[06:03] Hi [06:04] could somebody tell me how does cloud-init decide which network interface to configure? === evilissimo|afk is now known as evilissimo [07:36] smoser: ifup $interface [07:36] damn [07:36] ok, Ill make some adjustments [07:37] udhcpc will put the interface where it's gotten an address on in the environment, yes, as $interface [07:38] and disabling would be done in /etc/networking/interfaces, so no need for a DISABLED (which Ill remove) [08:29] harmw: is that related to my question ? [08:32] haha no, sorry :) [08:56] I should've bzr copied the default.script btw, since now it looks like cirros-dhcpc is completely new - which it obviously isn't [08:57] anyway, smoser , the merge request is updated [12:46] msitarz, it doesn't. [12:46] system configuration tells it. [12:47] on ubuntu, it blocks on anything 'auto' in /etc/network/interfaces. [13:17] smoser: so cloud-init does not setup the network [13:17] but the default OS configuration is just setup to use DHCP , that's why I get IP on eth0 ? [13:18] right. [13:18] thats how i would think things should work [13:18] eth0 should be expected to come up on dhcp and give route to network [13:18] now... [13:18] that is being fixedui do want to fix that htough [13:18] so that ocnfig drive can correctly insert networking === evilissimo is now known as evilissimo|afk [16:21] smoser, btw, i'm in NY for a couple week, so will be on/off (forgot if i told u :-P) [16:24] harlowja_at_home, sure. just make sure you vote for my talk. :) [16:24] ?? [16:25] where's that [16:25] http://www.openstack.org/vote-paris/Presentation/ecure-containers-in-openstack-using-lxc-and-user-namespaces [16:25] and while you're there, you should vote for this one too [16:25] https://www.openstack.org/vote-paris/Presentation/astrologer [16:26] cool [16:26] will do, will get some others to vote for that also :-P [16:26] others by canonical at http://insights.ubuntu.com/2014/07/31/voting-begins-for-openstack-summit-sessions-in-paris/ [16:26] look at that second link :) [16:26] if you have links to yours, i'll + them to [16:26] "highly effective astrological love problem solutions are good enough to solve all your problems - See more at: https://www.openstack.org/vote-paris/Presentation/astrologer#sthash.Gu4vFXan.dpuf"? [16:26] what the [16:27] i didn't put any up this time :-P [16:49] smoser, astrogger even put it under the networking category +1, ha [16:49] different type of networking, ha [18:35] smoser: ready to accept my merge request? :P [18:36] let me look. i think we were closer. [18:37] so the patch to busybox [18:37] you patch in a patch. is that right ? [18:37] is thats what happening there? [18:37] where does leasetime come from, do you know ? [18:37] ie, is it of any use for us to get that passed in from ifup ? [18:37] i dont know where it would get it from [18:40] harmw, do you mind if i bikeshed on some things? [18:40] i dont want to put you off, and i thikn its probably good enough. [18:40] but a few things. [18:42] here. i'll just make some changes here [18:42] and then propose for merge into your branch [18:42] then you sanity check [18:46] leasetime comes from the dhcpservice, we don't do anything with it [18:46] hm.. [18:46] i'd have thoguht so [18:46] but why does ifup have '%leastime%' ? [18:47] see line 37 of your MP. [18:47] hm, it does? [18:47] see how it can call dhcpcd with '-l %leasetime' [18:47] it seems odd [18:47] but i didn't understand. so asking you [18:47] :) [18:47] hmk, well, not all clients are using the same options though [18:47] in ifupdown.c [18:49] and yes, its a patch that creates a new (patch) file that will add the new dhcp client directive to that struct in ifupdown.c [18:49] whats so wrong with my stuff for you to b wanting to bikeshed? :p [18:51] give me 10 minutes. i'll push up a branch. [18:51] meh [18:52] just making stuff more localized and such. [18:52] i'll show [18:52] nothing big [18:52] you really did good. [18:52] thank you! [18:52] yea right, well obviously not good enough [18:52] :p [18:55] harmw, what is -x [18:56] as in the udhcpc parameter ? [18:56] yeah [18:56] asking for options [18:56] no [18:56] 'include option OPT in set packets' [18:56] ok. so if hostname is empty [18:56] then we dont want the -x either [18:56] right ? [18:57] damn, well... isn't hostname defaulting to cirros? [18:57] probably. but just in case. [18:57] god I hate you for that :p [18:57] i told you bike shed [18:57] you're probably right [18:57] i've got it thoguh [18:58] probably something like: [ -z `hostname` ] && OPT=bla [19:01] you should merge the updated buildroot as well, someday [19:01] yes. [19:01] i will merge that for sure. [19:02] but lets finish this udhcpc branch first, then you're all set on the buildroot, and then its time to drink beers and enjoy the weekend :> [19:15] harmw, so every time something ifup dhcps [19:15] it will kill resolv.conf [19:15] ie, last one wins [19:16] but thats not any different from the current situation, right? [19:16] i dont know. [19:16] problby not [19:16] just found it interesting [19:16] and odd [19:16] i'm gonna make it only do that if [19:16] domain or dns is set. [19:16] that part was already in default.script, which is called at ifup [19:16] alright. [19:16] i'll lave it as it is [19:16] probably wouln't hurt, no [19:16] we could potentially make that work [19:17] make it configurable [19:17] we could [19:18] i'd opt for first bringing in the major changes, and then perhaps clean this little issue :) [19:18] *fix [19:25] grrr, why doesn't centos7 ship with cloud-init [19:25] I thought RH had added that to RHEL [19:25] and while the releasenotes of centos7 mention there are cloud images underway, they have yet to arrive [19:29] harmw, lp:~smoser/cirros/udhcpc-wrapper [19:29] i'll be back in a bit [19:29] ok, ill have a look at merging that with my branch [19:43] done, pushed [19:45] ok, I'm building a new image now [19:46] hm, I dont see any reason for it to fail on the current buildroot - but I'm using my 2014 buildroot to testdrive this thing... (since that builds on centos) [19:46] ill check back in an hour :) [20:35] and another hour :p [20:35] so, busybox 1.20.1 in buildroot 2014 :| damn [20:35] yeah. thats kindof annoying. [20:36] lets get to a new buildroot [20:36] haha cool [20:36] make it happen [20:36] :) [20:36] i dont have an easy answer for that patch having to have the busybox verison in it [20:36] (that is annoying) [20:36] I had the exact same objection to it [20:36] but went for it nonetheless [20:37] anyway, ill be back in 60 minutes [20:37] k [20:37] so lets shoot to merge to your 2014.02 [20:37] and then merge to 2014.5 [20:37] (and lets not forget udhcpc) [20:38] right. [20:41] holy moley. theres a bunch there. [20:54] harmw, ok. i hope id ont break anything. but heres my plan. [20:54] a.) take the udhcp sutff without the patch disabled in the series file [20:55] so that it will still build fine, but will be inert [20:55] b.) get us merged to a newer version and address that later. [21:03] harmw, ok. pulled iwthout patch. [21:25] uhm, then you shouldn't have deleted default.script right? [21:26] since now we're not applying the patch to bb (to make it default to using cirros-dhcpc) it'll start udhcpc just like it always did [21:27] and that will be done with -s /usr/share/udhcpc/default.script [21:27] which doesn't exists [21:27] so please revert that file untill we will use the patch [21:27] thanks for merging though :) [21:28] be sure to update the chngelog btw