=== matsubara-afk is now known as matsubara === matsubara is now known as matsubara-afk [13:03] allenap: ping [13:10] roaksoax: otp, done soon. [13:19] is it possible to add maas server as a maas node? I would like to set up the maas server as the cloud-controller, glance and keystone. [13:20] guimaluf: yeah you can enlist localhost, but for juju is not gonna be usefull because the maas server is the onein chanrge of turning on the machines you need [13:22] roaksoax: Going to be a bit longer. [13:22] allenap: no worries, let me know when you done :) [13:24] roaksoax, so my maas server should be uniq for this purpose? cause I've a powerfull server and run only the maas server will be a waste of resources. [13:25] guimaluf: yeah, unless you deploy VMs with it [13:25] then should I put the maas server in a virtual machine ? [13:25] deploy vms with juju? [13:25] guimaluf: you could, or you could have VMs contorlled by maas [13:25] guimaluf: for juju they would only be nodes [13:26] hmm [13:27] I can't figure out how I'll use my 16 servers in juju + openstack IaaS [13:29] roaksoax, another issue: when I run juju status openstack-dashborad there is public-address:null [13:30] I dont know how will I make my cloud avaible. [13:30] when will my node get the hypervisor installed? [13:31] I'm bit confuse :( [13:31] and bothering you alot, I thin [13:31] think [13:31] TBH I have not done many openstack deployments with juju/MAAS [13:31] so I don't know much about those bits [13:31] guimaluf: someone who would be able to help you is adam_g [13:32] but he is at OSCON atm [13:33] :( [13:46] well, if juju doesn't have the address, it would seem a general juju<->maas issue. [13:59] roaksoax: What's up? [13:59] The stuff still hasn't gone upstream. [13:59] :) [13:59] allenap: ok, so, there's something wrong with maas-pserv [13:59] allenap: there's some weird thing happening. maas-pserv fails to start [14:00] but then magically, it starts [14:00] roaksoax: Do you have some logs for it? [14:00] allenap: there's nothing :) [14:00] allenap: can you spawn a canonicastack instace and install maas from quantal [14:00] and see it for yourself to see if you have any ideas of what it might be? [14:00] roaksoax: I wonder if this is still the celeryconfig problem? [14:01] allenap: might be, I imported the path [14:01] roaksoax: I've never used canonistack, so it might take me a while to figure it out. [14:01] roaksoax: If it's that, I'm working on it. [14:01] allenap: so upstart says twistd doesn't know who 'maas-pserv' is [14:01] allenap: so it is failing somewhere when the plugin does the import [14:01] roaksoax: Yeah, that sounds right. It can't import celeryconfig probably, but the import exception is being suppressed. [14:02] allenap: I added the path in provisioningserver/plugin.py so I know it is importing thepath correctly [14:03] roaksoax: Let me fix the celeryconfig thing. If the problem is still there then I'll spend more time on it. Will that block you? [14:03] allenap: yeah we have a broken maas in quantal :) [14:04] roaksoax: It's fairly broken in Precise too ;) [14:05] allenap: heh but precise at least works [15:41] roaksoax, when I expose a service I get the following error: http://pastebin.ubuntu.com/1098491/ [15:41] I'm completely discouraged [15:42] guimaluf: AFAIK, you cannot expose services with maas as that's like opening a port in ec2 [15:42] that's only a feature with ec2 i thi nk === matsubara-afk is now known as matsubara [15:43] so how will I make my service avaible? [15:51] guimaluf: the service is available [15:51] guimaluf: if you are deploying instances with juju inside openstack, then you can use the expose command I believe [15:52] roaksoax, I have onde node on my juju environment, I deployed, related and exposed many services, but where will I access my service? I cant get the public-address value, and I got this error on node juju log [15:54] guimaluf: "Service exposing works by opening appropriate ports in the firewall of the cloud provider." [15:54] if you are deployimng in baremetal that ncommand is not necessary [15:55] but when I access my node address nothing happens [15:55] all my services have agent-state:pending [15:55] :/ [15:56] that means that they are trying to get installed or something wrong is going on there [16:12] roaksoax, probabily something wrong in the node right? [16:12] guimaluf: yes in the node [16:14] there is nothing in /var/log/juju :/ only the error above [16:16] I'll install openstack without juju :/ time is near :( [16:17] guimaluf: is that on the client machines? [16:17] roaksoax, yep [16:17] /var/log/juju/provision.log [16:17] guimaluf: there's no machine-agent.log nor provision.log? [16:17] provision log http://pastebin.ubuntu.com/1098491/ [16:29] rvba: bug #1026228 [16:29] Launchpad bug 1026228 in MAAS "No way to DELETE or get a node back to 'Ready' from WebUI" [Undecided,New] https://launchpad.net/bugs/1026228 [16:29] guimaluf: there was an error with juju apparently, do you have any cloud-init logs? [16:39] roaksoax, http://pastebin.ubuntu.com/1098589/ [17:17] guimaluf: were you seeing this error? 2012-07-18 13:14:46,016:1882(0x7fe067770700):ZOO_ERROR@handle_socket_error_msg@1579: Socket [127.0.0.1:44068] zk retcode=-4, errno=111(Connection refused): server refused to accept the client [17:19] roaksoax, nop. only ZOO_INFO [17:20] sok [17:21] smoser: there seems to be an issue when importing the data source after the bootsrrap [17:25] roaksoax, ? [17:25] smoser: juju does not get installed after juju bootstrap with maas [17:26] i'm redploying [17:26] because cloud-init isn't getting its datasource [17:26] smoser: it is getting the ssh keys from MAAS [17:26] not the juju stuff though [17:27] roaksoax, can you give me more info and/or access? [17:28] smoser: is in my local environment, let me finish re-bootstrapping [17:28] smoser: apache: http://pastebin.ubuntu.com/1098663/ [17:31] smoser: ah never mind I think I found the issue [17:31] it seems to be proxy [17:34] smoser: for some reason sources.list ends up with ubuntu-mirror [17:34] smoser is that done by cloud-init? [17:35] roaksoax, yes. [17:35] you have broken dns [17:35] https://bugs.launchpad.net/ubuntu/+source/cloud-init/+bug/974509 [17:35] Ubuntu bug 974509 in cloud-init (Ubuntu) "cloud-init selects wrong mirror with dns server redirection" [Low,Confirmed] [17:36] cloud-init noticed that you had a host on your network (named 'ubuntu-mirror') and assumed that you wanted to use that mirror. [17:36] smoser: I don't this is MAAS on a VM [17:36] your issue is that you also have a host named "roaksoax-uses-a-broken-dns-provider-that-does-dns-hijacking" [17:36] dns/dhcp is libvirt [17:37] try it. [17:37] thats the problem. cloud-init does a dns lookup and its getting a valid response for that 'ubuntu-mirror' [17:38] (rather than a not found response) [17:38] smoser: right, but it is libvirt the one controlling DNS [17:38] well libvirt is letting dns pass through [17:38] dnsmasq is cacheing only for libvirt. [17:39] change your hosts dns to use 8.8.8.8 and restart libvirt. [17:39] (note, it is annoying, and something that will get fixed in one way or another). [17:41] smoser: did you happen to figure out how to select a particular dns server on a wifi connection at all times? [17:43] good luck with that [17:43] man dhclient-script [17:43] it would seem its possible there. [17:43] possible == hack [17:43] but that will likely cause issues on public wifi [17:44] its not a hack. its explicitly listed as how to do it. [17:44] oh. or also [17:44] /etc/dhcp/dhclient.conf [17:45] hm.. [17:45] but no, with resolvconf that wont work [17:45] and againa another reason why I hate the change in resolvconf :) [17:46] smoser: but anyuways, any way to disable cloud-init trying to use a mirror? What If I don't really want it to automatically use one [17:46] you can feed it one with user-data. [17:48] smoser: right, but shouldn't default behavior be not to try to do that? [17:48] smoser: cause, in my cause for exmaple, is breaking my maas implementation and might be breaking others too [17:50] smoser: cuase either way, that's the reason why we pass http://pastebin.ubuntu.com/1098696/ to the preseed [17:50] s/to the/in the [17:51] roaksoax, shouldn't the default behavior be to not have broken dns? [17:51] smoser: then I guess we should go all the way to resolveconf :) [17:51] but having this feature off by default would essentially defeat its purpose. [17:52] the idea was to not have to modify the image and get magical selection of your local mirror [17:52] i'm confused though [17:52] oh. [17:52] yeah, that does suck [17:53] you preseed the install, and that gets it right [17:53] and then cloud-init finds a better mirror [17:53] you can fix it through late_command here. [17:53] smoser: right, then I guess we should have an option within MAAS to disable that [17:54] well, this [miss]feature is probably going to change. [17:54] smoser: cuase, for example, in my particular case, how was I supposed to know I had a broken DNS on a clean machine and when I haven't played with DNS? [17:54] i wouldn't have notice at all if it wasn't for what cloud init found [17:54] so that means, like me, there might be others with the same issue [17:55] having mirrors on a MAAS network, but not wanting to use them [17:55] so.. i'm told that the way chrome handles this is that it tries dnslookups of never-will-exist domains. [17:55] and then says "oh, you have broken dns" [17:56] roaksoax, so.. [17:56] cobbler cobbler/password password xcobbler [17:56] is there any reason that i'd pre-seed that in quantal? [17:57] smoser: in pserv.yaml? [17:57] smoser: yeah it gets replaced on postinst [17:57] hm.. [17:57] in zimmer-build [17:57] ah lol [17:57] i used to do that [17:57] to set a password for cobbler [17:58] no not really needed [17:58] maas should take care of settings its own user now [17:58] so should i remove that even if on precise? [17:58] I believe so [17:59] what about updating /var/lib/cobbler/snippets/maas_proxy [17:59] any reason to do that on quantal [17:59] ? [18:00] yeah it should probably stay [18:00] ok... what about /etc/cobbler/cobbler.conf ? [18:00] is that still around? [18:00] smoser: you mean settings? yes still needed until we drop it [18:01] i'm looking at zimmer-build/ud-build.txt [18:01] and there are hacks/modifications for cobbler that are done [18:01] and i was just trying to see if each is still necessary [18:02] smoser: I don't really think we need any since all the setup is done by maas itself [18:02] though we need to ensure that maas set's the correc tproxy, and next_server [18:02] it updates /var/lib/cobbler/snippets/maas_proxy ? [18:03] err... wait. [18:03] is there a way that i can tell maas to use my *other* proxy? [18:03] that is what the proxy hack was doing [18:03] smoser: nope there's not. so that's why we should keep that. [18:03] and making sure next_server and server and MAAS_DEFAULT_ULR are set correctly [18:03] and then the cobbler init hack [18:04] are you looking at this? [18:04] i make cobbler update its network address on start [18:05] smoser: so the cobbler network address update should be kept, + make sure it is updated on /etc/maas/maas_local_settings.py [18:05] cause they should be the same [18:05] so i should just change that hack to read it from /etc/maas/maas_local_settings.py ? [18:06] from DEFAULT_MAAS_URL ? [18:06] smoser: yes, [18:06] that could also be a way to fix this [18:09] PYTHONPATH=/etc/maas python -c 'import maas_local_settings, urlparse; print urlparse.urlsplit(maas_local_settings.DEFAULT_MAAS_URL)[1]' [18:12] smoser: looks good to me [18:14] roaksoax, http://paste.ubuntu.com/1098737/ [18:14] i'm gonna test that [18:15] cool, looks good to me [18:23] roaksoax, wait. [18:23] does maas try to figure that value out on its own? [18:23] or how can i tell it to re-get it. [18:24] the problem is that i build the zimmer image, and then save it. and when it boots it has a new ip address. [18:25] smoser: ah riught, it tries to figure out on postinst, so you need to obtain it the way you were doing it, and update maas_local_settings, and update cobbler [18:25] bugger. [18:31] roaksoax, which of the /etc/init/maas-* files do i need to change so they dont start until MAAS_URL iset correctly? [18:31] smoser: actually that part is run by apache2 [18:31] through the wsgi.py file [18:32] fun. [18:41] roaksoax, so what do you think. [18:41] i get the right address [18:41] and then what? [18:41] do i update /etc/maas/local_settings.py ? [18:42] and dpkg-reconfigure maas ? [18:42] will that do it? [18:42] smoser: dpkg-reconfigure maas updates maas_local_settings.py and cobbler settings [18:42] updates it from what? [18:43] smoser: is a debconf questions that asks for the IP address [18:46] ok. so i'm going to just go with this: [18:46] sudo dpkg-reconfigure -pmedium maas === roaksoax_ is now known as roak === roak is now known as andreserl === andreserl is now known as roaksoax_ === Daviey_ is now known as Daviey === roaksoax_ is now known as roaksoax [20:48] roaksoax, [20:48] in apt-get install of maas (during zimmer build for quantal) [20:48] /tmp/maas.config.12013: 32: /tmp/maas.config.12013: dbc_go: not found [20:48] maas failed to preconfigure, with exit status 127 [20:48] Fetched 62.5 MB in 7s (8324 kB/s) [20:48] Selecting previously unselected package libavahi-common-data:amd64. [20:48] (Reading database ... 46830 files and directories currently installed.) [20:48] smoser: that seems like dbconfig-common is missing or failing or something [20:49] can you depend on that from config? [20:49] smoser: yeah === wgrant_ is now known as wgrant