[01:44] roaksoax: don't suppose you're about? [01:44] squashfs is too buggy so I want to remove it [01:49] what sort of bugs ? [01:49] * lifeless is curious [01:56] lifeless: bug 1076409, bug 1074167, bug 1077180 for starters [01:56] Launchpad bug 1076409 in MAAS "maas-import-pxe-files fails when run from trunk" [Critical,Triaged] https://launchpad.net/bugs/1076409 [01:56] Launchpad bug 1074167 in MAAS "maas-import-pxe-files fails to download squashfs ephemeral images." [High,Triaged] https://launchpad.net/bugs/1074167 [01:56] Launchpad bug 1077180 in MAAS "Not importing Quantal boot images" [Critical,Triaged] https://launchpad.net/bugs/1077180 [01:56] lifeless: maas-specific stuff :) [01:57] fun [15:04] So I've been trying to follow the instructions for setting up MaaS from here: https://maas.ubuntu.com/docs/install.html [15:06] And it seems that although the machine acting as both region and cluster controller is quantal (12.10), "maas-import-pxe-files" only grabs precise. That seems a bit unexpected. :-) [15:26] smoser: I'm starting to look at writing a Xen / XCP charm you can use for your OpenStack testing. Is there a KVM/libvirt charm I can use as a sort of template? I don't see one in the "Charm Store", or in the examples that come in the quantal juju package. [15:29] gwd, :-( maas-import-pxe-files is actually busted at the moment. (bug 1074167) [15:29] Launchpad bug 1074167 in MAAS "maas-import-pxe-files fails to download squashfs ephemeral images." [High,Triaged] https://launchpad.net/bugs/1074167 [15:29] i'll walk through and get you a workaround to get quantal [15:29] (it just wont use the squashfs based install, but the end should be the same) [15:30] smoser: I edited /etc/maas/import_pxe_files to include both precise and quantal, that seemed to work OK... [15:30] oh? [15:30] i would have thought it woudl faail if you did that. [15:30] (failing to get the squashfs image) [15:30] but if it works, thats good enough. [15:30] Well as you say, the squashfs / ephermeral stuff fails, but the normal ones seem to be working ok... [15:31] and install works? thats good enough then. [15:32] for "kvm/libvirt charm". the charms that we use to deploy openstack ... i'm looking [15:34] Hmm -- I set up some VMs with a private network, so that I could run dhcp / pxeboot without bothering our IT guys, but a lot of aspects of the maas / juju setup seem to assume this is not the case. For instance, "juju status" is complaining that it can't resolve hostname u1. [15:34] http://bazaar.launchpad.net/~gandelman-a/+junk/juju-deployer/files is the "deployer" [15:35] ping adam_g or jamespage in #ubuntu-server for more info [15:35] i'd have to dig, and they cna probably give you better reference. [15:35] join there, i'll introduce [15:44] smoser: Thanks for the intro. Back to maas -- "juju bootstrap" has produced an installed a juju controller, but "juju status" reports "ERROR Invalid SSH key". Is there a default login that I can use to log into the controller and see what's going on? [15:45] gwd, if you cannot get in to the installed node, then something is busted. :-( [15:45] are you able to see console logs on that system? === matsubara is now known as matsubara-lunch [15:47] smoser: You mean serial output or something like that? No -- unfortunately it's actually a VM. :-/ Is that information available from the maas controller anywhere? [15:48] gwd, wel, you shoudl actually get syslog installer info back to the maas controller [15:49] in /var/log/maas/rsyslog ... or somewhere down there. [15:49] for debugging at this point you need to set a password for the user and just login and then poke around. [15:49] (you could boot with init=/bin/bash or we can modify the preseed to the installer) [15:50] i'm really, really sorry you have an issue. [15:50] you imported your ssh keys ? [15:50] to maas? [15:51] smoser: I'm not too frustrated yet -- it's somewhat different than what I normally work on, so it's somewhat interesting. :-) [15:51] smoser: I did have 2 ssh keys I put into the web UI -- maybe the fact that there was 2 is the problem? [15:53] smoser: MaaS seems a bit confused at my setup as well -- the machines being controlled are on their own private sub-net, and NAT'ed through the controller. I had to do some massaging of the configuration files to get pxeboot to work right, for instance. [15:55] smoser: So I found the rsyslog files for my juju controller on the maas controller -- anything in particular I should be looking for? [15:58] gwd, so the install generally worked? [15:58] the system? [15:58] hm. [15:58] smoser: So it would seem -- I've got a machine (u1) that went through the install process, and gives me a login at the console. [15:59] smoser: And I can see the install logs on the maas controller (hostname maas), but I don't see anything in the logs about installing ssh keys. [16:00] right. [16:00] that happens on first boot. [16:00] i assume something is breaking, and cloud-init on the installed node did not get the metadata from MAAS [16:00] (which is how the keys wouldget pulled in) [16:01] so... we can change the maas server preseed for the install [16:01] so that it creates a user with a password [16:02] Right -- where is that, and how can I trigger a re-install? [16:04] * gwd goes afk for a few minutes [16:11] gwd, juju destroy-environment [16:11] and juju boostrap willd o it [16:14] smoser: OK cool; how about adding a password to the default user? [16:16] riight. [16:16] oh shoot. [16:16] did i not paste that. [16:16] somewhere there is a preseed file.. ( i dont have it in front of me, ,and i'm trying to juggle too many things right now) [16:16] but you want to basically do: [16:16] d-i passwd/root-login boolean false [16:16] d-i passwd/make-user boolean true [16:16] d-i passwd/user-fullname string ubuntu [16:16] d-i passwd/username string ubuntu [16:16] d-i passwd/user-password-crypted password $6$.1eHH0iY$ArGzKX2YeQ3G6U.mlOO3A.NaL22Ewgz8Fi4qqz.Ns7EMKjEJRIW2Pm/TikDptZpuu7I92frytmk5YeL.9fRY4. [16:16] d-i passwd/user-uid string [16:17] (some of that is already there) [16:17] you just want to add the password-crypted field [16:17] and that string is "ubuntu" [16:17] OK -- I'll look around for it. === matsubara-lunch is now known as matsubara === matsubara is now known as matsubara-brb [17:57] smoser: So it appears that there was a race between /home being mounted and the ssh keys being installed. [17:57] separate home partition? [17:57] smoser: The installer creates xvda1 to mount on /, and xvda6 to mount on /home. [17:57] i'll admit i've never considerd that. [17:58] hm.. [17:58] smoser: xvda1 has /home/ubuntu/.ssh with the proper keys in it; but while xvda6 has ubuntu/, it doesn't have any ssh configs. [17:59] gwd, its not a race [17:59] i think you're actually guaranteed failure :) [17:59] * gwd was being optimistic. :-) [18:00] this is interesting. [18:00] the auto partitioner chose to give you a /home. i dont think i'd seen that before. [18:02] Daviey, ^ that is interesting in maas. [18:02] smoser: So a work-around might be to add the following line: d-i partman-auto/choose_recipe select atomic [18:04] i dont know. its wierd to me that it did that. [18:04] so, since hte other wierd thing is xen, i suspect somethign related to xen [18:04] :) [18:15] smoser: Very weird... even with the preseed line I wrote above, it still made a /home partition. :-/ This is all with quantal. [18:17] gwd, well, i think we can work around this. [18:17] but it sucks. [18:17] it will require you to do a late_commadn to edit /etc/cloud/cloud-init.cfg === matsubara-brb is now known as matsubara [18:20] we basically ahve to apply this patch i think [18:20] http://paste.ubuntu.com/1353734/ === pcarrier_ is now known as pcarrier [18:23] smoser: eww [18:24] smoser: i had not seen that before ether.. Is it anew default in debian? [18:24] he's just using quantal [18:26] smoser: right, but i wondered if debian had started doing it, and we are getting it 'for free' as part of d-i? [18:29] smoser,Daviey: Well to compare, my maas controller VM was installed via mini.iso -> network install; so it 1) booted under xen but 2) only created one partition [18:29] Anyway, I've got to head out here -- thanks for your help, smoser. Hopefully tomorrow things will go smoother. :-) [18:29] gwd: right, but wasn't preseeded i am assuming ? [18:29] Daviey: That's right. [18:30] gwd: yeah, i'm wondering if preseeded installs generally have different behaviour [18:30] i don't think this is a maas bug, as such... more of a di one.. but that is nothing more than a hunch, with no facts to indicate it :) [18:32] Daviey: The Ians here on the Xen team know a bit about Debian; I'll ask them if they know anything about partman's behavior here tomorrow. [18:35] gwd: Ian Jackson knows not much about debian :) [18:57] * koolhead17 wakes up === Guest76575 is now known as dpb___