[00:00] I'm using the maas.img.bz2 for that node, because I did not want to mess with the DHCP [00:01] the status of the node in MAAS (I added it by hand with the right MAC and name) is: Commissioning [00:02] what I tried first is to boot from the 12.10 server CD and choose MAAS, use MAAS-server X (so the Avahi/multicast works) [00:02] but all it does it reboot [00:02] after you choose that option [00:03] any tips ? [00:03] Lennie_: it's not really tested without dhcp, why don't you want to change that? [00:04] Ahh, ok. I see [00:04] Well, I didn't want to change it, because I'm running this in a virtual environment but I thought it would still need Internet access ? [00:05] but I guess it doesn't. [00:05] the nodes need internet access to complete installation [00:05] this is probably what is going wrong for you [00:05] the master did install and download all the needed stuff [00:05] it does have internet [00:06] what do you mean by master here? [00:06] ok, I at least now know that without DHCP-changes it won't work. I'll switch stuff around [00:06] thanks [00:06] master=MAAS server [00:06] are you using the maas packages from 12.10? [00:06] yes I am [00:08] so the nodes need internet access to [00:08] too* [00:08] ahh, would that be different with 10.04 ? [00:08] I can just do a reinstall [00:09] no, it has always been the case [00:09] hmm.. ok, I see [00:09] when they commission/enlist, they download some extra packages from the archive [00:11] I just remembered I don't need to mess with DHCP, because it is virtual I can just tell iPXE to use next-server and pxelinux.0 [00:12] I have some doubts about it working, but I'll try anyway :- [00:13] :-) [00:13] ok :) [00:14] FWIW there's quite a few bugs we're fixing in 12.10, you can see the list here. It should be released to the archive soon. https://launchpad.net/maas/+milestone/12.10-stabilization [00:14] see I did try that before, what I get in that case is: Trying to load: pxelinux.cfg/some-long-number [00:15] and to be clear after that, nothing happends [00:15] right, it's trying to TFTP [00:16] the number is the MAC address [00:16] maas runs its own tftp server [00:16] the dhcp server tells the pxe boot where to tftp from [00:16] "next-server" [00:16] Yes I know and it is the same MAC address as before [00:16] but I expect you know all this :) [00:17] pxelinux.0 does get downloaded [00:17] why does it not download more ? [00:20] ahh, I found the bug...? it does not try to download the pexlinux.cfg/adsfsdf from the next-server, but the DHCP server [00:21] and that was after downloading the pxelinux.0 from the next-server :-) [00:22] ok, maybe it is because I'm faking it [00:22] probably [00:23] if next-server is not present, it defaults to the same address as the dhcp server [00:26] yep, that was the problem [00:32] we have a PPA with fixes for a lot of those problems if you want to test with it [00:34] euh... no ? I spend already 2 or 3 hours before I found out that the libvirt/NAT on my desktop where the virtual machines are running translates the multicast traffic to 224.something from Avahi to 10.0.0.3 (the address on the lxcbr0) [00:34] and I can't figure out why by the way [00:34] if you have an idea why I'd love to know :-) [00:35] it isn't a rule in iptables or iptables -t nat [00:35] or -t mangle ;-) [00:36] ok, the node failed to install: unmet dependencies: kernel package: linux-server [00:37] you need to use bridged networking [00:38] The NAT is a bridge ;-) anyway, I disabled network manager, created a bridge and attached eth0 and put my IP on the bridge and told libvirt to connect it to the bridge [00:38] NAT is not bridged :) [00:39] are you intending to use MAAS with real machines at some point? [00:40] well, maybe, probably. If I like it and I can convince my boss and coworkers at work :-) [00:43] what I wanted to do is play with OpenStack and Ceph, but the Ceph charm needed block devices, so I needed a way to automate deployment. So I'm doing a virtual bare-metal for deploying charms [00:45] ok [00:45] I was going to say that if you really wanted VMs for deployment, you don't really want to use MAAS [00:45] feel free to ping me directly if you want more help [00:47] have to admint it is all a bit confusing, I only started playing with OpenStack a couple of days ago and I know Ceph already for a couple of months, been following it's progress for more than a year. Now I'm adding juju, charms and MAAS. [00:47] maybe it is just to much in one go ;-) [00:48] there is a lot to learn, yes! [00:48] but if you take it one step at a time, it's not a problem [00:48] you won't get to learn it quickly though [00:49] and I wanted the OpenStack with Quantum and Ceph thin provisioning, I can tell you you need there are some corner cases there that don't work ;-) [00:49] and trying to get them all right in one go was a bad idea :-) [00:49] yes :) [00:51] hey roaksoax, did you get the dhcpd apparmor fix SRUed? [00:51] bigjools: is the install of 12.04 of MAAS less buggy or more buggy than 12.10 ? [00:51] Lennie_: vastly more [00:52] we're replacing 12.04 with what's in 12.10 soon [00:52] Ok, because I'm still stuck with installing the first node [00:52] so the bugs will be equal :) [00:52] yeah, it's not really tested on VMs so I am not surprised [00:53] is this VM related ?: "ok, the node failed to install: unmet dependencies: kernel package: linux-server" [00:53] it says in the title: "Congiruing linux-image-3.5.0-18-generic" [00:53] honest, not sure. is that in the console log on the node? [00:54] in the console yes [00:55] in the log, well console 4: unmet depend: linux-header-3.5.0-18-generic depends linux-headers-3.5.0-18 but it is not going to installed [00:56] and: unmet depend: linux-image-generic depends linux-image-extra-3.5.0-18 but it is not going to installed [00:58] I tried to chroot in /target and run apt-get -f install and it says: no space left on device [00:59] it seems it created a partition of 557MB [00:59] with over 500MB used [01:04] hmm, you can't force a MAAS installation to do something else than what it is programmed to do, when you are at the console, can you ? [01:05] not even by going to the menu when you have an error and choosing something [01:06] let's see what happends with I change it from virtio disk to ide [01:08] better [01:08] now it uses most of the disk [01:13] sorry, on the phone [01:14] bigjools: no worries, it's starting to work now :-0 [01:27] good to hear [01:33] bigjools: is this still true for juju ?: You'll want to make sure default-series is set to the Ubuntu Server release you want. (Only precise is supported for now.) [01:34] Lennie_: no, it can deploy quantal [01:36] because at this point I get: INFO Bootstrapping environment 'maas' (origin: distro type: maas)..." [01:36] "ERROR No matching node is available." [01:37] you have one of the bugs that's not fixed in the archive yet :( [01:37] it's a juju bug [01:37] there's a workaround, hang on [01:39] I already used juju 0.6 from ppa to be save, but I guess that wasn't enough ;-) [01:39] oh, hmm [01:40] try: juju bootstrap --constraints arch=i386 (or amd64 if that's what you have) [01:41] otherwise, check your maas console, you do have at least one node that's marked "ready" right? [01:41] yes, that is what I did wrong, I allocated them :-) [01:41] ah [01:41] it's all about learning the lingo :-) [01:41] you should not do that in the UI, juju does it for you [01:42] you have to use the maas-cli tool to unallocate them [01:42] yeah, that is what I understand now too [01:45] ok, that works, they are now ready, let's see if I can bootstrap them [01:46] bootstrap juju I mean [01:46] it installs the nodes using d-i, so takes a while [01:47] I should probably first attacht the extra disk [01:47] quicker if your squid-deb-proxy is populated [02:19] hmm... I don't understand it, why do I get this error: 'ERROR Error processing 'cs:quantal/ceph': entry not found' at this url http://jujucharms.com/charms/quantal/ceph there is one, but not here ?: https://store.juju.ubuntu.com/charm-info?charms=cs:quantal/ceph [02:23] → #juju [02:23] good idea :-) [02:47] while no1 mentioned it in #juju: I can probably use a branch instead: https://juju.ubuntu.com/docs/internals/charm-store.html#namespaces [03:20] bigjools: could the webinterface also automatically refresh every x amount of time ? (as in my case I have to manually start the vm's) [03:20] Lennie_: the dashboard page does that [03:22] that is the first page, not the nodes-page I assume. OK, that would be fine [12:01] Maas doesn't install node | http://askubuntu.com/q/222531 === matsubara is now known as matsubara-lunch === matsubara-lunch is now known as matsubara === cmagina_ is now known as cmagina [18:43] Hello [18:44] Anyone here? [18:47] I'm having a problem with commissioning nodes. [18:58] when I go into the MAAS web gui and click accept and commission, then I boot the node from the network it starts but at the login prompt it throws erros to the console 'http 401' and posts a lot of additional errors (which I cannot see b/c they go by too fast) and then shuts down. [20:31] hi