[05:07] bigjools: o/ [05:07] yo [05:08] bigjools: yo! [05:08] whats up? :) [05:08] should I ask what time it is there? :) [05:09] 1.10 am :) [05:10] nice [05:10] i drank colombian coffee and cant sleep :/ [05:10] I am only 14h ahead of you then :) [05:10] heh:) [05:10] when's the SRU going out mate? [05:11] bigjools: i uploaded on friday [05:11] gonna ping ppl tomorrow [05:11] ah it's in -proposed [05:11] (or later today) [05:11] there's another fix to go in, sadly, but I think you know. It can wait for later anyway. [05:11] bigjools: i can still upload i think [05:12] that'd be good [05:12] will look into that tomorrow/later today [05:12] just need to check the "fix committed" stuff on 1.2 that is not milestoned [05:12] I'll release the milestone once it's all in updates [05:12] ok cool [05:13] * bigjools high-fives roaksoax [05:13] bigjools: yeah finally we are almost done \o/ [05:13] it's been a long road ... [05:13] indeed [05:14] but almost there :) === jtv2 is now known as jtv [08:55] Howto perform OpenStack upgrade from Essex to Folsom on juju maas | http://askubuntu.com/q/267256 [10:07] UBUNTU MaaS Region and Cluster Controller in One Box | http://askubuntu.com/q/267284 [12:33] is there any doc on how to use /usr/sbin/maas-import-pxe-files in an environment not connectetr to the internet ? [12:33] like, downloading the iso manually, and feeding them into maas ? [13:21] melmoth: no... you need internet connection. it is a requirement === matsubara_ is now known as matsubara [15:08] racedo, here i am :-) [15:08] hey melmoth [15:08] hi roaksoax melmoth is trying to setup an offline environment with maas using mirrors [15:08] so, i have a problem, i m trying to use maas (1.2+bzr1360+dfsg-0ubuntu1~ppa1) on a lab not connected to internet [15:09] i have used reprepro to mirror archives.ubuntu.com (main updates security restricted and universe) [15:09] i also have mirrored in a separate subdir the ppa for juju and for maas (so i can install maas) [15:09] now, i have maas installed, but when i start a node, it try to contact archives.ubuntu.com [15:09] melmoth: when enlisting the nodes, do they try to go online to pull the packages? [15:09] yes [15:10] is it cloud-init configuring sources.list? [15:10] i want to be able to tell the machines where are my repos, and not to use the real upstream one [15:10] i have no idea [15:10] i never had to add anything related to cloud init anywhere before [15:11] btw, this is the step by step doc http://bazaar.launchpad.net/~pierre-amadio/+junk/c6100-isolated-maas-deployment/view/head:/README.txt [15:11] basically just how to set reprepro [15:11] i know some people workarouned this by trying to play with name resoltuion [15:12] and have archive.ubuntu.com point to the local repo, but i m not sure this is a good idea. [15:12] (plus i m not sure where to do that anyway ;-) ) [15:12] i would rather have the preseed file changed so they use _my_ mirrors. [15:13] melmoth: the images may already contain this, we may have to modify them [15:13] check /var/lib/maas/ephemeral/precise/ephemeral/amd64/20121008 [15:14] hmmm [15:14] hep [15:14] that s a good idea [15:14] sudo mount -o loop disk.img [15:14] and check the sources.list [15:15] yeh, i m trying that [15:15] looks good [15:19] melmoth: its probably easier to makecloud init do the repo change [15:19] smoser: ^^ [15:20] i dont know, i never used any mirror tool. I first try with apt-mirror, then i have been told to use reprepro because it ll be easier [15:20] note that one of the trick also is, i want to have a mirror with all the ppa stuff needed. reason is, i dont want anything to do with injecting gpg key with ppa styff [15:20] cause it ll be just fail too [15:21] grumble, still trying to access security and archive repo [15:22] melmoth: i mean makr cloud iniy use the internal repo [15:22] melmoth: http://askubuntu.com/questions/144393/how-to-let-maas-cloud-init-client-select-internal-mirror [15:22] oh [15:22] melmoth, you can probably accomplish this. but you might have to chagne the commissioning_user_data and the enlistment_userdata. [15:23] i *think* if you change the preseed that that will get refelected through and 'stick' after installation [15:23] doing this via proxy "should work" without any tampering from trunk now. [15:23] but i'm not sure about full mirror [15:23] that seems to be for the old version of maas, is it not ? (the above ask ubuntu link) [15:24] (personally, i'd recommend proxy) [15:24] melmoth: (yeah but thats somethibg similar) [15:24] i did try to change the preseed file in the generic file. [15:24] without success [15:25] the link above is something that i expect is now fixed. [15:25] (as maas tells cloud-init to 'apt_prserve_sources_list: true' itself) [15:25] even if you are behind a proxy, things fail (that s my personnal problem for 3 monthes now), as soon as you need a ppa. [15:26] but anyway, in this particular case, the pre requisite is "completly isolated network" [15:28] melmoth: yeah ppa's behind proxy indeed fail but there are work arounds [15:28] melmoth, you can proxy ppa for sure. squid-deb-proxy needs tweaking to make it do that. but you can do it. and it is not 100% offline, but its pretty good. [15:28] any link that explain the workaround for hte ppa behind a proxy ? [15:29] my ppa & proxy issue is https://bugs.launchpad.net/ubuntu/+source/software-properties/+bug/1089389 [15:29] Launchpad bug 1089389 in software-properties (Ubuntu) "juju bootstrap fail behind a proxy when a gpg key must be imported" [Critical,Confirmed] [15:30] the "injecting gpg key manually" workaround is not good enough, it must be done withing the zookeeper node during bootstraping. [15:30] melmoth: yeah you need to add late commands to add the ppa and stuf [15:30] like, changing th preseed file so it inject the gpg key manually ? hmmmm [15:31] gosh i hate preseed :) [15:31] yeah [15:31] didn't we hit something similar when on site last time? [15:31] so that way it is done automatically [15:31] racedo: we did [15:31] racedo, we currently have such a problem [15:31] that s the whole reason why i m trying to build stuff in an isolated env. [15:31] and we injected the gpg keys in the charms? [15:32] racedo: i think just in late_command [15:32] ok [15:33] * roaksoax looks for it [15:34] and melmoth for the ppa issue, would this be doable: mount -o loop disk.img, chroot to it and apt-add-repo? [15:34] racedo: no need [15:34] i dont know [15:34] ups not really [15:34] you know, i have no real understanding of what maas is actually doing. [15:34] you can modify enlist_userdata and commissioning-user-data [15:35] so, if it try to add the repo anyway, even if it still there, may be it ll fail. may be not...who knows. [15:35] roaksoax, and those are preseed snippet, right ? [15:35] melmoth: its really not maas itself [15:36] melmoth: it is external factor that we have to workaround in maas [15:38] i dont undertsand what are enlist_userdata and enlist file. Looks mor elike bash scrpit than pressed [15:39] so i m not sure if i need to add d-i stuff in it or echo > stuff < melmoth: enlist_userdata is cloud [15:40] melmoth: enlist_userdata is cloud init's user data [15:40] for enlistment process [15:40] /etc/maas/commissioning-user-data is cloud init's user adata for commissioning process [15:41] ok, as we say in france "c est comme le port salut, c est marqué dessus" (it s written on the name :) ) [15:41] now, i need to read doc about cloud init user data then. [15:43] ahh, yaml... [15:43] i need to read doc about yaml :) [15:45] racedo: great!! The Paste you are looking for does not currently exist. [15:45] Return to the Pastebin [15:45] racedo: all the stuff was in pastebins and now they are gone [15:45] oh ok [15:46] all the tweaks we did in colorado [15:47] i think we just include it in one charm [15:47] not sure how [15:47] racedo: yes, we included it for the cloud archive [15:47] racedo: did you or negronjl backed that up? === matsubara is now known as matsubara-lunch [15:48] i didn't negronjl was the charms person on site that week [15:48] let me quickly ping agy [15:49] racedo: already did :) [15:49] oh ok :) [15:53] Could not internally obtain zookeeper handle | http://askubuntu.com/q/267409 [15:57] racedo: ping [15:57] err [15:57] sorry [15:57] rvba: ping [15:57] rvba: ping [16:01] roaksoax: hi [16:02] melmoth, :-( forgot about that bug. [16:28] ah ah, enlisting seems to go a bit further. [16:29] gosh, please tell me i dont have to mirror the sources too [16:29] yeahhh, got one node declared \o/ [16:35] melmoth: so in your proxy environment, could you try things like this: sudo apt-key adv --recv-key --keyserver keyserver.ubuntu.com --keyserver-options http-proxy=http://192.168.123.2:8000/ 93EE8CC5 [16:35] melmoth: or sudo apt-key adv --recv-key --keyserver keyserver.ubuntu.com 93EE8CC5 [16:35] in the first you'd need to specify the proxy you are using [16:35] it ll work, it s actually written in th ebug if i remember correctly [16:36] melmoth: ok so something like that is what you need to add [16:36] to a late_command [16:36] in the preseed [16:36] and manually add the ppa repository [16:36] ahh, yes. that could be an option, in a late command, that, or, as i was just reading cloud init stuff apt_sources: in the cloud init config file [16:36] i will try it out next time i have nothing else to do :) [16:37] hopefully next week [16:37] depends how this isolated stuff turns out [16:38] yes... ready state \o/ [16:46] hmmm, it does not like my mirror :-( [16:48] when i bootstrap, the zookeeper node download the Release file from my mirror [16:48] there is a Codename: precise in it. [16:49] i can wget it all rightr from the console of the zookeeper node, but still, it complain the mirror does not support the precise release. [16:50] melmoth: are you telling the preseed to use the mirror? [16:50] yes, it does it all right, i see it downloading the release file from my mirror [16:51] melmoth: did you remove the proxy option from the preseed? better ocmment it [16:51] hmm. not sure [16:52] melmoth: you'll probably need to do that [16:52] melmoth: so installations don't access the proxy in maas, and use the mirror directly [16:53] not sure what to do, there is a /usr/share/maas/preseeds/generic file that define a proxy marco or something [16:54] hmm, actually, i already edited this file, to put my mirror ip in it. [16:54] i have a strange feeling i have no idea what i m doing. [16:54] let s remove the "d-i mirror/http/proxy string http://{{server_host}}:8000/ [16:54] " line [16:54] just in case. [16:56] melmoth: yeah [16:58] now, it seems to complain because my mirror has no /ubuntu/dists/precise/Release.gpg [16:59] melmoth: maybe you need to correct the path for the mirror [16:59] in the preseed [16:59] and that would be it [17:00] well, i do not have any file ending with .gpg in my repo. [17:01] then that's probably whhy [17:01] i think i start to understand the "VerifyRelease: blindtrust" option i used in reprepro [17:17] How to PXE boot a virtual machine from another virtual machine which are in same virtual network? | http://askubuntu.com/q/267435 === matsubara-lunch is now known as matsubara [20:43] Hi, I'm trying to automate my MAAS cluster with the maas cli. I've written a script to do this but I can seem to get the maas nodes to allocate to a user after they've been released. [20:43] My basic algorithm is: [20:43] maas-cli login mezeoadmin [20:43] maas-cli mezeoadmin node release [20:44] … Do this for all the nodes. [20:44] maas-cli mezeoadmin nodes accept-all [20:44] mross: accept-all is only done when they are first enlisted in MAAS [20:44] ssh -o "StrictHostKeyChecking no" ubuntu@node "sudo reboot" [20:45] Then I'm going to the web interface and pressing 'start node' for each node that has been released. [20:45] This seems to work, but I would like to automate it completely off the cli. [20:46] What am i missing. [20:46] ?? [20:46] @roaksoax OK. So I can remove that. [20:46] mross: 1. why don't you use juju ? :) [20:47] Because I haven't written a juju plugin for our service yet. [20:47] I'm just getting familiar with maas right now. [20:47] mross: 2. when you 'Start node' on the web UI, they get turned on to install ubuntu on them [20:47] mross: you could use juju deploy ubuntu [20:47] mross: and that will simply install ubuntu on the nodes [20:47] and on topof that you could deploy your service [20:48] So juju deploy ubuntu to nodes that are in a ready state will install ubuntu on nodes that haven't been allocated? [20:48] Or do I have to allocate them to a user then deploy? [20:49] mross: nodes are allocated when a user wants to install ubuntu on them, in the case of juju, it will take nodes that are in 'Ready' state to deploy ubuntu on them and install a service [20:52] OK thanks. Is this a decent workflow for what I'm trying to do. [20:52] mross: so in order words, you do not allocated a machine to a user, so the user can install ubuntu on it [20:53] however, the user can use a machine from the Ready pool, once it selects that machine to be used, the machines becaomes allocated to the user [20:53] Correct [20:53] mross: personally, I'd recommend you to use juju [20:53] makes things so much simpler [20:53] if you need to automate the dpeloyment of your service, you'll need to write a charm that installs and configure your service [20:53] a juju charm [20:54] Any good examples you can send me for my particular use case? [20:55] mross: so for example, you could use: juju deploy wordpress, juju deploy mysql, juju add-relation wordpress mysql [20:55] that deploys wordpress and mysql in different nodes, and adds a relation between them saying that wordpress will use the mysql node as its backend db [20:59] OK. I think I know where you're headed here. [20:59] And where I need to go. Roak thank you very much! [21:17] mross: more infomrmation in maas.ubuntu.com and juju.ubuntu.com :) === kentb is now known as kentb-afk [23:35] roaksoax: re. bug 1154805 [23:35] bug 1154805 in MAAS "Juju/MAAS does not provide feedback when there are no available nodes are found" [High,New] https://launchpad.net/bugs/1154805 [23:36] I am pretty sure this is just a juju bug [23:36] see the recent thread on juju-dev where I suggested that the user experience is poor [23:36] bigjools: yeah i was thinking that, but just to make sure I included maas [23:37] roaksoax: it's a provider bug in juju [23:37] or perhaps a core bug [23:37] I dunno [23:37] bigjools: yeah that's why I wasn't sure either [23:37] but maas returns a 403 IIRC [23:38] bigjools: maas does show an error [23:38] bigjools: but juju doesn't [23:38] roaksoax: in the log? [23:38] bigjools: yeah maas shows NodesNotAvailable: No matching node is available. [23:38] bigjools: but for juju ois just as like nothing happened [23:38] bigjools: i'm more concerned on the contraints bug though [23:39] bigjools: ppl have been complaining about constraints in customer deployments [23:39] ah no it returns httplib.CONFLICT [23:39] roaksoax: it's definitely a juju bug [23:39] it has no notion of not being able to get a node, it just waits for one forever (I think it retries) [23:40] yeah maybe, TBH i didn't look much into it since i needed to get that stuff working [23:40] feel free to retarget accordingly [23:41] done :) [23:41] bigjools: i think this is more important though: https://bugs.launchpad.net/maas/+bug/1154803 [23:41] Launchpad bug 1154803 in MAAS "Default constraints prevents the allocation of nodes if they are equal to the HW." [High,New] [23:41] looking [23:42] roaksoax: well that sucks :/ [23:42] bigjools: yeah i was thinking might be related to node having 1, and the constraints 1.0 amybe? [23:42] roaksoax: I don't know who will fix that [23:42] bigjools: mgz? :) [23:42] it won't be us I'm afraid, we're super busy on another deadline [23:43] roaksoax: should be easy to fix if you write a test to re-create it :) [23:44] bigjools: sure, I'll try to do it this week, im packed getting other stuff to work [23:44] * bigjools hears you [23:44] roaksoax: FWIW, all of my team are with me here in Brisbane next week [23:44] in case you expected them to be around to ping :) [23:45] bigjools: hehe ok :)