[00:11] hi all -- quick question, when anyone gets a minute [00:12] is there a native way to specify that cloud-init configuration baked into an image can override configuration passed as user data? === harlowja is now known as harlowja_away [08:09] smoser: I'm feeding the raw image to glance, which hands it over to my Ceph backend [08:09] come to think of it, isn't Glance able to convert images on the fly these days..? [08:10] anyway, supplying a vhd to serve Hyper-V would be nice [13:06] harmw, glance doesn't (or shouldn't) do it. [13:06] generally thats the philosophy with glance. [13:06] its just a registry. [13:07] hypervisors need to just deal with it. [13:07] and i guess then that that means where nova provisions something "to volume" that it needs to do it too. [13:08] id' happily cahnge to some other format if that format was as easily consumable and producable as qcow and was also sparse and compressed. [13:08] especially if it streamed. wget | convert-to-raw | dd of=block-devic [13:08] that is ideal. i guess s/wget/curl/ [13:09] ah, well [13:09] (so yeah, you touched on somewhat a religous topic for me) [13:09] haha :) [13:09] it is absolutely absurd to expect "image producers" (or appliance producers or whatever) [13:10] to have to offer the same bits in 6 different formats. [13:10] and you (as the user) to have to know which one to get. [13:10] true, though it is conveniant [13:10] the unfair benefit for me of qcow is that it "just works" with kvm. [13:10] which was what I'm after :) [13:10] so that is me being biased a bit. [13:17] harlowja_away, please ping me when in. [13:17] i really need to call 0.7.5 by like end of tomorrow [13:17] so if there is stuff that you think is or should get in, please let me know. [13:17] harmw, same for you above [13:17] or anyone else. [13:18] hm? ah, cloud-init? [13:18] yeah :) [13:18] we also talk about cloud-init here sometimes [13:18] hehe [13:19] well the fbsd stuff is in, and I've not worked on that for quite a while now [13:19] mostly waiting for it to arrive in ports [13:19] and to busy with cirros :p [13:20] again, thanks for your help with that. [13:20] sure np [13:23] https://git.openstack.org/cgit/openstack/cinder/commit/?id=e066158b5235a3879fe90fa3bd813fc3363c01f5 that looks like Glance auto-converting any image type to raw volume [13:24] or it converts the image to raw, making cow volumes possible [13:24] (meh, Ill just have to read the source at some point time) [13:27] smoser: just have Canonical donate me a nice HP Gen8 Microserver and we'll be another step closer to world domination :> [13:28] harmw, NUCs are the new hotness. [13:28] microservers crap [13:29] I'm in for one of those as well :) [13:30] nucs are really neat actually. they also have 'eamt' which gets you vnc to the system and remote power control an dsuch. [13:30] but to our experiments no serial over lan [13:31] so they include some kind of bmc? [13:31] essentially. [13:31] its consumer grade [13:31] but pretty neat. [13:31] sweet, didn't know that [13:31] it shares 1 NIC with the host [13:31] ay [13:31] its actually available on lots of system syou probably didn't know about [13:31] ie, if you have a thinkpad of < 2 years old they all have it. [13:31] I haven't bought hardware in ages, so... :) [13:32] yeah, my thinkpad is 4 years old. it missed the eamt by 1 generation [13:32] :) [13:32] and all servers we have have here come with idrac/ilom or whatever decent thing they have [13:32] *installed [13:32] eamt is really kindofo hookey, but neat. the vnc works by taking the hosts port 5900 [13:32] wtf [13:32] ie, you cannot get to that host's IP address on 5900. [13:32] it shares the IP [13:33] aargh [13:33] thats just sux [13:33] well, consumer [13:33] :) [13:33] true :) [13:33] what do you want for $300 in a hocky puck sized server [13:33] hehe [13:33] 2nd nic and 2nd 2.5" hdd :) [13:33] ideal compute nodes [13:34] atleast in a homelab [13:34] yeah. they're really neat. [13:34] for the money i think they beat the pants off of hp microservers [13:34] at least the onest that we had some of. [13:35] ofc, but the new Gen8 microservers are way better compared to the first 3 generations [13:35] but enough about that :) back to world domination [13:36] when you're gonna branch cirros 0.4? [13:46] btw smoser, how about a little tool in bin/ to change to root password when building from source? [13:47] booooooooo [13:47] this year is our year (cubs) [13:47] perhaps, but here we are rather clueless on just wtf the cubs are and why they should win :> [13:48] as above where i like to selfishly inflict my preferences of kvm on the world, i also like making people type "cubswin:)" [13:48] someone disappointed that "cubswin:)" doesn't show anything about cirros in google [13:49] hehe [13:49] if I created such a tool, would you merge it? [13:50] actually... i've dreamt of "cirros-tools" [13:50] as a package [13:50] err... a separate project. [13:50] fine to start in cirros [13:50] hmk? [13:50] but that do things like: [13:50] cirros-util start lxc [13:50] cirros-util download [13:50] ... [13:51] cirros-util set-passwd cubs-lose [13:51] it'd make testing things easier too. [13:51] hm hm, interesting [13:51] for lxc, though, its in lxc now. 'lxc create -t cirros' (although that there needs a feature to add user-data/meta-data) [13:52] actually... i wonder how close 'backdoor-image' would come to working on cirros. [13:52] i might have even tested it at one point [13:52] (it has code that changes passwords) [13:52] https://code.launchpad.net/~smoser/+junk/backdoor-image [13:53] so that might actually work as it is. [13:53] backdoor-iamge --user cirros cirros-image.img [13:54] (typo) [13:54] what is 'sl' ? [13:54] isn't that the steamlocomotive? [13:54] oh. funny. [13:54] :) [13:54] ubuntu ships with it iirc [13:55] 26k though installed. [13:55] wonder if there is a trim version :) [13:55] :) [13:56] did you ever see the "wheres chuck" meme ? [13:56] http://www.jonobacon.org/2011/11/16/wheres-chuck/ [13:56] i had a ascii art version. that i was going to shove into cirros [13:57] and have it show it if you did the konami code on the console [13:57] i'd love to have some easter egg like that [13:57] damn right :) [13:57] I don't know this specific meme though [13:57] but we have similar memes here :p [14:08] https://launchpad.net/cirros [14:08] harmw, you have almost as many points as i do! [14:09] :) [14:09] hehe cool [14:09] i just branched 0.3. so now there is lp:~cirros-dev/cirros/0.3 [14:09] and lp:~cirros-dev/cirros/trunk [14:09] ah yes, I see [14:09] the second is the target of the ilnk 'lp:cirros' [14:09] and the latest version got updated :) [14:10] and your name is in the changelog [14:10] fame and fortune will come your way soon [14:10] o m g [14:10] your telling ppl to just build the image themselves? [14:10] no. [14:10] i jsut don't knwo what to do. [14:11] as i really dont want to amnually upload stuff to launchpad [14:11] and download.cirros-cloud.net is actually akami CDN'd so it would be faster anyway [14:11] *and* i can more easily get logs of those downloads. [14:11] so i really don't want people looking at launchpad, basically. [14:12] ok, well a link from the source page or something would be nice so ppl know where to look for prebuilt images [14:13] this is true. [14:13] :) [14:13] http://bazaar.launchpad.net/~cirros-dev/cirros/trunk/view/head:/ChangeLog [14:13] especially since cirros-cloud.net redirects you to launchpad [14:13] most epic changelog ever [14:14] having some decent html on cirros-cloud.net would be nice btw, instead of merely redirecting [14:14] my wife suzanne has promised a cirros logo [14:14] and i've wanted to have a shirt for ODS [14:14] lol nice [14:14] i agree on all of this. [14:15] is this your 'whatever, I'm down with everything'-day? [14:16] :) [14:16] https://launchpad.net/cirros [14:16] that should look better now. [14:16] 2 links to download.cirros-cloud.net [14:16] lol [14:21] did you see this: [14:21] https://bugs.launchpad.net/cirros/+bug/1273159 [14:21] obviously you did [14:21] but is this right: [14:21] This can be worked around by adding this line to the eth0 stanza of /etc/network/interfaces [14:21] ? [14:21] i can't see how it is [14:23] but maybe [14:23] https://www.mail-archive.com/busybox@busybox.net/msg03985.html [14:25] hm, if thats right it would require the addition of -O staticroutes to go in there as well [14:25] yeah. [14:25] can't verify that right now though [14:25] well, source code makes it seem unlikel [14:25] $ grep -r nodefaultopts . [14:26] shows nothing [14:26] in busybox git [14:26] hm, I believe I've seen that -o somewhere though [14:26] to not ask for default options [14:27] plus, it's no_default_options [14:30] wrt the retries.... on the metadata service [14:30] there were issues on ec2 [14:30] where the metadata servie woudn't come up right away. [14:30] crazy stupid. [14:30] hmk [14:31] but when our ubuntu images first got there we were booting and hitting it before it was up [14:31] and we'd just say "nothing there!" [14:31] so ... poll and retry :-( [14:31] hmk, so 20 retries made sense? [14:31] i think they're probably much better now. [14:31] well, 20 retries covers i think 60 seconds? [14:31] somethin glike that. i think. [14:34] ah. itmeout is 10 seconds . on the curl request [14:34] so it could be up to 20*10 + 20*2 (the nap length) [14:35] yea well, it's a pita having to wait 5 minutes because it takes to long to acquire an ip and trying to contact a non-existant ec2 api :p [14:37] that is kind of silly. we can probably make it do max of 60 seconds. [14:37] it is unreasonable in the first place for the MD to not be there. [14:38] depends on the env, when I'm testing cirros I certainly don't do that in my openstack setup [14:38] but just with qemu, or hyper-v [14:39] no ec2 api's on either of those [14:53] right. that is reasonable. [14:53] so i'd like to have a (non-root) way to boot the instance with metadata [14:54] ie, like ubuntu images do (http://ubuntu-smoser.blogspot.com/2013/02/using-ubuntu-cloud-images-without-cloud.html) [15:03] ok, so when there is no ec2 available it firesup a local ds and uses that [15:05] well, fires up.. it just reads in the user-data file which was already there [15:05] perhaps the existence of such a file could (should?) make it skip contacting ec2 in the first place [15:06] yeah, thats what it does harm. [15:06] and you can do that with 0.3.2. [15:06] if you seed those directorries, then it will read from them. [15:07] cool [15:07] then we realy need a proper seed-tool :) [15:10] agreed. [15:10] the thing that sucks is root [15:10] oh. actually, kyou could still attach a config-drive disk. [15:10] and that should work. [15:10] and maybe i did it to allow a "nocloud" [15:10] maybe [15:11] yeah, actually it should [15:11] harmw, you can just attach a nocloud disk. [15:12] to much hassle :> [15:12] to avoid root [15:12] its worth it. :) [15:13] hehe ,well, I'm cool with using sudo to manipulate images... but yes, have a root-less solution is cool as well [15:13] one other thing you could do in a raw image is have some place in the disk that a tool could just edit straight away [15:13] bu that doesn't work with qcow or any other format [15:13] hehe, raw +1 :> [15:13] and you probably only get like 512 bytes there (this is how grub does things for its 'environment' file) [15:14] well, it knows how to read extX filessystem, but then to just write into those places. [15:14] might be just enough for nocloud [15:14] wll, i dont know. limits user-data. [15:14] one thing that is very useful, and tests a lot of it is just lxc [15:14] true, but normal users probably have ec2 for that [15:14] and with 14.04 you can use lxc without root [15:15] but we'd need to make cirros work well there. [15:15] fair enough [15:15] its not too much hassle to attach a disk. really. [15:15] and if you have: [15:15] no ofc not :) [15:15] cirros boot kvm --user-data=foo [15:15] then, you dont know of such hassle. [15:15] so true [15:15] but what if I'm testdriving in a hyper-v vm [15:15] which is kind of a pain already [15:16] well, then yo uhave to download the install shield installer [15:16] and click yes-i-accept [15:16] and then get some DLLs from google some where [15:16] and then reboot [15:16] and *then* you can do it. [15:16] ah yes, the one that requires .net4.5 and which isn't supported on a hyper-v server running Windows Server Core [15:16] funny you [15:16] cirros.msi :> [15:17] windows is *so* well designed for automation [15:17] spare me... [15:17] * harmw manages several dozens of Windows systems [15:17] someone pointed me at this a few days ago. [15:17] haven't read it all [15:17] http://www.thoughtworks.com/insights/blog/cloud-based-devops-possible-windows [15:20] "However, it isn’t all sunshine and lollipops. WinRM is actually pretty painful and fiddly to use and PowerShell is an ugly and procedural language." [15:20] so true === harlowja_away is now known as harlowja [17:12] smoser hey, just got in [17:12] lazy west coast people [17:12] utlemming, did you see my comments ? [17:12] haha [17:13] smoser one that would be nice @ https://code.launchpad.net/~harlowja/cloud-init/local-before-net/+merge/211783 [17:14] to fix the issue where cloud-init-local starts after networking [17:14] ugh. [17:14] well, that only fixes in sysvinit [17:14] (not rhel6, right?) [17:14] *and* that would then differ from other distros [17:15] sure, idk the ordering of other distros, since systemd and ordering isn't so easy to figure out [17:16] right. i dont think its guanrteed [17:18] sure, so don't the other files also need to have this start before networking? [17:18] *this == cloud-init-local [17:21] harlowja, ? [17:21] i dont understand [17:22] guess the question is should the other files be adjusted also? [17:23] harlowja, theres no way to do it in ubuntu. [17:23] i dont think [17:23] well, at least not more invasive than i'd lke to go at this point. [17:25] k, so i guess then maybe systemd needs to be adjusted, in the rhel5/6 that i'm y! using we are using those sysvinit scripts [17:26] brb [17:36] ok back [17:38] * harlowja had to do apple security update crap [17:38] anyways smoser we can debate that later, i'm fine with the next release afaik [17:39] be nice to have sean here push the freesbsd stuff, but i can't seem to find him [17:39] oh, i found him online [17:39] lol [17:44] but he's not responding, sad face [17:44] lol [17:58] harmw so sean is currently sucked back into the mail vortex :( [17:58] haha lol [17:58] maybe more blackhole [18:05] harmw bugging him about maybe when he'll have some time free from mail [18:13] he says maybe end of this week :-/ [18:15] ok :) === harlowja is now known as harlowja_away === harlowja_away is now known as harlowja === harlowja is now known as harlowja_away === harlowja_away is now known as harlowja