away is now known as kadams54 === scuttle|afk is now known as scuttlemonkey === kadams54 is now known as kadams54-away === kadams54-away is now known as kadams54 [15:58] PSA: we're migrating jujucharms.com to the new environment and experiencing some glitches atm. Work in progress === JoshStrobl|AFK is now known as JoshStrobl === kadams54 is now known as kadams54-away === kadams54-away is now known as kadams54 [16:45] PSA: migration aborted for now, things should be back to 'normal' === braderhart_ is now known as braderhart [17:00] When adding a machine, should it set up a virbr0 with the ip 192.168.122.1 like it does on the bootstrap node? [17:00] using manual provisioning [17:04] Destreyf: I'm not sure, I think that virbr0 is only for KVM stuff? [17:05] my 2nd machine is trying to connect 192.168.122.1 [17:05] and it can't reach it === arbrandes_ is now known as arbrandes [17:28] Destreyf: juju uses hardcoded ip addressing with host only networking when deploying to containers/kvm [17:29] lazyPower: so there's no way to mix lxc and physical machines? [17:29] Destreyf: this is a known issue, and we have some solutions to work around it using flannel, calico, and a few other SDN providers that are being worked on [17:29] ah okay [17:29] lazyPower: that was the information i needed, thank you [17:29] Destreyf: there is, but if you need reachability into that instance, you should look at flannel-docker charm which will spin up a flannel0 interface [17:29] and that gives you a private networking space to interfact with services on your network [17:29] if you're using LXC containers [17:29] there's additional work that needs to be done [17:30] we dont have a magic bullet charm that will configure your containers properly atm [17:30] well, kind of, but its a weird deployment pattern [17:30] theres an older flannel charm in ~hazmat's namespace [17:30] and if you deploy that charm to a host *first* [17:31] as in before doing juju deploy --to lxc:# [17:31] it will configure flannel based networking for any lxc containers created after the fact [17:31] lazyPower: I'm trying to do something stupid, but its a test case scenario [17:31] fortunately it has the binaries in it, so old but still works ;-) [17:31] hazmat: o/ :) [17:31] lazyPower: any experience with using juju on debian? [17:31] hazmat: its funny you say that, i just updated that entire ecosystem [17:31] lazyPower: i know [17:31] because juju... is fantastic in its relationship management [17:32] Destreyf: i haven't personally, no [17:32] i've used juju on windows, and juju on ubuntu so far [17:32] there's one or two centos based charms out there too that our partners cloudbase have been working on [17:32] but i have yet to get my hands on a debian charm [17:32] Destreyf: there's some patches out there for it, but its not out of the box afaicr, the cloudbase folks might know more [17:32] hazmat: its inc. actually [17:32] the work has been done [17:32] i dont think its landed yet tho [17:33] lazyPower: oh.. awesome [17:33] hazmat: mbruzek was looking for dockercon hackathon participants btw, do you have any interest in hackathon'ing with us? [17:34] hazmat, lazyPower thank you very much for your time, you've answered questions that i've been racking my brain on for 2 days now [17:34] Destreyf: anytime :) [17:34] and on that note i need to grab a quick lunch, if you need anything else dont hesitate to ping Destreyf [17:34] i'll circle back when i get back from lunch [17:35] lazyPower: i'll be there, haven't decided on anything yet in terms of projects, i've got two ideas, one is really boring, but just a pull request idea i wanted to work on (ie. won't win, but accomploshies a personal goal), the other is more like a gee whiz thing, also not a winner. preference is to hack in go at the event [17:35] hazmat: i think they would be receptive to *any* ideas [17:35] i came up with some good suggestions on how to get ideas but no clear idea winners. so far its just mbruzek and wwitzel3 for camp juju [17:35] they just released that you can have up to 10 ppl in a hack team [17:36] which is a big change from the eventbrite page we signed up on so many moons ago [17:36] it was originally 3 per pod [17:36] so nutshell one was adding proxy support for building images, the other was setting up an ssh server that attached to a container namespace. [17:36] so we're scrambling to either find ppl or come up with a game plan to show up and just pollenate other teams [17:36] oh dude [17:36] ssh server + container namespace = interesting [17:36] whats the objective behind that? [17:37] lazyPower: darned if i know ;-).. solomon suggested it to me at the ams dockercon hackathon [17:37] haha [17:37] fair enough :) [17:37] allright, brb o/ [17:40] hazmat: on flannel [17:40] the 10.10.x [17:40] is that x the machine number? [17:40] so machine 0 is 10.10.0.0/24 [17:40] and machine 1 is 10.10.1.0/24 [17:41] Destreyf: basically yes, the 10.10.0.0/16 is just the cidr for the cluster, each machine gets a /24 for containers on it [17:42] hazmat: so the 3rd octet is different for each machine though? [17:42] Destreyf: ie. its more like a container on machine 0 has 10.10.0.1/28 and a container on machine 1 has 10.10.1.1/28 .. the machines themselves have subnets assigned to them, but they don't have ips from the cluster range [17:42] k [17:44] Destreyf: upstream has a few more modes added since my version of the charm was first published, lazyPower would know more about the latest version, most importantly they support configuring under iaas substrates for zero overhead, and using in kernel data plane paths (vxlan over udp) for faster performance.. the version in my charm was just doing udp [17:44] overlay afaicr. [17:49] hazmat: do you recommend i use it from yours or the upstream? [17:50] Destreyf: i'd use lazyPower's newer one, i'm not actively working on charms atm [17:51] Destreyf: not sure if the newer one also works with lxc, there's also limited juju support in 1.24 for managing iaas routing for container addressabilty (lxc) with ec2 (only using default vpc if enabled on an account and maas) [17:51] hazmat: hsi is listed as flannel-docker [18:05] right, it will install the flannel0 bridge if related to a non-docker host [18:05] otherwise it reconfigures docker for use with flannel [18:06] there's more work that needs to be done w/ flannel to make it work with LXC containers. namely lxc introspection, reconfiguring of existing containers, and ensuring lxc is reconfigured to start new containers w/ the flannel networking enabled by default [18:06] looks like i'm uzing hazmats :D [18:06] ya know, small stuff :) [18:06] i would for now, its older, but gets the job done. [18:06] once we have the newer stack updated for lxc, migration should be a short hop jump skip and --replace away [18:06] er --switch === kadams54 is now known as kadams54-away === kadams54-away is now known as kadams54 === kadams54 is now known as kadams54-away [18:50] * mbruzek reads scrollback [18:50] Thanks for the ideas hazmat [18:50] hazmat: do you have any interest in the hackathon competition? === kadams54-away is now known as kadams54 === natefinch is now known as natefinch-afk === kadams54 is now known as kadams54-away [21:51] hazmat: sorry to bug you so much, how hard would it be to change the ip address inside of flannel? [21:53] Destreyf: the ip address you get back from flannel is determined by etcd [21:54] hazmat: so under hoooks, i see flannel_network set to 10.10.0.0/16, this would be the spot right? [21:54] hooks.py* [21:54] oh [21:54] lazyPower: replied not hazmat [21:54] :D [21:54] i saw a reply and assumed [21:54] that's bad of me [21:54] Destreyf: yeah if you change that it should change the A block [21:55] awesome, what would be the best way to go about routing connections to these? so i can access from say my pc (currently) 10.10.81.254 [21:55] my ip is 10.5.1.36 [21:56] Destreyf: im not sure I understand what you're trying to do [21:56] me either, let me think it over [21:56] i appreciate your help [21:56] are you trying to setup a vxlan between your desktop and those containers? [21:57] if thats the case, you have 2 options. 1) setup a vpn style connection to a machine that has a flannel0 interface. Such as using sshuttle [21:57] 2) Install flannel on your pc to gain teh flannel0 interface and join in the vxlan. [21:57] basically, say i install a service, say wordpres on a lxc container, using hazmat's flannal charm, how can i access the wordpress that' d be hosted there [21:57] ah, why not deploy haproxy on the parent node [21:57] and reverse proxy to the wordpress charm? [21:58] and you are geneous [21:58] you dont even need flannel for that :) its host-only networking [21:58] lazyPower: reason for flannel is HA setups between machines, containers and the likes [21:58] on the LXC side [21:59] Destreyf: check this out [21:59] http://blog.dasroot.net/container-networking-with-flannel.html [21:59] its a little dated now [21:59] i need to do an updated post with the new stuff [22:00] also lazyPower i haven't looked in depth, but any chance you guys support (or would support) linode as a backend? [22:00] we currently support linode with the manual provider [22:00] lazyPower: that's what i figured [22:00] but you do lose some of the magic, as you manually add machines to your env. [22:00] lazyPower: i'm currently doing manual adding anyways for tests/local hardware without MAAS [22:00] there's been talk of writing a linode manual provider plugin based on hazmat's work w/ the DO provider plugin but nobody's taken on the project yet. [22:01] lazyPower: i might just do it one of these days [22:01] awww the link's broken [22:01] XD [22:01] which link Destreyf? [22:01] on that article you linked [22:01] hmm.. shouldn't be [22:01] it mentions his do provider [22:02] http://blog.dasroot.net/juju-digital-ocean-awesome [22:02] lame :| [22:02] http://blog.dasroot.net/juju-digital-ocean-awesome.html [22:02] yeah [22:02] found it [22:02] thats what i get for migrating from ghost to pelican without running a broken link checker against my content [22:56] * hazmat yawns [23:16] this might be a flawed finding but i noticed that juju log logs any std out that your program writes regardless if you call log or not. Is that weird? === blahdeblah_ is now known as blahdeblah [23:57] cholcombe: nah... expected [23:57] but probably not well documented :) [23:57] o/ hazmat [23:58] thumper: greetings [23:58] hazmat: do you know much about tox? [23:58] hazmat: I have a MP up for python-jujuclient [23:58] but couldn't get the tests working [23:58] thumper: what was the error? [23:59] * thumper finds the mp [23:59] https://code.launchpad.net/~thumper/python-jujuclient/jes-cache-file/+merge/261586 it is on there