21:02:29 #startmeeting 21:02:30 Meeting started Tue Aug 2 21:02:29 2011 UTC. The chair is ttx. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:02:31 Useful Commands: #action #agreed #help #info #idea #link #topic. 21:02:42 Welcome to our weekly OpenStack team meeting... 21:02:51 Today's agenda is at: 21:02:56 #link http://wiki.openstack.org/Meetings/TeamMeeting 21:03:10 when do we add incubated projects to the agenda? 21:03:29 notmyname: sorry, that was my action item I think when ttx was away. 21:03:30 i'm here 21:03:57 notmyname: ah. We can have an incubated news topic, after Nova 21:03:58 ttx: my apologies. dropped the ball on that one. I was supposed to make sure dash and keystone were updating on this meeting. 21:04:01 willdo 21:04:11 #topic Postmortem feedback for 1.4.2/diablo-3 release 21:04:24 Last week we released Swift 1.4.2 and the diablo-3 milestone for Glance/Nova 21:04:38 It went well, though we had a bit of delay for diablo-3 to try to sneak some Glance bugfixes 21:04:51 Anything that went wrong from your perspective and that we need to fix ? 21:05:17 ttx: https://blueprints.launchpad.net/openstack-ci/+spec/glance-upgrade 21:05:29 ttx: we need to work on that one for next milestone release... 21:05:45 sounds like a good idea 21:05:56 jaypipes: we also need a bzr-tarball-delta job 21:06:09 that would have caught the absence of glance-scrubber early 21:06:12 yup. 21:06:36 we have a CI bug for that. mtaylor: do you prefer a Blueprint ? 21:06:48 anything else ? 21:07:14 ttx: bug is fine 21:07:18 #topic Swift status 21:07:23 notmyname: o/ 21:07:32 Do you have a timeframe for 1.4.3 already ? 21:07:34 1.4.2 is done 21:07:42 I was looking at 1.4.3 today 21:07:45 Vek: sorry, reading scrollback - in the new world order, you will get an email 21:07:50 diablo-4 is scheduled on August 25, in case you want to align :) 21:08:15 I think we will have time for only one more release before diablo (the version that will be in diablo) and still have time for docs, etc 21:08:36 I was thinking sometime around the first week of september, but I haven't settled on anything yet 21:08:45 yes, makes sense 21:08:48 8-25 is a good date to know. I'll keep it in mind 21:08:55 notmyname: Any features already planned for that version ? 21:09:47 nothing big. whatever we can get done. we've finished everything that was promised for diablo, so I expect 1.4.3 to simply be bug fixes and perhaps a few small things. we're investigating time-limited files, for example 21:10:09 would be a cool feature, indeed 21:10:14 indeed 21:10:24 notmyname: Other announcements/comments ? 21:10:37 ya 21:11:18 I hope that scaling down swift deploys can be done in essex (deploy on <5 servers). we need community input there 21:11:53 you need input before the design summit ? 21:11:59 it's something that's needed, but not something current devs have a lot of insight to since we have existing, large clusters 21:12:01 no 21:12:15 nothing needed before the design summit 21:12:39 Raise your hand if you have questions on Swift... 21:12:47 also, I guess PTL elections are coming soon, so get your commits in if you want to vote 21:12:56 0/ 21:12:56 notmyname: I might have some input for you on that with FreeCloud... I'll ley you know. 21:13:11 notmyname: I committed version changes, does that count ? :) 21:13:18 What's the status of keystone authN integration with Swift? 21:13:18 lol 21:13:23 ttx: I think you're already there :-) 21:13:28 cool! 21:13:35 heckj: great question! ask the keystone people :-) 21:13:45 heckj: there's a middleware checked in to keystone; that's about all I know. 21:13:58 notmyname: swift is already to go once Keystone finishes something? 21:14:08 heckj: as far as we know 21:14:24 notmyname: cool, thanks 21:14:32 +#topic Glance status 21:14:37 ow 21:14:41 #topic Glance status 21:14:41 https://launchpad.net/glance/+milestone/diablo-4 21:14:53 Looks good to me... 21:14:55 We're kickin' ass on D4. 21:15:12 jaypipes: keystone integration blocks another diablo-4 spec, how far is it ? 21:15:37 thx to the Brians, Vek, s1rp, jkoelker, and johan 21:15:55 ttx: Vek needs to complete a functional test with auth spun up. 21:16:02 * Vek threw a glance_auth_token.py middleware into keystone 21:16:22 jaypipes: Announcements, comments ? 21:16:38 ttx: we need to have a common way of starting Keystone servers, daemonized when there could be other keystone servers runing on the box... 21:16:52 ttx: basically, what glance-control gives us in Glance (and swift-init in Swift) 21:17:27 ttx: no announcements. currenlty a few outstanding packaging things mtaylor is working on. we're looking good to hit D4. 21:17:35 ok 21:17:41 Raise your hand if you have a question on Glance. 21:17:43 ttx: oh, and moving to Gerrit/GH on Thursday morning. That little thing. :) 21:17:53 bah! almost nothing. 21:18:03 heh 21:18:09 piece of cake 21:18:37 #topic Nova status 21:18:44 vishy: yo! 21:18:48 heyo! 21:18:52 Looking at https://launchpad.net/nova/+milestone/diablo-4 21:19:04 Still a bit work in progress, with some specs being added 21:19:19 like the networking integration stuff that is being split into more trackable bits 21:19:50 vishy: I hope that we can get it into order by the end of the week ? 21:20:06 ttx: just broke up the blueprints today 21:20:12 yes 21:20:20 vishy, soren: I had one question already, about EC2 Id compatibility 21:20:21 should be able to merge prop at least one of them in the next few days 21:20:23 blueprints should be prioritized properly in the next couple of days 21:20:34 vishy, soren: do we have a clear plan forward there ? 21:20:48 vishy, soren: it's marked essential, so I'm getting nervous 21:21:28 because from where I stand it's still very much under discussion. 21:21:37 I'm actually comfortable shipping without changes and saying that ec2_api is only supported in one zone configs 21:21:57 Me too. 21:21:57 vishy: ok, so we can at least downgrade to "High" 21:22:08 It's not "essential" for sure. 21:22:13 soren: ok, so we can at least downgrade to "Medium" :) 21:22:34 * ttx sets to "High" to drop some pressure 21:22:41 soren: iirc you were opposed to the idea of a mapping layer at the top for ec2_ids 21:22:54 soren: although I still think it is the easiest way forward 21:23:38 danwent: saw that, thanks. 21:24:03 vishy: I think that's a dreadful idea, yes. 21:24:35 well, at least you don't feel that strongly, soren. :) 21:24:42 haha 21:24:48 could also use someone for this blueprint: https://blueprints.launchpad.net/nova/+spec/aws-api-validation 21:24:49 I'm not sure I've seen a better idea. 21:24:57 vishy: ...but hardly anyone seems to care about the EC2 API, so there seems to be little chance of going any other route. 21:25:20 soren: what would be an alternate solution ? 21:25:47 jaypipes: nice one ;) 21:25:48 ttx: we discussed it in great detail in the ml thread. 21:25:49 ttx: Using an ID generation mechanism that doesn't require one of our API's to have its own mapping system. 21:26:13 soren: which involves chaging the rest of Nova again. gotcha 21:26:20 ttx: I don't think I'm int he sort of mood where I can give an objective run-down of the options. 21:27:16 soren: wel, you can't be asked to implement a solution you find dreadful, so at the very least, that would need to be reassigned 21:27:25 soren, vishy: is this something that can be resolved in the next few days? 21:28:10 soren: I agree with that train of thought, it just feels rather limiting given how ec2 instance ids are used. 21:28:27 ttx: It does seem natural that the person who actually seems to care about the EC2 API actually makes sure it works. 21:28:53 soren: fair enough 21:28:55 soren: there's more than you that care about the EC2 API, trust me. Lots of NTT folks do as well. 21:29:01 jaypipes: I doubt it. I've given up pursuing happiness on that particular endeavour. 21:29:18 soren: ok then. let's shelve this for offline. 21:29:32 ok -- In all cases, the diablo-4 plan should result in a lot of branches landing. 21:29:45 We need to propose early and review early, and merge what can be merged asap 21:30:12 For example, comments on the last part of boot-from-volume have apparently been adressed: 21:30:17 https://code.launchpad.net/~yamahata/nova/boot-from-volume-2/+merge/68496 21:30:28 blamar, devcamcar, other nova-core: please rereview it and get it off the table if it's ready. 21:30:37 vishy: more comments ? 21:31:05 just main focus on early reviews would be great, since we're trying to get a lot of stuff in 21:31:27 Questions for Nova PTL ? 21:31:56 #topic Post-D4 branch handling for Nova and Glance 21:32:12 jaypipes, vishy: I want to discuss how to handle the post-d4, feature-frozen pre-diablo-release timeframe 21:32:23 Like I already told you in May, I think we have two options: 21:32:37 "Long" one: No more Diablo features after August 22, which is when the diablo-4 milestone branch needs to be created 21:32:52 We use the last milestone branch as the 2011.3 release branch. Trunk development switches to Essex. 21:33:07 Unrestricted bugfixes land in release branch and get ported to trunk, until we switch to targeted bugfixes 21:33:22 (after that only specific fixes land in release branch and get ported to trunk, others go directly/only to trunk). 21:33:36 "Short" one: Features should have landed by diablo-4, but 2011.3 release branch is only cut on September 8th 21:33:48 so what lands in Diablo remains in pure nova-core control for two more weeks 21:34:02 After that date only targeted bugfixes are accepted, and trunk development switches to Essex 21:34:19 Advantages of long one: diablo-4 contains all features and serves as beta. Trunk is always open(though switched to Essex early) 21:34:34 Drawbacks of long one: August 22 is early. 4 long weeks of tracking and proposing bugfixes to two parallel branches 21:34:49 Advantages of short one: More team focus on bugfixes. "Only" two weeks of parallel branches 21:34:55 ttx: I would prefer the shorter one for Glance. 21:35:04 Drawback of short one: Features are not very welcome in trunk for 2 weeks (soft feature freeze) 21:35:16 jaypipes: i also kinda prefer the short option 21:35:23 vishy: ? 21:35:36 short 21:35:42 but it runs a bit counter to the "always open trunk" philosophy we decided at the design summit 21:35:58 If you both are comfortable with a soft feature freeze, then I'm ok too 21:35:59 true 21:36:10 ttx: sure, but this is only once every 6 months... 21:36:18 and only two weeks. 21:36:20 ttx: and the goal here is integration with other projects and bug fixing... 21:36:34 and testing 21:36:36 i've noticed that merging in fixes is a little painful because people don't often make them off of the release branch 21:37:01 so you have to rebase the changes and repropose them against the release branch, which also creates divergent history 21:37:06 yep. 21:37:32 vishy: and the long option would only make that process longer... 21:37:43 right 21:37:59 so not merging features into trunk for two weeks seems ok 21:37:59 OK, let's go with "short" and try to see how to we can simplify dual-proposal process 21:38:18 its not as if they can't still propose them for review to get eyes on them 21:38:34 vishy: right, at least postpone merging them for a couple weeks, until Essex opens. 21:38:42 ya 21:38:54 #topic Incubated projects news 21:39:13 devcamcar: anything on dash you want to bring up? 21:39:15 is quantum incubated yet ? 21:39:22 zykes-: nope. 21:39:29 zykes-: keystone and dashboard right now. 21:39:32 ah 21:39:37 * jaypipes searches for zns.. 21:39:39 zykes-: not proposed yet. 21:40:02 zykes: you can stay in the room for the quantum discussion at the top of the hour 21:40:31 danwent: ? 21:40:35 ah 21:40:56 ok, sounds like nobody is there from incubated projects 21:41:03 * jaypipes asked dolphm to join us for a status update on keystone. 21:41:05 jaypipes is trying to get one 21:41:21 dolphm: welcome! 21:41:22 I'll make sure they get the news that they have a meeting topic now :) This was added a bit late. 21:41:27 jaypipes: thank you 21:41:49 dolphm: wondering if you want to update the community on keystone? anything you want to say about progress made, etc? 21:42:36 hmmm... i don't have too much to say... 21:42:56 dolphm: ok, no worries if you don't. just wanted to give you all an opportunity. 21:43:15 dolphm: is there anything that the community can assist you with? anything you want to bring up regarding Gerrit? 21:43:21 i'm glad to be on gerrit / jenkins - that's pretty much the story of my week 21:43:25 dolphm: Does Keystone rock ? 21:43:38 ttx: yes, by default 21:43:40 ok, that's the spirit ! 21:43:42 lol :) 21:43:51 ok, enough harrassing dolphm 21:44:00 it's a specially shaped rock, in fact. 21:44:07 heh 21:44:08 #topic Open discussion 21:44:13 hey all 21:44:14 * ttx opens the bar 21:44:18 and one rock to bind them.... 21:44:22 brief note on hyper-v wins 21:44:37 primeministerp1: I like wins. 21:44:38 ttx: i have a nova scalability issue i'd like to point out 21:44:50 primeministerp1: go first 21:44:53 we'll be discussing the hyperv/openstack cloud in the interop lab at novell's brainshare in october 21:44:55 for "sucks" report: trying to merge code that depends on features that have been added to something else. 21:45:14 primeministerp1: ++ w00t. 21:45:17 hopefully we'll also be discussing it during the upcoming openstack conf as well 21:45:26 primeministerp1: where/when is that exactly ? 21:45:30 so the novell one 21:45:37 is in Salt Lake City 21:45:46 the week after the openstack conf in boston 21:46:04 ideally we present the same material in both 21:46:33 so not really sure what the venu will be like, but it's a good opportunity to spread the word 21:46:50 primeministerp1: is that a brainstorming discussion, or a full-fledged presentation ? 21:47:02 full fledged presentation 21:47:08 ok, so more for the conference 21:47:12 yes 21:47:24 so I figure we can talk about the brief history 21:47:26 primeministerp1: the CTP should go out any time now 21:47:27 what we did 21:47:28 etc 21:47:49 comstud: ok, bad news now 21:48:01 ttx: great i'm planning on emailing spectorclan tomorrow about it 21:48:04 ok.. we have a bug: lp797770 21:48:17 Vek: agreed, this is why I've been loathe to break out projects willy nilly 21:48:18 socket closed errrors on high load 21:48:21 bug 797770 21:48:22 Launchpad bug 797770 in nova "'Socket closed' during API stress test" [High,Confirmed] https://launchpad.net/bugs/797770 21:48:46 comstud: bug which I'd like to see fixed before release, yes 21:48:55 was talking with eday... we've determined that the mysql engine sqlalchemy is using by default uses the mysql C library 21:49:05 so it's using libc socket calls... 21:49:08 eventlet cannot wrap these. 21:49:17 this means that any calls to mysql block until completed 21:49:33 *cough* 21:49:40 *cough* 21:49:46 right 21:49:56 so eventlet can't switch greenthreads and do other things while mysql queries are in progress 21:50:04 comstud: your "by default" seems to imply there is a solution 21:50:04 ttx: is there any like "events.openstack.org" ? 21:50:08 there _is_ a 'pymysql' engine option for sqlalchemy... 21:50:16 it's a purely python module... but I'm not sure of it's stability 21:50:25 zykes-: hrm... maybe 21:50:27 comstud: you could put the db operations in a thread pool 21:50:38 creiht: Or that was my other suggestion :) 21:50:39 just another option 21:50:49 assuming the C code will unlock the GIL 21:50:55 which I'm sure it probably does 21:50:56 i've heard others use the thread pool trick 21:51:05 we do that for the sqlite dp operations in swift 21:51:09 db 21:51:23 i mean specifically to get around the gevent/eventlet issue 21:51:35 In any case.. it's possible using 'mysql+pymysql://' for the engine will get around the current issue... but I've been unable to verify it so far. 21:51:47 I just wanted to make ppl aware of this 21:51:57 what's keeping you from verifying it? 21:52:16 Vek: time... ant tried, but got some weird exception 21:52:22 comstud: thanks for the heads-up 21:52:24 I plan to look at it in the next couple of days 21:52:26 'k 21:52:34 and for looking at that sort of bugs 21:52:38 if this doesn't solve it, it's possible we may want to think about threads. 21:53:05 zykes-: maybe http://openstack.org/community/events/ 21:53:14 comstud: http://eventlet.net/doc/modules/db_pool.html 21:53:15 0ok 21:53:37 abstracts some of that away, but not sure what would need to be done to make it work with sqla 21:53:49 anything else before we close ? 21:53:57 creiht: will take a look 21:54:27 I appreciate the Keystone guys keeping their doc updated in openstack-manuals! 21:54:48 annegentle: have we been doing that?! 21:54:52 :) 21:54:55 haha :) 21:55:03 lol 21:55:20 on that good note... 21:55:25 #endmeeting