=== harlowja is now known as harlowja_away === praneshp_ is now known as praneshp === evilissimo|afk is now known as evilissimo [09:28] hi guys, anyone here? :) have some questions about cloud-init [10:34] so, I've specified the EXTERNAL_DHCPC=y thingy in buildroot.config, but it still doesn't execute /sbin/ifupdown [10:34] and calling /sbin/ifupdown directly complains about unknown mappings [10:34] so prhaps I should take the easy route afterall, and just change udhcpc to /sbin/bla :p [10:41] helo, i have instaled cloud-init on my centos machine, and i am stuck now. [10:42] it's some strange configuration with centos on hyperv [10:42] in which way are you stuck :) [10:42] is there any chance i could add cloud-init config file over mounted disk or url ? [10:42] well i dont really know what to do now :) [10:42] i want to specify configuration to run at machine boot [10:42] one time [10:43] i know what i want in cloud-config and how to define it [10:43] you have an amazone-like api machine setup? [10:43] like openstack, cloudstack [10:43] i have default installation from yum packet manager on centos [10:43] so, that would be - no [10:43] ah [10:43] cloud-init is used to configure cloud images [10:43] and basically, that requirs a metadata service in your network [10:44] - the web url you mentioned [10:44] to download the config from [10:44] usualy lives at http://169.254.169.254/ [10:44] ah [10:44] ok, is there any way to conigure metadata service only ? [10:45] i deploy instances in "cloud" manner [10:45] hmm... maybe i could get some layer in front of hyperv ? [10:45] just make sure there is a metadata api accessible at that ip, and make sure your vm is able to reach it :) [10:46] ok, but how to install // configure metadata api ? [10:46] it's part of openstack right ? [10:46] btw, cloud-init can also use configdrive - eg. put the json config on a usb stick (well, image) and add that to the vm [10:46] yea [10:46] or any other cloud platform [10:47] depending on your environment, you could also script something yourself [10:47] afaik, using configdrive would be sufficient for environments that lack a metadata service - like yours [10:48] great [10:49] hm, just don't know since whn cloud-init supports it... [10:49] this was very helpful [10:49] hey my cloud-init process is not able to get user-data file on my aws setup [10:49] why not :) [10:49] 2014-02-12 16:16:13,268 - util.py[DEBUG]: Writing to /var/lib/cloud/instances/i-xxxxxxx/user-data.txt - wb: [384] 0 bytes [10:49] it says 0 bytes downloaded [10:49] indeed it does [10:49] do i need to setup something extra for configdrive? [10:50] readthdocs.org :) [10:50] about it format [10:51] docs.openstack.org/user-guide/content/config-drive.html [10:51] yeah, find it [10:51] great :) [10:52] http://developer.openstack.org/api-ref.htmlext_config_drive.html not working :| [10:52] paresh: you could try curl on the user-data url from within the instance [10:53] http://docs.openstack.org/user-guide/content/enable_config_drive.html [10:53] nvucinic: I think you should be able to use that [14:05] nvucinic, you can use the 'nocloud' datasource, and 'seed_from' to probably accomplish what you want. [14:05] or you could configure on the disk the url to the ec2 mtadata service and then stand something like that up. [14:06] nvucinic, http://bazaar.launchpad.net/~cloud-init-dev/cloud-init/trunk/view/head:/doc/examples/cloud-config-datasources.txt [14:06] that shows how to configure 'NoCloud' to give it a 'seedfrom' [14:07] then just stand something up there with user-data nd meta-data as described in [14:07] http://bazaar.launchpad.net/~cloud-init-dev/cloud-init/trunk/files/head:/doc/examples/seed/ [14:07] you can also pass the seedfrom informatoin on the kernel cmdline if you had easier access to that. [14:09] smoser: great, thank you === evilissimo is now known as evilissimo|afk [15:08] harmw, what complains about unknown mappings ? [15:08] your patch surely seemed sane to me. === evilissimo|afk is now known as evilissimo === evilissimo is now known as evilissimo|afk === evilissimo|afk is now known as evilissimo [16:05] nah, forget it :) [16:05] its almost done, Ill push the last bits tonight === evilissimo is now known as evilissimo|afk === evilissimo|afk is now known as evilissimo === evilissimo is now known as evilissimo|afk === harlowja_away is now known as harlowja [19:37] smoser: https://code.launchpad.net/~harmw/cirros/udhcpc-wrapper [19:38] ive had to build this patchd into my updated buildroot env, since thats the one that builds on CentOS [19:39] though it should compile without issues on ubuntu [19:44] so: https://code.launchpad.net/~harmw/cirros/udhcpc-wrapper/+merge/229114 [19:48] harmw, seems odd you'd have to set that. [19:48] CONFIG_FEATURE_IFUPDOWN_EXTERNAL_DHCP [19:48] since i thought previously we were running ghrough that code [19:48] in order to get udchpc invoked with the options [19:51] there is a big if EXTERNAL_BLABLA around the part that decides which dhcp client to run [19:51] it's N by default [20:00] so how does udchpc run in 0.3.2 ? [20:00] that seems "external" :) [20:02] please check out static int FAST_FUNC dhcp_up(struct interface_defn_t *ifd, execfn *exec) in ifupdown.c [20:02] it ends with: [20:02] return execute("udhcpc " UDHCPC_CMD_OPTIONS " -p /var/run/udhcpc.%iface%.pid " [20:03] and before that function is the stuff that concerns CONFIGURE_FEATURE_IFUPblablabla [20:04] this all revolvs around L575 :) [20:12] ah. ok. [20:12] i believe you [20:12] :) [20:16] harmw, that looks really nice. [20:17] actually.. [20:17] it looks to me like its busted for non-eth0 [20:17] as when busybox calls us, it will set that in the environment i presume [20:17] but then you override it by reaading /etc/default/udchppc [20:17] and also i don tlike the global 'DISABLED' [21:12] you dont? [21:12] and that interface... hm, I suppose I shouldve paid a little more attention to that [21:12] ill just remove that from the config file :) [21:13] and disabled as well, if you dont like it [22:28] ah, so [22:29] commnting between watching netflix isn't very wise either :p [22:29] smoser: udhcpc uses eth0 by default [22:29] so using a config file to make it changable for whatever reason shouldn't be a problem [22:31] so, we can configure on which interface we want it to acquire an address [22:31] and udhdpc will in turn set $interface with the interface it got an adress from/on [22:32] http://git.busybox.net/busybox/tree/networking/udhcp/README.udhcpc?h=1_1_stable&id=179f41778880d0a233435f5c9ed04e17316f479a#n73 [22:34] whats wrong with DISABLED, or is there some other easier way of saying 'no, please move on and leave me without an address!'? [23:47] harmw, hm.. [23:47] i was thinking when someone types: [23:47] ifup eth0 [23:47] and /etc/networking/interfaces says to bring eth0 up as dhcp [23:48] then 'ifup' invokes cirros-dhcpc [23:48] as: [23:48] cirros-dhcpc up [23:48] i assumed it somehow conveyed which interface should be brought up [23:50] harmw, yeah, so our command there (line 33 of https://code.launchpad.net/~harmw/cirros/udhcpc-wrapper/+merge/229114) [23:50] needs to : [23:51] cirros-dhcpc up %iface% [23:51] at least. and can't see why not put in some othe other things avialalbe (like leasetime). [23:52] then, when cirros-dhcpc calls udhcpc it will pass on the provided %iface. [23:52] i suspect then that udhcpc then stuffs that interface into the environment [23:52] and we'll be able to expect it in environment on renew|bound