21:00:13 <gholms> #startmeeting Cloud SIG (09 Sep 2010)
21:00:13 <zodbot> Meeting started Thu Sep  9 21:00:13 2010 UTC.  The chair is gholms. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:00:13 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
21:00:18 <gholms> #meetingname cloud
21:00:18 <zodbot> The meeting name has been set to 'cloud'
21:00:24 <gholms> #topic Init process
21:00:30 <gholms> Who's here today?
21:00:34 * jsmith is here
21:00:40 * obino is here
21:00:44 <gholms> #info rbergeron is unable to attend today
21:00:46 <jsmith> rbergeron is on a plane to OLF
21:01:17 <gholms> jsmith is here!
21:01:25 <gholms> #topic EC2 image status
21:01:27 * jforbes is roughly here
21:01:35 <gholms> How goes image building?
21:02:26 <jforbes> It went well, I had to make some changes to my login bits, but got sidetracked with kvm stuff.   I should be able to get back to them tomorrow
21:02:42 <gholms> #chair jforbes
21:02:42 <zodbot> Current chairs: gholms jforbes
21:03:04 <brianlamere> here
21:03:09 * jforbes looks over last weeks minutes
21:03:29 <jsmith> jforbes: That's good news... mind if I ask what changes?
21:04:21 <jforbes> jsmith: I messed up the ssh authorized_keys response
21:04:25 <jforbes> for root
21:04:59 <jsmith> Ah...
21:05:20 <gholms> What are you using for login bits right now?
21:05:31 <brianlamere> jforbes:  I got an email from Ben (the aws os contact) about a patch that "significantly increases EBS volume performance" - patch is based on 2.6.35, and he's curious how it best to get it in.  Shall I just point him at you?
21:05:36 <jforbes> trivial bits, but it meant respin and I didn't get back to them
21:05:46 <jforbes> brianlamere: yes
21:06:23 <jforbes> gholms: right now it is all in post manually modifying bits, that needs to change (and I think someone discussed making a package to handle most of that)
21:06:55 <gholms> Yeah, we talked about working with Canonical's cloud-init scripts.
21:07:04 <brianlamere> jforbes:  aye, that should be a package.  Someone somewhere should pipe up for that
21:07:35 <jforbes> brianlamere: I thought someone did, but I could be wring there
21:07:42 <gholms> I looked over it all last week.  We can use about half of its scripts wholesale, but the rest will need modification to make any sense on Fedora.
21:07:47 <brianlamere> note that the cloud-init scripts are pretty basic; if they don't do exactly what is wanted, it's easy enough to just spin new stuff with that as inspiration
21:08:10 <gholms> If we can factor out the package management code that will fix the biggest hurdle right there.
21:08:15 <brianlamere> I think several people suggested it should be, I just don't recall who said they'd be the person
21:09:11 <brianlamere> gholms: well I'm still waitin on rsc to need boto help; if you want an additional set of eyes on that, lemme know ;)
21:09:48 <gholms> I'm involved with a bunch of wedding prep stuff this weekend, but I might be able to look at cloud-init next week.  They really have a good framework, and its upstream maintainer is willing to take patches that help with portability.
21:10:25 <obino> brian: I heard boto help. do you want me to put you in contact with Mitch?
21:11:24 <gholms> brianlamere: ^
21:11:39 <brianlamere> Obino: oh hey!  was speaking more of the python-boto package, but it may be that rsc can find better synergy with you all
21:12:08 <gholms> obino: In case you don't already know, rsc is the packager for boto here.  ;)
21:12:16 <obino> got it: thanks
21:13:22 <gholms> brianlamere: If you're feeling adventurous you could always look over the cloud-init scripts and libraries.  Maybe we can figure out how best to make use of them and adapt them to our needs.
21:15:38 <gholms> #action brianlamere to forward an EC2 kernel performance patch to jforbes for upstream submission
21:15:56 <brianlamere> certainly, I sortof did that to some degree myself for my personal images anyway.  Do we have a spec that ultimately describes what our needs are?  Is that still very much a wip?
21:16:45 <brianlamere> jforbes: just sent Ben's email to you, it's traveling internet-speed as we type
21:16:54 <jforbes> brianlamere: thansk
21:16:58 <jforbes> thanks even :)
21:17:38 <gholms> The biggest things their scripts do are fetch SSH keys and update packages.  We can use the former nearly unmodified, but the other should really be broken out into a small python class that calls apt or the yum api.
21:18:16 <skvidal> apt api?
21:18:30 * skvidal reads context
21:18:33 <gholms> Do they have a python api?  Right now they just shell out.
21:18:54 <skvidal> ah
21:18:58 <skvidal> they have python bindings
21:19:03 <skvidal> hopefulyl we're not using apt on fedora
21:19:07 <skvidal> b/c that makes me sad
21:19:31 <gholms> Nah, we need to break that out into some python classes that handle things in platform-appropriate manners.
21:19:49 <jforbes> apt on fedora would be sad indeed
21:19:58 <brianlamere> I think his interest was more in just making a call to do X things to the appropriate API based on host, so that the tools could be shared ;)
21:20:04 <gholms> Yeah
21:22:11 <gholms> You can do some neat configuration with their utilities, so if we can adopt the same system that would benefit everyone.
21:22:48 <gholms> brianlamere: Can I jot you down for looking into cloud-init and we can discuss our thoughts later?
21:23:17 <brianlamere> gholms:  certainly
21:24:01 <gholms> #link https://launchpad.net/cloud-init/
21:24:05 <gholms> #action brianlamere and gholms to look over, discuss cloud-init
21:24:14 <gholms> Anything else image-related?
21:25:23 <gholms> Okay!
21:25:30 <gholms> #topic EC2 kernel status
21:25:44 <gholms> Are we able to use the stock kernel wholesale now?
21:26:03 <jforbes> Yup
21:26:05 <brianlamere> haven't we been, since pv-grub?
21:26:28 <jforbes> Well, since shortly after, we had to get a patch in
21:26:28 <brianlamere> though, Ben is messing that up with the patch I shot to jforbes ;)
21:26:38 <gholms> It was on the wiki page last I checked, so I'm double-checking.  ;)
21:26:51 <brianlamere> fair enough
21:26:52 <gholms> But that can go upstream, right?
21:26:54 <jforbes> Yeah, wiki hasn't been updated in a while
21:27:24 <jforbes> Umm, the patch from ben, I hope so, haven't seen it yet... the other one we are carrying, not a snowballs chance, but it is a 1 line patch
21:27:36 <gholms> What's that one for?
21:27:59 <jforbes> making pv_ops kernels boot on some older hypervisors
21:28:09 <gholms> #info Fedora's kernel works as-is
21:28:13 <brianlamere> oh yeah, forgot about that one
21:28:39 <gholms> #info This includes one (now two?) patches on top of the mainline kernel
21:29:01 <jforbes> we haven't added the other, and it is not a foregone conclusion that we will
21:29:06 <jforbes> But I will take a look at it
21:30:43 <gholms> Anything else kernel-related?
21:30:52 <jforbes> not that I am aware of
21:30:58 <gholms> #topic EC2 feature status
21:31:05 <gholms> #link http://fedoraproject.org/wiki/Features/EC2
21:31:37 <jforbes> nothing to discuss there (wiki really needs to be updated for Cloud SIG, you are going over meeting bits from July)
21:31:50 <gholms> So, the EC2 feature page hasn't been updated in a while.  What has changed in the last month as far as that goes?
21:32:08 <jforbes> nothing, there is nothing to do until beta
21:32:37 <jforbes> though things like cloud-init would be really nice to have
21:32:42 <gholms> Sorry.  Feature completion deadline is coming up and the page hasn't been updated in a month.
21:33:10 <jforbes> And what is the completion percentage?
21:33:16 <gholms> 75%
21:34:12 <jforbes> Oh, someone moved it
21:34:12 <gholms> Right now we need to document the process for creating and publishing EC2 images and set up mirror infrastructure.
21:34:12 <jforbes> gholms: it was at 100% until you moved it
21:34:24 <jforbes> neither of those are required for F14 release
21:34:27 * gholms looks again
21:34:35 <jforbes> https://fedoraproject.org/w/index.php?title=Features/EC2&diff=195445&oldid=190039
21:34:54 <gholms> Oof, you're right.  Sorry, I was really tired.
21:35:33 * gholms goes to fix that
21:35:37 <jforbes> Anyway, both of those issues are ongoing concerns for the cloud SIG, but neither are required for F14 concurrent release
21:36:02 <gholms> We don't need an image-building process?  Okee dokee.
21:36:30 <gholms> (You guys have a much better idea of how this works than I do, after all)
21:36:54 <jforbes> We cannot depend on getting the fully open image building tools into Fedora in time, so no.  Once those are in, we do need an image-building process for releng to follow
21:37:27 <gholms> Fair enough.  I presume we've tried euca2ools 1.3.1 and found that they don't work?
21:38:18 <jforbes> No, we haven't tried them that recently.
21:38:23 <jforbes> It would be nice if they do
21:38:29 <brianlamere> (2 pennies:  I agree that the image itself is what is important, versus documenting how anyone else can recreate it)
21:38:45 <gholms> Yeah, we can still release without it as long as we're making progress.
21:38:49 <obino> jforbes: if they don't I be that mitch would like to fix them
21:39:13 <jforbes> obino: I wll give them a spin tomorrow to see
21:39:36 <obino> great: let us know if we can help. Feel free to Cc: mitch since I'll be on vacation
21:39:48 <gholms> #agreed Will try building images with euca2ools, though this is not a blocker for concurrent F14 release
21:40:22 <gholms> Anything else EC2-related, or shall we go on to Eucalyptus?
21:41:14 <gholms> #topic Eucalyptus packaging
21:41:44 <gholms> #info euca2ools 1.3.1 in updates-testing; please test
21:42:11 <gholms> Any progress on this front?
21:42:23 <gholms> obino: ^
21:42:59 <obino> gholms: as I mentioned to you I'll be on vacation soon, but we are coming up with new documentation to use the official RPMs instead of ours
21:43:22 <obino> after that we'll pull our RPMs from the download pages and we'll keep them only on the nightly builds
21:43:58 <obino> 1.3.1 itself was tested, but I'm trying to talk to engineering to add tests for the officials pacakges
21:44:29 <gholms> If you have tested it on Fedora, please go to https://admin.fedoraproject.org/updates/ and say so.  It *really* helps.
21:45:22 <obino> got it: will try to do so, but I'm taking off tomorrow: I'll do what I can
21:45:58 <gholms> If anyone else at Eucalyptus tested it then he or she can do so as well.  It doesn't have to be you specifically.
21:46:33 <obino> I'm reaching out now over my desk :)
21:46:40 <gholms> Anyone done anything with Eucalyptus proper?
21:47:18 <obino> are you asking for the packages?
21:47:40 <gholms> As far as improving their state with regards to getting into Fedora
21:47:45 <obino> got it
21:48:05 <obino> not yet: I'm trying to gather a 'task-force' of people interested in Fedora
21:48:12 <gholms> Ok
21:48:17 <obino> the first steps are to get interested and involved
21:48:35 <obino> I already got Mitch involved at least with euca2ools
21:48:44 <brianlamere> tell them there's pie involved.  I don't know from who, but I'm sure there is.
21:48:48 <obino> I hope he will get more understanding of rpm packages from there
21:48:54 <obino> hehe
21:49:01 * gholms wonders if Eucalyptus is hiring  :)
21:49:16 <obino> yes we are: want to apply? :D
21:49:22 <brianlamere> first step is just a process for getting it working on a fresh install, then on from there
21:49:39 <gholms> obino: Sure.  We can talk afterwards.
21:49:54 <obino> well eucalyptus works on fedora on a fresh install
21:50:00 <obino> both from source and our pacakges
21:50:13 <gholms> Yeah, the changes are all about making it guideline-compliant.
21:50:16 <obino> we need to integrate better with selinux (<-- understatement)
21:50:22 <obino> and firewall
21:50:24 <gholms> ...and that.
21:50:37 <brianlamere> well, ok - the packages have the extra in it however.  So maybe "fresh install" isn't really what I meant ;)
21:51:44 <gholms> #help Packaging, selinux help wanted for Eucalyptus RPMs
21:52:33 <gholms> Anything else Eucalyptus-related?
21:53:14 <gholms> #topic OpenStack packaging
21:53:41 <gholms> Last week openstack-swift was approved.  Are we doing anything more with OpenStack?
21:54:44 <gholms> ...or shall we go to open floor?
21:55:42 <gholms> #topic Open floor
21:57:01 * gholms will close in 60 seconds unless someone has something to say
21:57:58 <gholms> Thanks for coming, everyone!
21:58:01 <gholms> #endmeeting