Perl 6 - the future is here, just unevenly distributed

IRC log for #puppet-openstack, 2013-04-04

| Channels | #puppet-openstack index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:25 dachary joined #puppet-openstack
07:22 dachary joined #puppet-openstack
08:07 _ilbot joined #puppet-openstack
08:07 Topic for #puppet-openstack is now Place to collaborate on Puppet/OpenStack tools: logs at http://irclog.perlgeek.de/puppet-openstack/today
08:34 derekh joined #puppet-openstack
11:13 dachary joined #puppet-openstack
17:17 d3u joined #puppet-openstack
18:01 EmilienM joined #puppet-openstack
19:54 dachary joined #puppet-openstack
20:23 dachary joined #puppet-openstack
20:32 openstackgerrit joined #puppet-openstack
21:10 dachary joined #puppet-openstack
21:50 openstackgerrit joined #puppet-openstack
22:15 openstackgerrit Dan Prince proposed a change to stackforge/puppet-keystone: Run 'keystone db_sync' as keystone user.  https://review.openstack.org/26154
22:18 openstackgerrit Dan Prince proposed a change to stackforge/puppet-keystone: Run 'keystone db_sync' as keystone user.  https://review.openstack.org/26154
22:23 openstackgerrit Dan Prince proposed a change to stackforge/puppet-glance: Add swift_store_auth_version to swift backend.  https://review.openstack.org/26158
22:28 openstackgerrit Dan Prince proposed a change to stackforge/puppet-glance: Make swift backend not enherit glance::api.  https://review.openstack.org/26160
22:39 bodepd we have gerritt notifications in the house!
22:40 mgagne :D
22:40 dprince joined #puppet-openstack
22:41 bodepd and it looks like the core modules are on their way to being what is used with smokestack!
22:41 bodepd dprince: welcome!
22:41 dprince bodepd: They are.
22:41 dprince bodepd: getting these understackforge should make all of this easier.
22:41 bodepd dprince: I have a lot of TODOs
22:42 bodepd I probably need to write a contributors guide
22:42 bodepd derek started on in etherpad
22:42 * bodepd looks for link
22:42 dprince bodepd: my question for you earlier was what *you* were planning on using for initial testing.
22:42 mgagne will there be a "Jenkins/CI" reviewer?
22:43 bodepd right now, I have the following patch submitted:
22:43 dprince mgagne: there will, yes.
22:43 bodepd https://review.openstack.org/#/c/26148/
22:43 bodepd this is only for unit testing.
22:43 bodepd I want integration tests to gate
22:43 bodepd (unit test == rspec puppet)
22:44 mgagne I guess this will replace travis =)
22:44 bodepd yes, this replaces travis
22:44 bodepd (it more or less emulated when travis was doing)
22:44 bodepd except, I am currently only testing the version of ruby that happens to be installed on the openstack CI nodes
22:44 mgagne got it
22:45 dprince bodepd: awesome.
22:45 bodepd (no 1.9.3, 1.8.7,head) in the matrix
22:45 bodepd clearly unit tests are not enough
22:45 dprince bodepd: so for integration are you doing something too? or will SmokeStack be the primary focus for that?
22:45 bodepd dprince: I have something
22:45 bodepd dprince: I need to understand more about smokestack
22:45 bodepd what OS's does it test on?
22:46 dprince bodepd: cool. SmokeStack is really the only thing upstream that I know of that does packages.
22:46 bodepd (right now my integration tests are for Ubuntu 12.04 and Centos 6.x)
22:46 dprince bodepd: currently Fedora.
22:46 dprince bodepd: Soon RHEL.
22:46 bodepd I have CI stuff that I integrate with github.
22:46 dprince bodepd: For upstream stuff (which I would like these modules to be) you'll have to have recent packages in order to gate things.
22:47 dprince bodepd: The idea w/ SmokeSTack is latest and greatest packages and config MGMT (puppet) should always work.
22:47 bodepd randmonly selected example: https://github.com/puppetlab​s/puppetlabs-cinder/pull/38
22:47 bodepd I totally understand that
22:47 dprince bodepd; gotta run... back in a few minutes to read scrollback.
22:48 bodepd please bear with me while I wrap my head around how things are going to change
22:49 bodepd right now, I have integration tests that perform some really basic tests for Redhat and Precise, based on the folsom packages
22:50 bodepd and based on the latest grizzly packages for Precise
22:50 bodepd I would like to at least initially, continue to run these on incoming pull requests to at least get a sanity check on it they appear to break things
22:51 bodepd I understand that running from the released packages is slightly less than ideal for your purposes
22:51 bodepd but I still want to run some kind of tests on Ubuntu (and eventually expand to Debian)
22:52 bodepd dprince: I have two questions for you
22:52 bodepd 1. in terms of process, do you have any issue with rspec-puppet. It will be tremendously helpful when combined with the unit testing gate
22:53 bodepd when I mention that we need a process doc, I want that to be a requirement for contribution
22:53 bodepd 2. I understand that you just want to verify that things work with master, but what are your thoughts on backwards compat?
22:54 bodepd ideally, I would like the modules to be able to support both master and the last released version
22:54 bodepd I understand this may not be possible which is why I say ideally.
22:54 bodepd the reasons this is important to me:
22:55 bodepd 1. b/c of the nature of the modules, some changes and development will be speciifc to the previously released version of openstack
22:55 bodepd 2. I would like the same modules to work one version back to ease with upgrade
22:55 bodepd I woudl at least like to say, let's try until we know that it is impossible
22:56 bodepd dprince: that is all for now ^^^
22:57 mgagne a visual representation of the flow would be cool =)
23:03 dprince 1) rspec-puppet *sounds* fine. I haven't used it too much though.
23:04 dprince 2) for backwards compat there are a couple options. Ideally we'd just have branches that follow each project (folsom, grizzly, etc)
23:05 dprince maintaining backwards compat is somewhat painful when new (required) services get added.
23:05 mgagne dprince: how would it work with puppetforge?
23:05 dprince mgagne: branches could get released to puppetforge.
23:06 dprince mgagne: So puppet-nova could have a 'grizzly' branch that gets released to puppetforge.
23:07 dprince mgagne: I personally won't ever be using puppetforge for openstack core components. I may use it for dependencies like mysql/rabbit/qpid, etc.
23:07 mgagne aren't they going to conflict? How will I add a dependency in my Modulefile or whatever to get the nova/grizzly version?
23:07 dprince mgagne: your module specific version scheme could support that.
23:08 dprince mgagne: just bump the major version number for each openstack release or something.
23:08 mgagne hmm, I did not consider that approach =)
23:09 dprince bodepd: My *concern* with you gating on old packages is... they are going to get in the way at some point.
23:09 bodepd dprince: I totally agree.
23:09 bodepd especially if we are not attempting any kind of backwards compat
23:09 dprince bodepd: If I use this for SmokeStack... and need a change to land within say an hour, I'm not going to want to wait for that.
23:09 dprince bodepd: If I have to wait, I'll fork them again and we'll end up where we are today (I have my modules, you have yours)
23:10 dprince bodepd: So I'd really like to see us treat the stackforge/puppet* modules treated just like upstream openstack in this regard where we cut stable releases, etc.
23:11 bodepd how are the stable branches going to be merged?
23:12 bodepd potetially, there will be a flood of requests in the stable branches initially when they come out (when people start using the modules)
23:12 dprince bodepd: And, as far as Debian/Ubuntu testing goes I would love for that to be in SmokeStack too. There is an element of "a chase" there though because packaging changes have to be made. I make those for Fedora.
23:12 dprince stable branches can have changes in Gerrit like any other project.
23:12 dprince 1) make a branch.
23:12 dprince 2) gerrit review stable/folsom
23:12 dprince or whatever.
23:13 dprince When that change lands you can repost a new puppetforge version if you want too.
23:13 bodepd how do we ensure that the right patches go to all of the appropriate branches?
23:13 bodepd I think the idea for puppetforge is.
23:13 dprince bodepd: are you talking about backports?
23:13 bodepd when there is a release, master gets moved to a branch, that makes a major release update to all modules
23:13 bodepd dprince: yes (or forward ports, or whatever)
23:14 dprince bodepd: we'll need guys to sign up for stable maintenance of those things (if required).
23:14 dprince bodepd: you should *never* have forward ports.
23:14 dprince bodepd: if there is a rule there is that.
23:14 bodepd fair enough
23:14 dprince bodepd: it lands upstream first... or it doesn't land.
23:15 dprince bodepd: (perhaps there is a corner case here or there)
23:15 bodepd I have a concern that while that rule is ideal, these modules may require things to occassionally only be added to older branhces
23:16 dprince bodepd: I think that might happend too. but it wouldn't need forward porting then right?
23:16 bodepd correct.
23:16 bodepd things would either go in through master, or they only go to a specific branch
23:16 bodepd ah, grizzly was released today
23:16 dprince bodepd: way I see it. A model like this gets us puppet modules that should always just work with upstream.
23:16 bodepd gotcha
23:17 dprince bodepd: so, when we hit havana it'll just be a matter of creating some branches and things should work (for packages we test on at least)
23:18 bodepd dprince: can you help out with the contribution guidelines? I assume most of these rules are consisten with docuemnted opensack guidelines?
23:18 dprince bodepd: sure. Derek had a good start no?
23:19 bodepd yes, I will start adding to that
23:20 dprince bodepd: cool. I'll take another look myself too.
23:20 bodepd also, if you need a 1 hour turn-around, how it that going to work if you need a +2 ?
23:20 dprince bodepd: Uncharted territory. I could just send it and ask forgiveness.
23:21 bodepd I'm not the biggest fan of that...
23:21 dprince bodepd: honestly with smokestack I usually know breaking changes before they happen if I'm paying attention.
23:21 dprince bodepd: I'm open to ideas.
23:21 dprince bodepd: trying to come up with something that continually works.
23:21 bodepd dprince: I starting looking at ruby-openstack (slightly off subject, but I noticed you were a contributor)
23:22 bodepd dprince: I will think about.
23:22 dprince bodepd: I am. Or was.
23:22 bodepd dprince: was?
23:22 bodepd dprince: what about a branch called cowboy?
23:22 dprince bodepd: It is based off of the old RUby openstack-compute. So that was my main contribution (I did that one)
23:23 dprince bodepd: cowboy could work. sounds like I'm just forking though (so why not stay where I am)
23:23 dprince bodepd: This is the problem we need to figure out.
23:24 bodepd ideally, the answer to that is that it makes sense to use the same modules that are used for prod deploys with smokestack gate.
23:24 bodepd and that the value is worth some amount of extra overhead
23:24 dprince bodepd: Right. What I'm talking about is something totally different here. I'm talking about a breaking Nova change which (in order to pass) requires that I change puppet-nova.
23:25 bodepd 2 part answer
23:25 dprince bodepd: That is why real gating w/ puppet modules is tricky. I might prefer just relying on pre-gating tests.
23:25 bodepd 1. the modules should be flexible enough where this does not happen, the biggest offender atm is the command line parsing
23:26 bodepd that is why I was looking at ruby-openstack
23:26 dprince Sure. But there are always corner cases.
23:26 bodepd 2. when things are completely broken (example: nova-conductor)
23:26 dprince quirky permissions issues that crop up, etc.
23:26 bodepd token_format
23:27 dprince Sure. Well I spend a good bit of time trying to fix those things.
23:27 dprince But it happens.
23:27 dprince I have to run again.
23:27 bodepd dprince: productive conversation.
23:27 dprince Lets pick this up again tomorrow.
23:27 bodepd we'll make this work
23:27 bodepd I'm off tomorrow
23:27 dprince bodepd: cool. cheers

| Channels | #puppet-openstack index | Today | | Search | Google Search | Plain-Text | summary