2014/10/08/#cloud-init.txt

=== Guest69465 is now known as mfisch
=== mfisch is now known as Guest67039
=== harlowja is now known as harlowja_away
[10:59] <harmw> smoser: ping
[13:00] <smoser> harmw, here. whats up?
[13:00] <harmw> I'm working on some code to get the proper fbsd interfacename
[13:01] <harmw> should be ready soon
[13:01] <harmw> can you change the instance hostname in AWS?
[13:28] <harmw> btw, don't nuke that instance since I'm debugging on it :p
[13:35] <smoser> harmw, you can't change a hostname in aws
[13:35] <smoser> i can give you another instance if that'd help.
[13:38] <harmw> hmk
[13:38] <harmw> so ip-10-0-0-158 is the current hostname?
[13:38] <harmw> nah, I think that'll work
[13:43] <smoser> harmw, wlel, http://169.254.169.254/latest/meta-data/local-hostname would tell you
[13:43] <smoser> but you seem to have no sane tools on that system (curl or wget)
[13:43] <harmw> fetch :)
[13:43] <smoser> i remember a time before ubuntu whne i liked to play with sharp pointy objects.
[13:43] <harmw> :>
[13:43] <smoser> rather than things that worked ;)
[13:50] <harmw> smoser: https://code.launchpad.net/~harmw/cloud-init/amazon-fbsd-fixes/+merge/237595
[13:51] <harmw> come to think of it, it's only relevant for configdrive setups since that's the only case where the nic is even used :p
[13:51] <smoser> hooray!
[13:52] <harmw> but nonetheless, it's always better then just using 'some' hardcoded name :p
[13:53] <smoser> harmw, oh, cause it reads that in interfaces.
[13:53] <smoser> yeah, thats all slated to change
[13:53] <smoser> where we'll get info on the mac and find it that way
[13:54] <harmw> yea, it should indeed be some higherlevel solution
[13:54] <harmw> but for now this'll do the trick
[13:54] <harmw> if only hiren_ would wake up
[13:54] <harmw> since he actually uses configdrive
[13:55] <smoser> harmw, the only coments i have on that merge proposal is that you should in general re-use 're.'
[13:55] <smoser> with re.compile
[13:56] <smoser> but its fine.
[13:56] <smoser> ie:
[13:56] <smoser> for line in ...
[13:56] <smoser>   m = re.match('^\w+', line)
[13:56] <smoser> is going to re-compile that regex
[13:56] <smoser> where it could have used a cached version
[14:03] <harmw> so -always- compiled regexes?
[14:04] <harmw> ah, it's inside that for loop
[14:04] <harmw> damn
[14:05] <smoser> i dont know really about a good rule for when to do it and hwen not to
[14:05] <smoser> but clearly in that case on the loop, you're wasting cpu. 
[14:06] <smoser> dont worry about it.
[14:06] <harmw> well in this case I honestly overlooked my own loop
[14:06] <harmw> stupid though
[14:07] <harmw> else I would've used a compiled regex, since you told that on some other review a while back :)
[14:37] <jaxxstorm> is this the right place to discuss cloud-init bugs?
[14:38] <harmw> jaxxstorm: yes
[14:38] <harmw> bring it on
[14:39] <jaxxstorm> I'm using the phone_home module and it's failing because of SSL, but my URL is http:// and on port 80
[14:39] <jaxxstorm> output:
[14:40] <jaxxstorm> https://gist.github.com/jaxxstorm/0b2f7b20718bcb462fc6
[14:40] <harmw> could you use paste.ubuntu.com or something please :)
[14:41] <harmw> anyway, it says this is the url: http://foreman.hostname.com80/unattended/
[14:41] <jaxxstorm> yeah I've redacted the URL
[14:41] <jaxxstorm> but basically it's http://foreman.companyurl.com:80/unattended
[14:41] <harmw> ah, and that 80 is deliberatly without prepending :?
[14:41] <jaxxstorm> yeah that comes from the foreman
[14:42] <harmw> UrlError: [Errno 1] _ssl.c:492: error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify faile
[14:42] <harmw> you're using a selfsigned cert?
[14:43] <jaxxstorm> well my foreman instance does have a self signed cert
[14:43] <harmw> exactly
[14:43] <jaxxstorm> but it's also available on port 80 without SSL for callbacks
[14:43] <jaxxstorm> so why would it even attempt to use SSL on port 80 with http?
[14:43] <harmw> isn't foreman just redirecting (forcing) to use ssl?
[14:44] <harmw> c-i won't just start SSL by itself 
[14:45] <jaxxstorm> if I call the exact same URL with curl it doesn't use SSL
[14:45] <jaxxstorm> and doesn't redirect
[14:47] <harmw> weird
[14:48] <jaxxstorm> yeah
[14:48] <jaxxstorm> in any case, is there a way to have it not verify the SSL cert?
[14:48] <harmw> that's something smoser would know, though I think it is possible to import a custom CA
[14:48] <harmw> which would make c-i accept the certificate aswell
[14:49] <jaxxstorm> yeah I saw that module and I may go dow that route if needed
[14:49] <harmw> it would be the most elegant solution :)
[14:52] <harmw> smoser: could you accept the merger?
[14:55] <smoser> merged. thanks.
[14:55] <harmw> sweeeet
[14:56] <jaxxstorm> okay, never mind, sorry guys
[14:56] <jaxxstorm> < Status: 302 Found
[14:56] <jaxxstorm> getting redirected to https
[14:56] <jaxxstorm> :(
[14:56] <harmw> I told you :P
[14:56] <harmw> ah, you forgot curl -l
[14:56] <harmw> *probably
[14:56] <harmw> or actualy.. ah nvm
[14:57] <jaxxstorm> I guess I can add the CA cert, but that'll be a pain
[14:59] <smoser> jaxxstorm, you should probably be able to insert the ca cert in cloudconfig
[15:00] <jaxxstorm> yeah I think I can
[15:00] <jaxxstorm> does it rely on a package being present for CentOS?
[15:00] <jaxxstorm> how does it add the cert?
[15:00] <jaxxstorm> I want to make sure all the required packages are installed before I attempt it
[15:00] <smoser> i think it shoudl all be contained in cloud-init (and usage of python-requests)
[15:01] <harmw> cloudinit/config/cc_ca_certs.py: distros = ['ubuntu', 'debian']
[15:01] <harmw> would that mean it's not supported on RHEL?
[15:03] <jaxxstorm> it seems that way...
[15:03] <jaxxstorm> :(
[15:04] <harmw> jaxxstorm: open up a bugreport, this is something worth adding I'd say
[15:04] <jaxxstorm> would love to, where do I go?
[15:04] <harmw> https://bugs.launchpad.net/cloud-init
[15:06] <jaxxstorm> done
[15:06] <jaxxstorm> https://bugs.launchpad.net/cloud-init/+bug/1378874
[15:10] <harmw> that module is currently using some Debian-specific stuff, but if the logic applies to RHEL it should be fairly easy to get it to support that
[15:11] <jaxxstorm> thanks :)
[15:12] <jaxxstorm> I'm cheating anyway
[15:12] <jaxxstorm> runcmd: /usr/bin/curl -k <%= foreman_url('built') %> 
[15:13] <jaxxstorm> btw, the app is amazing, has really helped my provisioning process, so thanks!
[15:14] <harmw> if foreman is redirecting you, you need a curl -L to follow the 30x
[15:18] <jaxxstorm> ah, v. true
[15:36] <smoser> jaxxstorm, hm.. i dont knwo. i'm not sure how specific that would be. 
[15:55] <smoser> harmw, harlowja_away https://github.com/cloud-init 
[15:55] <smoser> now to get source there.
[16:54] <harmw> why?
[16:54] <harmw> smoser: what's wrong with LP? 
[17:08] <harmw> btw, you may nuke that instance now
[17:10] <smoser> harmw, ok. i'll kill instance
=== Guest67039 is now known as mfisch
[17:10] <smoser> nothing wrong with lp, but want to have something on github for people to find it, and or to contribute. there. some people just more comfortable there.
[17:11] <harmw> sounds legit
=== mfisch is now known as Guest72593
[17:12] <JayF> smoser: is the copy in git or bzr going to be canonical?
[17:12] <JayF> I guess I could pick a better word
[17:13] <harmw> :p
[17:13] <JayF> I mean lowercase "c" canonical not uppercase "C" ubuntu
[17:13] <JayF> :P
[17:14] <smoser> :)
[17:14] <smoser> JayF, i dont know. not thought of that yet.
[17:14] <JayF> That's the most important thing to know, isn't it?
[17:15] <JayF> split brain problem and all that
[17:18] <smoser> yeah, its clearly important to know.
[17:39] <harmw> +1 for keeping LP master :p
[17:50] <JayF> +90001 for making it github
[17:50] <JayF> you make it github and all the sudden you'll see more contributions from me
[17:50] <JayF> that's for sure
=== harlowja_away is now known as harlowja
[18:15] <harlowja> holy crap, github, woah
[18:15] <harlowja> although the PR mechanism sux still 
[18:15] <harlowja> *imho
[18:52] <gholms> On, the bright side, it means not having to use bzr.
[18:53] <kwadronaut> heh
[18:53] <kwadronaut> only the maintainer(s) have to take care of keeping everything in sync
[18:55] <gholms> Yep
[18:57] <kwadronaut> for git-git you can use hooks or send a message to github asking to mirror it (which means they'll pull in changes twice a day or so and it's more obvious for users that it's not the canonical one)
[18:58] <kwadronaut> there's both bzr-git and git-bzr fwiw. besides that, i dislike both platforms.
[18:58] <kwadronaut> (if we're shedding bikes, i'm taking part and now i'll run again)
[19:02] <JayF> I paint my bikeshed git of any kind
[19:02] <JayF> smoser: another option would be to put cloud-init in stackforge/ and use openstack processes
[19:03] <JayF> smoser: I have no idea how you'd feel about that, but you'd get some quantity of testing infra for "free"
[19:03] <JayF> smoser: plus you'd be more tightly integrated with one of your biggest users
[19:41] <harmw> stackforge sounds cool
[19:49] <smoser> JayF, this is true, and i've considered that before. but i'm not tied to openstack, and do not really want to be. 
[19:51] <harlowja> i'd have to lean towards not also, i've got about 3 stackforge projects and an oslo one, so i've been through it, i'm not sure its buying much for cloud-init that github + travis couldn't buy (i do personally like the gerrit that openstack has over pull requests, but this is a managable dislike)
[19:52] <harlowja> although the stackforge one could have an interesting connection into the rest of openstack if that stackforge repo somehow got plugged into the image-building
=== Guest72593 is now known as mfisch
[19:53] <harlowja> would be nice to have a new image-built when cloud-init review happens, and tested accordingly, thats like the true functional test that would be nice to have automated
=== mfisch is now known as Guest6813
[19:54] <harlowja> but if said thing doesn't exist ( JayF might know more about the image building stuffs that could be possible) then i'm not sure it matters, github, stackforge...
=== Guest6813 is now known as mfisch
[19:55] <harlowja> i've had https://github.com/yahoo/Zake working with travis and for a project without ton of activity its worked out ok
[20:02] <smoser> harlowja,  i want an easy "patch image with cloud-init" for sure.
[20:02] <smoser> but i dont think building the image from scratch is useful.
[20:02] <harlowja> well scratch or whatever other mechanism that i believe the tripleo folks/ironic people are using
[20:02] <smoser> and 'mount-image-callback' or libguestfs makes patching it in trivial.
[20:02] <harlowja> sure
[20:03] <smoser> http://ubuntu-smoser.blogspot.com/2014/08/mount-image-callback-easily-modify.html
[20:03] <harlowja> be nice to have whatever that is part of the review 'check' pipeline
[20:03] <smoser> this is ture, but openstack would specifically be better off using a "stable" cloud-init for testing (ie, grabbing latest ubuntu image).
[20:03] <smoser> and probably even latest image of stable release.
[20:04] <harlowja> :)
[20:04] <harlowja> so the oslo program has done something sorta neat for that
[20:04] <harlowja> it has a check pipeline that checks the existing  (stable) and from source 
[20:04] <harlowja> so u can get early warnings from the source 'check' and still not affect users of 'stable'
[20:05] <harlowja> ex @ https://review.openstack.org/#/c/126791/ (the gate-*-src-taskflow are testing this source version)
[20:05] <harlowja> while afaik the other users of taskflow are not using that source version (but the last release thats avaialble)
[20:07] <harlowja> so this could do something similar
[20:14] <smoser> hm..
[20:14] <smoser> so that would seem like a 'nice to have'
[20:14] <smoser> but stackforge would not seem like a requirement for that.
[20:15] <smoser> in that they have "sources" that live other places.
[20:15] <smoser> such as cloud-images.ubuntu.com or fedora.mirror.org/foo
[21:03] <harlowja> smoser agreed
[21:04] <harlowja> its not a requirement, but could be simpler to make happen (emphasis on could)
[21:04] <harlowja> since we already have people here in this channel that work with openstack and the infra there
[21:26] <harmw> openstack? what's that
[21:26] <harmw> sounds scary
[21:32] <JayF> I suspect if you moved cloud-init into stackforge/ you could easily convince people to help you get some of this testing working
[21:32] <JayF> because of the large number of crossover between openstack devs/opers and cloud-init users
[21:32] <JayF> I'm not saying *I* would (I owe this project other work already), but I'm saying people would :)
=== shardy is now known as shardy_z
[21:44] <harlowja> JayF so that means u would?
[21:44] <harlowja> lol