21:02:28 <danwent> #startmeeting Quantum Team Meeting 21:02:29 <openstack> Meeting started Mon Aug 27 21:02:28 2012 UTC. The chair is danwent. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:02:30 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:02:31 <openstack> The meeting name has been set to 'quantum_team_meeting' 21:02:38 <danwent> huh.. someone updated meetbot :) 21:02:55 <zhuadl> cool! 21:02:55 <danwent> agenda: http://wiki.openstack.org/Network/Meetings 21:03:11 <danwent> ok, so this won't be a fun meeting, as we really need to take an axe to things for the Folsom release 21:03:24 <nati_ueno> hi 21:03:35 <danwent> the key thing to remember is that those features can still get into grizzly 21:03:35 <garyk> and it is late in the evening :) 21:03:43 <SumitNaiksatam> hi 21:03:55 <danwent> we just need to focus on tightening up existing functionality, and documenting it. 21:04:03 <danwent> otherwise, quantum as a whole will not be useful 21:04:09 <danwent> #topic announcements 21:04:20 <danwent> #info Quantum Folsom RC1 targeted for 9/10. Two short weeks out. 21:04:43 <danwent> given that we're behind, we decided to move the RC1 date back, but this does not mean exstra time to add new stuff 21:04:47 * cdub_ checks agenda 21:04:57 <danwent> we'll need the time for bug fixes and documentation 21:04:59 <Salv-away> Hi everyone! 21:05:01 <amotoki> danwent: meetingbot update https://lists.launchpad.net/openstack/msg16053.html 21:05:16 <danwent> amotoki: ah, no time to read all those emails :)( 21:05:17 <danwent> :) 21:05:25 <amotoki> just info. 21:05:37 <danwent> #info python-quantumclient will also be following a feature-freeze policy 21:06:18 <danwent> the motivation for this is that the clientlib is used heavily by the main codebase, not just the CLI, so we want to feature-freeze that as well, even though its not strictly required by openstack 21:06:34 <danwent> amotoki: yup, thanks 21:06:46 <danwent> #topic Folsom-RC1 21:07:01 <danwent> #info F-RC1 status https://launchpad.net/quantum/+milestone/folsom-rc1 21:07:17 <danwent> we really need the full attention of all core devs over these next two weeks 21:07:35 <danwent> this is crunch time, only happens twice a year 21:07:47 <garyk> how about we do a table like last time - 2 per patch? 21:07:53 <danwent> bug fixes, reviews of bug-fixes, and docs 21:08:05 <danwent> garyk: i'm open to all ideas 21:08:14 <danwent> one tricky thing now is that we shouldn't have ANY large patches 21:08:28 <danwent> so really we should have many small patches 21:08:39 <danwent> which makes the overhead of a list higher 21:08:51 <danwent> but if people like having the list, I'm cool with it. 21:08:54 <garyk> danwent: ok. maybe we should have at least 2 reviewers a day? 21:09:01 <danwent> garyk: I like that idea. 21:09:10 <cdub_> 1 list, lots of things...amortizes :) 21:09:39 <danwent> and the nice thing is that many of these changes should be small, meaning the likelhood of them requiring many re-reviews is lower 21:09:46 <danwent> cdub_: overhead is per entry added to the list :) 21:10:00 <danwent> or at least the over head i was talking about 21:10:29 <danwent> Salv-away or garyk, do one of you want to update the reviewdays page to have two entries per day for the next two weeks? 21:10:32 <cdub_> heh 21:11:09 <garyk> danwent: i'll update it (in the morning) 21:11:15 <danwent> Ok, so we REALLY need to make sure we're spending core dev resources well, so here are some RC rules I'm proposing 21:11:19 <danwent> garyk: thx 21:11:30 <danwent> garyk: I may do it myself then, so we can get people signed up today 21:11:57 <danwent> core devs: only bother to look at reviews that are for a bp/bug that is assigned to RC1 21:12:19 <danwent> everyone: please do not even propose branches to gerrit if they are not targeted for RC1 21:13:02 <danwent> only core devs can assign a bug to RC1. if you aren't a core dev, file it, and make sure a core dev knows about it, either directly, or by emailing the ML 21:13:11 <danwent> does that seem reasonable? 21:13:30 <danwent> again, my goal is to focus resources on things that are valuable community wide, and make sure core devs time is spent well. 21:14:03 <danwent> any concerns or suggestions on these rules? 21:14:26 <garyk> danwent: sounds good. 21:14:31 <danwent> I feel like we're at risk for the Folsom release, so I'm erring on the side of clamping things down pretty tightly 21:14:33 <mestery> danwent: I guess bug triage becomes quite important here if non-core devs open bugs. 21:14:34 <markmcclain> seems reasonable 21:14:46 <danwent> mestery: agreed. I did a full triage last night. 21:14:53 <mestery> danwent: You rock. :) 21:15:04 <danwent> I wonder if we should schedule a bug triage frequently though over the next two weeks. 21:15:28 <danwent> #todo #danwent as part of review days, include bug triage for all bugs filed the day before 21:15:31 <mestery> danwent: That's not a bad idea. 21:15:47 <garyk> +1 21:15:49 <danwent> mestery: thanks for mentioning that. 21:16:06 <cdub_> hmm, fedora test day is going to be 9/4, if we can get some good test cases written (and a snapshot package built) we might be able to get some decent test cycles there 21:16:14 <danwent> ok, before we move on, want to make sure people are on the same page about this. i'll send a not to the lsit about it. 21:16:29 <danwent> cdub_: yeah, garyk mentioned that. will be very valuable. 21:16:37 * markvoelker is on the page 21:16:42 <danwent> :) 21:16:49 <danwent> ok, now to the even less fun part.... 21:17:16 <danwent> My feeling is that we really need to free up core dev cycles to focus on testing, bug fixes, bug fix reviews, and docs. 21:17:31 <danwent> so we really need to take an axe to anything that isn't progressing well. 21:17:53 <danwent> its not that any particular feature couldn't be squeezed in if we had to, but the sum of all such features is way to much. 21:17:59 <danwent> so on that list for me are: 21:18:02 <danwent> xml v2 support 21:18:05 <danwent> multi-host dhcp 21:18:08 <danwent> per-port host routes 21:18:15 <danwent> nova metadata service + overlapping IPs 21:18:20 <danwent> nova proxy for floating ips 21:18:40 <danwent> all of those items are things I'd love to have in Folsom, but they aren't moving forward from a review perspective, so I think we need to drop them. 21:18:55 <danwent> of course, I know each of these features is someone's pet feature 21:19:09 <nati_ueno> per-port host routes should be in RC1 because it impact quantum v2 api 21:19:39 <danwent> but I think we need a strong majority of the quantum team behind the feature for us to keep moving forward 21:20:00 <danwent> nati_ueno: I suspect we may need to amend the API spec here if the spec does not match the code 21:20:28 <danwent> per-port host routes where kind of a grey area that we were hoping to squeeze in. 21:20:55 <danwent> but I just don't see it moving forward, and the reality is that making sure Quantum as a whole is stable and well documented just strikes me as vastly more important 21:21:02 <garyk> nati_ueno: i feel that we still have a lot of work to do here - the initial patch does not include the dhcp support and we need to do test ccycles 21:21:39 <nati_ueno> garyk: dhcp part is very small. Just 5-10 line 21:21:54 <danwent> Since we have a big chunk of the team here, I wanted to see if there was anything on this list that the team as a whole things we need to make a priority 21:22:32 <danwent> nati_ueno: as I said, we could nit-pick about any given feature. the reality is that the deadline is well past for Folsom features. 21:22:33 <nati_ueno> danwent: I'll abandon multi-host. But per-port host looks very important 21:22:46 <danwent> what do other people think about per-port host routes? 21:23:01 <danwent> like I said, pretty much every one of those features on the list is something I would like 21:23:11 <garyk> danwent: nati_ueno can you guys give a explanation of the feature so that all can understand better 21:23:15 <danwent> we just need to make some tough choices to insure overall quality 21:23:40 <danwent> right now, the routes that a VM will get via DHCP are defined on a per-subnet basis 21:23:42 <nati_ueno> garyk: This is functionality to set unique host_route per hosts 21:23:49 <jrd-redhat> What's the pain level associated with omitting per-port host routes? 21:23:54 <danwent> one might want to override the subnet policy on a per-port basis 21:24:24 <danwent> Use cases for this exist, but are not the common case, in my experience. 21:25:03 <markmcclain> seems like this could easily be worked around with chef or another config mgt tool 21:25:09 <garyk> danwent: i am in favor of quality and stability. how about priotitixing this as high for g1? 21:25:12 <cdub_> what about api impact? 21:25:25 <zhuadl> is it an essential feature for RC1? 21:25:29 <danwent> Salv-away_: I believe the API currently implements the per-subnet model only 21:25:39 <danwent> Salv-away_: can you confirm? 21:25:54 <nati_ueno> danwent: It is per-subnet only 21:26:05 <danwent> my bias is quality and stability as well. 21:26:27 <danwent> Ok, I am not hearing strong enough support for any of these items. 21:26:35 <danwent> So let's move on. 21:26:50 <danwent> I wanted to talk about the status of a few items that are still live, at least for a few more days 21:26:55 <danwent> jrd-redhat: rootwrap 21:27:01 <Salv-away_> Api does not implement per port routes 21:27:03 <jrd-redhat> Yes! 21:27:12 <danwent> Salv-away_: thanks, that's what I thought. 21:27:14 <jrd-redhat> danwent you want update? 21:27:28 <danwent> jrd-redhat: yes please. I did a review on friday. I think we're in good shape 21:27:49 <jrd-redhat> I think so too. Will feel more confident with more review, because I've been wrong before. 21:27:55 <jrd-redhat> But I believe it's ready to go. 21:28:00 <garyk> jrd-redhat: i have reviewed and tested with ovs + l3 agent and it looks good 21:28:17 <garyk> jrd-redhat: + dhcp agent 21:28:25 <danwent> Ok, for rootrwap, I'm going to say it needs to be merged by Thursday, or it doesn't make it. Is that OK? 21:28:27 <amotoki> I got it work with ovs + dhcp + l3 too. 21:28:48 <danwent> note, this is merged, not under review 21:28:48 <cdub_> danwent: just make sure it gets on the review list (or the non-list ;) 21:29:02 <jrd-redhat> danwent sure, as far as I'm concerned I'm on borrowed time already, as it took longwer than I expected. 21:29:13 <danwent> cdub_: garyk, how about you and I commit to getting that one merged? 21:29:26 <jrd-redhat> My understanding is if I can get another positive review from a core dev, then we're done, right? 21:29:40 <danwent> jrd-redhat: yes, and that can probably be me, I just need to re-review 21:29:47 <garyk> danwent: sure 21:29:50 <jrd-redhat> Ok. THanks for looking it over. 21:30:00 <danwent> rkukura: here? 21:30:03 <rkukura> yes 21:30:09 <danwent> provider networks branch 21:30:16 <danwent> i'm pretty worried here 21:30:18 <rkukura> it needs reviews 21:30:27 <danwent> Salv-away_: you are reviewing? 21:30:31 <rkukura> working fine in my multi-node tests of flat, vlan, and tunnel networks 21:30:34 <danwent> along with arosen? 21:30:52 <danwent> rkukura: ok, I think some reviewers had trouble getting it working, which this late in the game worries me. 21:31:03 <danwent> but I think they are working with you on this, perhaps it was a misconfig 21:31:06 <garyk> danwent: rkukura ; i am also looking at it. hopefully tomorrow i'll have some more input 21:31:20 <danwent> Ok, are you guys ok with a thursday drop-dead date? 21:31:26 <danwent> merged or its not int? 21:31:28 <danwent> in? 21:31:33 <rkukura> the most recent devstack patch helps get it configed 21:31:49 <danwent> rkukura: ok, that's still under review in devstack? 21:31:54 <rkukura> I've done all I can - it really needs to reviews 21:32:07 <rkukura> yes the devstack update is under review - they go together 21:32:14 <danwent> #todo #danwent identify two core devs for review of provider nets. 21:32:22 <cdub_> danwent: iiuc, it's pretty important for feature parity, so needs proper review attention 21:32:36 <danwent> ok, well, we have to make deicsions we're not happy with, so I'm going to stick a thursday drop-dead on this. 21:32:50 <danwent> cdub_: I agree, that's why its still alive at all :) 21:32:57 <cdub_> heh, ok ;) 21:33:11 <danwent> ok, final FFE, test agent 21:33:16 <rkukura> I'm OK with Thursday iff I get the comments tomorrow at latest 21:33:20 <danwent> nati_ueno: where does this stand 21:33:39 <nati_ueno> danwent: I'm writing unit tests. I'll push it to review in Today 21:33:39 <danwent> rkukura: ok, i'm going to do my best to wrestle up core dev resources on this, and will review it myself if needed. 21:33:47 <rkukura> thx 21:34:24 <danwent> nati_ueno: ok, thanks. this test agent is important for devstack gating, and since we want the gating working on Folsom, its important that this agent makes it into folsom 21:34:38 <danwent> nati_ueno: where are we review-wise? 21:34:55 <danwent> i know I was one core reviewer 21:35:09 <danwent> did we have another already (and I hopes it not garyk… he's already reviewing a lot) 21:35:12 <nati_ueno> danwent: What's meaing of review-wise? 21:35:31 <danwent> sorry, just trying to figure out if there's another core dev actively reviewing already 21:35:34 <danwent> it seems there is not 21:35:44 <danwent> (from looking at the review) 21:35:53 <danwent> anyone want to step forward here? 21:36:02 <danwent> otherwise, I guess this will have to drop as well. 21:36:07 <garyk> danwent: i can review but it will come at the expense of documentation 21:36:30 <danwent> garyk: you're on too many other key reviews, and I need you on docs, so I'd prefer not 21:36:43 <garyk> danwent: ok 21:36:44 <danwent> anyone other than salvatore or garyk? 21:37:14 <rkukura> I can take a look 21:37:24 <markmcclain> I can take look too 21:37:47 <danwent> ok, thanks guys. I'd probably prefer markmcclain, at least until the provider nets stuff is done, as rkukura, that will have to be your primary focus 21:37:54 <danwent> but if you want to review this as well, that's fantastic 21:38:11 <danwent> Ok. So the last topic is docs 21:38:26 <danwent> we haven't made progress on docs at all this past week, which is part of the reason I'm feeling we're so behind. 21:38:44 <danwent> ideally, I think our team would be spending time split between testing, bugfix/review, and docs 21:38:55 <danwent> garyk is going to help me on the admin docs side 21:39:06 <danwent> Salvatore isn't here, but I think he has some API docs for review 21:39:18 <danwent> #todo #danwent contact #salv-orlando about sending out API spec link 21:39:30 <danwent> garyk and I will be working on an outline for admin docs 21:39:46 <danwent> then we will ask certain people to help write the docs for features that they added. 21:40:16 <danwent> we'll also have to work closely with annegentle and the docs team, as quantum info will be going into a doc that is about all of openstack networking, including nova-network-ing 21:40:35 <danwent> we really need to focus on docs, as releasing an RC without docs is not very useful for people to test 21:41:05 <danwent> cdub_, rkukura , garyk are things looking good for fedora packaging? 21:41:07 <cdub_> danwent: what kind of help do hou need? 21:41:22 <danwent> cdub_: on the docs? 21:41:25 <cdub_> *nod* 21:41:34 <danwent> mainly writing content and creating diagrams 21:41:43 <rkukura> danwent: we've got work to do, and hope to make progress this week 21:41:58 <danwent> documenting install, basic use cases, advanced use cases, config options, etc. 21:42:12 <danwent> med_: you around? 21:42:31 <danwent> rkukura: ok, sounds like you'll target having packages by Sept. 4th though, for the fedora testing day? 21:42:40 <danwent> anyone from ubuntu or other distros listening in? 21:42:43 <mnewby> danwent: Oy, meant to be here earlier :( 21:42:58 <danwent> mnewby: http://wiki.openstack.org/Network/Meetings 21:43:08 <mnewby> danwent: I know, apologies. 21:43:09 <danwent> mnewby: can fill in details offline later 21:43:13 <mnewby> danwent: danke 21:43:35 <danwent> #todo #danwent contact ubuntu team about Folsom packaging for Quantum 21:43:50 <danwent> Oh, one other key topic I forgot 21:43:54 <danwent> devstack gating 21:44:08 <danwent> nati_ueno: gating is currently blocked due to a missing dependency? 21:44:27 <nati_ueno> danwent: It looks yes. netaddr looks not installed 21:44:29 <danwent> we should file a critical bug on that 21:44:53 <nati_ueno> danwent: yes 21:44:54 <danwent> I noticed that quantum does not even seem to have a file in file/apt directory of devstack 21:45:22 <danwent> #todo nati_ueno danwent make sure there's a devstack gating bug filed 21:45:23 <nati_ueno> danwent: I can't figure out why it was working. 21:45:34 <danwent> nati_ueno: python-netaddr is installed by nova 21:45:40 <danwent> so likely it was picking it up from there 21:45:46 <danwent> anyway, let's discuss on the bug 21:45:54 <nati_ueno> danwent: Ah I got it 21:45:56 <danwent> #topic open discussion 21:46:01 <danwent> anything else to talk about? 21:46:28 <cdub_> random...anybody in San Diego this week for LinuxCon/CloudOpen/Plumbers? 21:46:32 <danwent> remember, if you're a core dev, you really need to be putting cycles in in these next two weeks 21:46:44 <danwent> in terms of testing, docs, and reviews. 21:47:02 <nati_ueno> Should I avandon patches which is not in RC1 now? 21:47:34 <nati_ueno> or keep the review request until G starts? 21:47:35 <danwent> nati_ueno: you mean those that are already on gerrit? I'm fine if you just -2 them with a note saying they should not be looked at until the RC drops and master opens for grizzly 21:47:51 <nati_ueno> danwent: I got it 21:47:52 <mnewby> danwent: Have cycles to spare, will be focusing on reviewing unless something else has priority. 21:47:52 <danwent> core devs can know to not look at patches that have a -2 21:48:04 <danwent> mnewby: great. that will be very helpful, as we have a backlog 21:48:11 <nati_ueno> danwent: It looks I can't -2 21:48:18 <danwent> hehe, my bad. 21:48:35 <mnewby> danwent: will also be working on tempest tests for quantum. need tempest support for internal deliverable. will coordinate with you on that. 21:48:36 <danwent> ok, i will do it for you. I will be doing this for all non RC1 patches later today anyway 21:48:44 <nati_ueno> danwent: Thanks 21:48:44 <danwent> mnewby: that would be REALLY cool. 21:49:07 <danwent> mnewby: nati_ueno is working on some devstack tests, so would be good to coordinate with him about coverage 21:49:27 <garyk> guys, i am really beat i am going to crash. good night 21:49:30 <danwent> mnewby: from talking to CI team, it seems like gating checks are supposed to move to all tempest in the future 21:49:37 <danwent> garyk: ok, get some rest :) 21:50:01 <danwent> mnewby: so if there is a question of where effort is best placed, its probably tempest 21:50:05 <mnewby> danwent: roger that. the exercise tests and tempest would seem to have the same aim. 21:50:06 <danwent> ok, any other open discussion? 21:50:16 <nati_ueno> garyk: Good night 21:50:22 <mnewby> garyk: 'nite! 21:50:23 <danwent> mnewby: yes, goal is to make exercies scripts only simple tutorials, not real tests. 21:50:36 <med_> danwent, I'm around now. 21:50:38 <danwent> ok, anything else before we can all say good night/afternoon/morning? 21:50:44 <danwent> med_: ok, let's chat after meeting 21:50:48 <med_> sure. 21:51:06 <danwent> ok, thanks folks. remember, two key weeks to folsom RC1 for quantum :) 21:51:11 <danwent> #endmeeting