19:59:32 #startmeeting heat 19:59:33 Meeting started Wed Nov 28 19:59:32 2012 UTC. The chair is asalkeld. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:59:34 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:59:36 The meeting name has been set to 'heat' 19:59:54 #chair shardy zaneb 19:59:55 Current chairs: asalkeld shardy zaneb 20:00:00 http://wiki.openstack.org/Meetings/HeatAgenda 20:00:04 #chair shardy zaneb asalkeld stevebake 20:00:05 Current chairs: asalkeld shardy stevebake zaneb 20:00:29 well done shardy on agenda 20:00:38 guys, I need to head out in about 40mins so may need to leave early 20:00:46 ok 20:00:55 well lets be quick 20:01:05 roll call? 20:01:07 +1 on being quick ;) 20:01:21 here 20:01:26 here 20:01:26 o/ 20:01:27 o/ 20:01:27 here 20:01:38 #topic packaging 20:01:57 so we need pypi ? 20:02:10 Yeah, so there was a request from one of the HP guys for pypi 20:02:21 do we put the package on the openstack pypi 20:02:22 I don't know if we need it, but it was requested 20:02:36 are other openstack projects in pypi? 20:02:49 honestly not sure 20:02:54 not sure either 20:02:55 my recollection is that to the extent they are, which is not great, it is a total pita 20:03:04 some of the clients are 20:03:14 I think a really outdated version of swift is 20:03:16 ya, for tests 20:03:19 it would be good for project visibility, and maybe our ubuntu instructions could use it until there is a PPA 20:03:21 key issue seems to be ease of use on ubuntu rather than pypi specifically 20:03:32 jpeeler: good point, and that makes sense 20:03:33 for several users (all our users?) 20:03:49 well maybe talk to the guys asking 20:03:59 and really find out what to do 20:04:04 ubuntu ppa > pypi imo 20:04:07 python-heatclient should be in pypi anyway, since its a lib 20:04:14 yip 20:04:24 yes, client definitely in pypi 20:04:38 #action put heatclient in pypi 20:04:53 so as a first step, I'll get the getting started and tools/openstack working properly on ubuntu, that should satisfy most of the initial issues 20:04:54 is there a bug/bp for the packaging? 20:05:11 yep, linked in the agenda ;) 20:05:14 cool well done shardy 20:05:22 doh 20:05:31 well lets move on 20:05:43 #topic State of ubuntu host/guest support 20:06:11 so two issues - ease of install, and ubuntu guest functionality 20:06:22 I'll tackle the first as mentioned above 20:06:34 so we addressed packaging (ya) 20:06:46 what about the guests 20:06:57 jpeeler: did you look at the cloud-init path issue? 20:07:16 a while back, not recently 20:07:33 changes between different versions of cloud-init is the problem 20:07:49 can we just increase the Importance on ubuntu bugs 20:08:11 i started looking for a way to query for the version, but that was only recently added. that's where i stopped 20:08:20 Ok, so we need to figure that out, and update the ubuntu tdls to install a recent version of boto (need a bug for the latter) 20:09:00 what about integration testing - that currently only targets Fedora 20:09:17 #action make a bug to update the ubuntu tdls to install a recent version of boto 20:09:35 shardy, sounds like a good idea 20:09:45 to have a mix of guests 20:09:54 we have the all-clear to request some help from the CI team, but not sure what exactly we need atm 20:10:15 ok 20:10:33 to get the functional tests running? 20:10:43 mmm 20:11:01 should we also talk to Derek Higgins to see if there are some lab resources we could use? 20:11:25 So if I get the ubuntu getting started markup sorted, what would it take to plumb that in with the existing automated tests zaneb? 20:11:57 (the ones we are not running) 20:12:11 we do really need to run those 20:12:12 shouldn't be too bad, but would have to get the existing ones running again first 20:12:34 #action start running the integration tests 20:12:34 need better visibility of the results, so it's noticed when we break them 20:12:40 need to copy over the ubuntu ISO and build a jeos for it 20:13:06 shardy: problem is we never got them all working on the server 20:13:18 (and now the server is broke too) 20:13:24 doh 20:13:47 ok all in all - not good 20:13:58 auth to beaker is not working, I'll have to file a ticket to get it sorted 20:14:14 but I really do feel that we would be better off migrating to tempest 20:14:20 this is becomming a resource sucker 20:14:54 we need to have a plan for this 20:15:04 Any ideas how we can get better value out of the existing tests without committing lots of resources? 20:15:05 I have spent at least 3 months of this year on it already, it was always a giant resource sucker 20:15:14 how do other projects handle this? 20:15:24 shardy: yes, tempest ;) 20:15:34 ok, lets do that then ;) 20:15:46 is that more work? 20:15:46 Dan Prince is Mr Tempest, we should ask him 20:15:55 yip 20:15:55 that way we don't need to do any infrastructure stuff 20:16:04 well that is good 20:16:28 big issue that I see is that we somehow need a guest image to launch 20:16:29 anyone interested in finding out about it 20:16:40 yip 20:16:57 to the tests have to build a new image every run? 20:17:15 Could they not just download a pre-built jeos? 20:17:18 well they could download the example images 20:17:23 stevebake: no, although it has to be up to date if we keep changing the cfn-tools 20:17:42 zaneb: automate rebuilding the pre-built images 20:17:55 sdake already wrote a script for that IIRC 20:17:58 or get cfn tools packages 20:18:10 Yep, cfntools packages sounds better to me 20:18:14 or get cfn tools packaged (how is that goind) 20:18:25 then we can have ostools which use the rest api 20:18:25 +1 20:18:49 yip 20:18:52 yeah i was going to package cfntools 20:18:55 *or* use the AWS cfn tools, which are presumably already packaged ;) 20:19:08 can we do that ^^^ 20:19:11 zaneb, we have an extension for ha 20:19:18 ok so no 20:19:25 to restart services 20:19:40 asalkeld: can we inject it through cloud-init? 20:19:50 yes 20:20:00 let's do that 20:20:02 we could have cloud-init do that update 20:20:04 wouldn't that let us use other jeos images off amazon? 20:20:21 no 20:20:36 we would need the cloud-init config 20:20:51 do AWS images not have cloud-init? 20:21:12 yes, they do but not that default config 20:21:23 I don't know enough about how the native aws cfn images work 20:21:23 to install our cfn-init 20:21:51 I think we should move on 20:21:55 that's what I'm suggesting, use Amazon's cfn-init 20:22:15 yeah, moving on 20:22:33 some unresolved issues there 20:22:45 #topic RHEL/CentOS support? 20:23:02 so do we see that as important? 20:23:05 is this for guest/host 20:23:18 clearly nobody is going to deploy Fedora in a production environment 20:23:19 looks like host 20:23:30 I was thinking host 20:23:40 but TDLs for RHEL would be good too 20:23:44 shardy, I think we need to focus on upstream work 20:23:52 less disto 20:24:06 else we will never get any work done 20:24:15 Ok, the problem is our users are demanding distro-specific instructions... 20:24:20 I would think rhel support will come at some point, maybe not now? 20:24:20 but yes to testing many guests 20:24:33 is there a public date for RHEL7? 20:24:38 stevebake: no 20:24:41 ok, cool, lets drop that item for now then 20:24:49 shardy, mainly ubuntu right now 20:24:58 I guess the issue is if you need packages in rhel it's better to start sooner rather than later 20:25:02 stevebake: "second half of 2013" is what I found on google 20:25:05 yeah, I'm trying to get some users who're not on ubuntu ;) 20:25:11 because it can take a long time 20:25:28 #topic Way forward for integration tests 20:25:37 I think we have done this 20:25:47 that's what I thought ;) 20:26:02 #topic Launchpad issues 20:26:22 I think I have done that 20:26:23 think this is sorted after last nights reorg 20:26:29 I'm attempting to flesh out my blueprints 20:26:50 but keep checking bugs 20:27:00 but for the benefit of everyone - new features == blueprint, bug==bug 20:27:10 zaneb: so is REST done now? 20:27:11 and "weeding the garden" as steve would say 20:27:26 stevebake: to a first approximation, yes 20:27:40 zaneb, set the state on that bp 20:27:45 ok 20:27:56 still stuff like XML support to add 20:28:06 but all the basic functionality is there 20:28:07 all of you see if a bp matches the work you are doing 20:28:15 and update the state 20:28:38 #topic jpeeler to update re state of snapshot/prebuild investigation 20:28:39 * stevebake sees no xml bp 20:29:15 i was surprised to see this on the agenda since i thought it was decided it was too much work 20:29:25 especially after seeing this, which i didn't know existed: https://github.com/heat-api/heat-prebuild 20:29:27 so this idea sucked and we are moving on? 20:29:50 ok, I wasn't sure, I thought it was unresolved so added it to the agenda 20:30:11 that sounds like a good plan, but I don't think we'd actually _decided_ that before 20:30:31 jpeeler, how would you summerize the idea/project 20:30:47 so we had some grand ideas 20:30:47 the snapshot or the prebuild? 20:30:53 snapshot 20:31:37 is it doable/too much work, what are the show stoppers 20:31:41 the snapshotting was going to handle automatic caching with the option of allowing instances to launch either an exact clone or one that had all the packages installed, but would receive new user data 20:32:31 it's doable, it is a bit of work. one problem is handling simulatenous caching operations gracefully and a way to signal when user data execution was about to begin 20:33:04 userdata comes from the nova metadata server? 20:33:11 yeah 20:33:24 so we want that to run each time 20:33:25 you could use a waitcondition (or similar method) for the latter 20:33:27 ok, so no easy synchronisation there 20:33:44 I think we can use our time better 20:33:50 +1 20:34:03 jpeeler: what's your recommendation? 20:34:36 jpeeler, it would be good to write a brain bump on the wiki 20:34:56 so we don't loose the info you have learned 20:34:57 it is a lot of work, but at the same time i'm guessing part of the reason it was brought up was for me to dive into the code 20:35:11 i can find other things to work on though 20:35:18 cool 20:35:41 +1 20:36:00 well we are getting to the end of topics, i suggest we go back to integration tests 20:36:05 +1 20:36:18 asalkeld: PTL? 20:36:39 #topic when do we need to choose a ptl 20:37:15 so I was asked about it at project meeting 20:37:21 it would be nice to be able to wait until sdake is back on deck 20:37:28 but bit sad as steve is sick 20:37:33 yes 20:37:54 so the qu. is when do we _have_ to do it by 20:37:58 otoh maybe there should be somebody as the designated email contact when the TC need to bug us 20:38:26 sure, stand in 20:38:27 how hard is it to change the PTL later? 20:38:29 when were re-elections? is it about 2 months before release? 20:38:46 well elected every 6 months - no? 20:39:02 just before each summit 20:39:08 not sure how the rules apply to incubated projects 20:39:19 since we don't affect the makeup of the TC, it may not matter 20:39:26 #action ask at next project meeting 20:39:52 (or just ask mark) 20:40:14 should we wrap up? 20:40:21 nice short meeting 20:40:35 is anyone going to FOSDEM? 20:40:55 I can't go, but would be good to get a heat talk there if anyone is going 20:41:03 where is it? 20:41:04 I think somebody needs to at least propose a heat talk 20:41:08 asalkeld: Brussels 20:41:22 ok zaneb it's right next to you 20:41:33 it's a weekend, which is why I can't go (sons birthday) 20:41:44 shadower: any interest? 20:41:51 btw, can y'all start thinking about documentation, I'll put it on the agenda for next week 20:41:59 zaneb, trying hard to avoid ;) 20:42:11 I would like us to one day sort out the integration test 20:42:16 I went last year 20:42:24 it was freezing! 20:42:28 zaneb, maybe 20:42:45 (I am doing aussie linux conf in jan) 20:42:46 I was thinking about talking Heat on the Devconf in Brno feb 20:43:12 shadower: that's not until the week after ;) 20:43:17 ok - the end ... 20:43:27 #endmeeting