[01:15] hey wallyworld_, did you see my message on Friday about authorized_keys? [01:15] um [01:15] about modifying for the ubuntu user [01:15] if i did i forgot [01:15] i don't think i saw [01:16] wallyworld_: maybe I'm just missing something, but I don't see where the ubuntu user is specified when the worker modifies authorized_keys [01:16] it's not [01:16] shouldn't it be? that's what we log in as [01:17] the current auth keys file with the key from bootstrap doesn't specify the ubuntu user does it? [01:18] so i'm just adding additional keys in the same way [01:18] wallyworld_: I'm pretty sure cloud-init will put it in the ubuntu user's one [01:19] well, it must, since we log in as that during bootstrap [01:19] i assumed that the auth keys would be added to /home/ubuntu/.ssh/... [01:19] i guess it depends what ~ resolves to [01:20] i'm sure i checked though [01:20] and the ubuntu user auth keys file was the one that got updated [01:20] but now i'm doubting myself [01:21] wallyworld_: I didn't see it getting updated, but I did see /root/.ssh/authorized_keys had some updated keys in it [01:21] hmmm. maybe i didn't check then. but i thought i did [01:22] i'll do a fix if necessary [01:22] I'll have another look in a bit [01:22] i could have sworn i checked etc but maybe not [01:22] i guess if the processes are running as root, the ~ will not resolve to ubuntu [01:23] right [01:23] too bad that they are running as root :-( [01:23] I was wondering whether we should be using something less generic than "ubuntu" now [01:23] i sorta assumed perhaps that we would not do something that silly :-( [01:23] like create a "juju" user at init time [01:24] then manual provisioning would be less of a special case regarding key management [01:24] that makes sense. get ready for a whole lotta bikeshedding though :-) [01:24] :~( [01:24] well maybe not, there's always a first time :-) [01:28] so cloud init does explicitly rely on a specific user being set (default ubuntu) [01:28] so looks like i need to do a fix [01:30] okey dokey [01:31] wallyworld_: will things break horribly if the ubuntu user doesn't exist? [01:31] (manual provisioning doesn't yet create it) [01:31] i don't *think* so, but am not sure [01:31] no worries, I can test later [05:24] axw: i fixed the ssh auth keys user issue. turned out to be a bit of a bitch cause of the way our tests override ~. if you get a chance, could you take a look? https://codereview.appspot.com/40650049 [05:25] wallyworld_: cool, thanks, will have a look now [05:25] ta [05:25] i tested live on hp cloud [05:28] wallyworld_: why not use osenv.Home() on Windows too? [05:28] axw: does windows support $HOME? i wasn't sure and didn't think it did [05:29] wallyworld_: just had a look, osenv.Home() does different things for Windows [05:29] not $HOME, but $HOMEDRIVE [05:29] or %HOMEDRIVE% at any rate :) [05:29] but will that return the user home dir [05:29] or the drive letter? [05:29] umm [05:30] wallyworld_: it's %HOMEDRIVE%%HOMEPATH% [05:30] with a slash in between [05:30] either way, our tests won't work on windows as is [05:30] true [05:30] so i think a todo is ok and we can fix the tests wheneever we want to run on windows [05:31] fair enough [05:31] our tests also just assume all paths with ~ are of the form ~/foo [05:47] axw: thanks for looking, i'll fix the little issues [05:47] wallyworld_: nps, thanks for fixing [05:48] including nuking bob :-) [05:48] :) [06:16] axw: that's merged now [06:16] wallyworld_: thanks! [06:16] I'll test it with manual later [06:16] np, thanks for letting me know about the issue [06:17] i'm off for a bit to take the dog for a bath. he rolled around in some shit at the park yesterday and he stinks :-( [07:41] grr, my wireless keeps dying [08:18] try another channel? [08:51] well it recovers if I restart networking, so I don't think it'd be interference === gary_poster|away is now known as gary_poster [16:34] fwereade, are you around? === makyo_ is now known as Makyo === gary_poster is now known as gary_poster|away