15:00:21 <eglynn> #startmeeting ceilometer 15:00:28 <openstack> Meeting started Thu Jan 15 15:00:21 2015 UTC and is due to finish in 60 minutes. The chair is eglynn. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:32 <openstack> The meeting name has been set to 'ceilometer' 15:00:32 <sileht> o/ 15:00:43 <DinaBelova> o/ 15:00:50 <llu-laptop> o/ 15:00:54 <cdent> o/ 15:00:59 <ildikov> o/ 15:01:01 <fabiog> hello 15:01:06 <eglynn> hey folks 15:01:10 <eglynn> #topic Kilo-2 status 15:01:10 <nealph> g'morning 15:01:17 <mesterm> morning 15:01:23 <eglynn> #link https://launchpad.net/ceilometer/+milestone/kilo-2 15:01:31 <_elena_> o/ 15:01:51 <eglynn> we've made some progress in landing specs reviews 15:02:05 <eglynn> still a couple of BPs marked as blocked though 15:02:20 <idegtiarov> o/ 15:02:37 <yeungp> o/ 15:02:54 <eglynn> ceilometermiddleware has a +2 on https://review.openstack.org/142129 so should land soon 15:03:18 <DinaBelova> eglynn, hm there is one bug with my +2 - waiting for more cores https://bugs.launchpad.net/ceilometer/+bug/1372442 15:03:30 <DinaBelova> somehow I though it was already landed 15:03:35 <DinaBelova> thought* 15:03:43 <eglynn> DinaBelova: cool, I'll look after meeting 15:03:48 <DinaBelova> eglynn, thanks 15:03:54 <eglynn> might have to bump https://blueprints.launchpad.net/ceilometer/+spec/instance-autorestart if there's not more traction on it soon-ish 15:04:37 <eglynn> nealph: were you gonna to file a BP for the DB pipeline config part 1? 15:04:52 <eglynn> nealph: ... IIRC we discussed splitting this over kilo-2 and kilo-3 last week 15:05:05 <nealph> eglynn: sorry, I thought https://blueprints.launchpad.net/ceilometer/+spec/ceilometer-configuration-via-data-store was sufficient? 15:05:13 <eglynn> nealph: cool, thanks! 15:05:33 <nealph> eglynn: added the linked https://blueprints.launchpad.net/ceilometer/+spec/configdb-api 15:05:39 <nealph> as discussed. 15:05:45 <nealph> for the k3 work. 15:06:19 <eglynn> nealph: excellent thanks ... it was just missing the targeting to kilo-2, done now 15:06:20 <DinaBelova> nealph, a-ha, as I remember you've discussed the work to be done as a part of https://blueprints.launchpad.net/ceilometer/+spec/configdb-api with idegtiarov 15:06:21 <nealph> eglynn: my launchpad skillz are negligible...let me know if I need to do additiona. 15:06:52 <DinaBelova> nealph, let's I guess assign it to idegtiarov? 15:07:17 <eglynn> DinaBelova: the config API work assigned to be to idegtiarov? 15:07:38 <nealph> eglynn: DinaBelova: let's hold off, I think fabiog and I need to chat offline on that... 15:07:42 <DinaBelova> eglynn, at least idegtiarov and nealph had the discussion (succeded as far as I remember) about it 15:07:47 <DinaBelova> nealph, ok 15:07:50 <eglynn> nealph: cool 15:08:06 <DinaBelova> it looks like I got it wrong a little bit 15:08:09 <DinaBelova> :) 15:08:30 <nealph> DinaBelova: no, no, I think I jumped the gun on that. my fault. 15:08:46 <DinaBelova> nealph, np :) 15:09:15 <eglynn> cool, anything else on kilo-2? 15:09:34 <cdent> gabbi stuff is pending acceptance into global requirements but that seems likely by the endof this week 15:09:49 <eglynn> cdent: excellent :) 15:10:27 <eglynn> #topic gnocchi update 15:10:47 <eglynn> interesting thread about gnocchi performance on the ML 15:10:50 <eglynn> #link http://lists.openstack.org/pipermail/openstack-dev/2015-January/054534.html 15:11:11 <cdent> I think it's not clear to the casual observer that carbonara is a dev environment 15:11:33 <DinaBelova> eglynn, the only issue with this thread (for me, at least) is the name of mail-sender :) 15:12:10 <DinaBelova> cdent, I guess the overall issue with gnocchi now is lack of up-to-date documentation 15:12:17 <DinaBelova> with these moments described at well 15:12:18 <jd__> I'm a bit worried about the slow progress wrt Ceilometer integration 15:12:28 <jd__> we didn't do what we talked about for k1 and k2 is approaching without much progress 15:12:39 <cdent> jd__ makes very good points 15:12:52 <jd__> DinaBelova: the documentation is up-to-date, but it might be incomplete :) 15:13:00 <DinaBelova> jd__, yeah, sorry :) 15:13:04 <DinaBelova> wrong words 15:13:10 <jd__> you stand corrected ;) 15:13:16 <DinaBelova> lol 15:13:38 <jd__> other than that I'm pretty happy with the progress we made so far 15:14:02 <eglynn> jd__: I'll have a bit more time to contribute directly now that the juno-derived distro is mostly complete 15:14:02 <jd__> and I'm gonna make efforts the next months to bring more attention to Gnocchi in general 15:14:17 <jd__> eglynn: you see me glad and happy to hear that 15:14:35 <DinaBelova> jd__ - I'm kind of still concerned with one moment I've probably mentioned before about gnocchi -> ceilo integration. It's kind of very useful to store history of changes somehow... 15:14:46 <DinaBelova> and I see no simple way for doing this 15:14:53 <eglynn> jd__: can we put together a list of high priority gnocchi tasks over the next week? 15:14:56 <DinaBelova> (if we'll merge repos I mean) 15:15:09 <gordc> DinaBelova: history of policy changes? 15:15:19 <eglynn> DinaBelova: git history do you mean? 15:15:25 <DinaBelova> gordc, I would way git history 15:15:27 <DinaBelova> eglynn, yeah 15:15:43 <jd__> DinaBelova: I don't think merging repository is on the table righ tnow 15:16:02 <DinaBelova> jd__, I'm mostly about some future 15:16:23 <jd__> and I'm pretty sure now that it's not gonna happen 15:16:24 <eglynn> DinaBelova: the stackforge/gnocchi repo could continue to exist in a "frozen" state after any merge (as a "fossil record" of the history) 15:16:59 <DinaBelova> eglynn, well, yeah, probably 15:17:01 <jd__> there's like 0 point of merging an entire independant subsystem in a bigger one 15:17:20 <DinaBelova> jd__, well, if so, that's not the big deal 15:17:35 <DinaBelova> jd__, I just remember discussions about this before 15:17:35 <eglynn> yeah, having multiple repos is less of an issue in my mind than having separate core teams 15:17:40 <jd__> right now the big deal is to make Gnocchi and Ceilometer work completely together, which is not yet done 15:18:07 <jd__> eglynn: though we have separate core team in Oslo which is not a problem in the end 15:18:20 <jd__> but it's not a problem we have right now 15:18:33 <DinaBelova> nellysmitt and idegtiarov are working on adding all resources (changes coming...) 15:18:40 <eglynn> it would be good for more ceilo cores to get involved in gnocchi reviewing, at least, IMO 15:18:45 <jd__> DinaBelova: ah good to know, I didn't 15:19:03 <DinaBelova> jd__, well, they had lots of testing env issues... 15:19:03 <jd__> eglynn: you said that ~2 months ago and that didn't happen, as I unfortunately expected :) 15:19:04 <eglynn> nellysmitt: was your OPW project topic finalized? 15:19:20 <gordc> eglynn: i sense a subtle hint there. 15:19:29 <jd__> I think we need to bring more attention on Gnocchi (and Ceilo) to have more people (even external to Ceilometer) starting to contribute 15:19:41 <eglynn> gordc: ... subtle, moi? ;) 15:19:47 * gordc will try to review today... was meaning to but got sidetracked. 15:19:48 <DinaBelova> eglynn, not yet - as I had long Xmas holidays it was frozen 15:19:53 <nellysmitt> eglynn, no project topic yet 15:19:59 <DinaBelova> I found one really interesting task although 15:20:10 <cdent> you're right, of course, jd__ but what if we've asked for time and haven't been able to getit? 15:20:11 <DinaBelova> with making new resource adding more openstacky to gnocchi 15:20:31 <DinaBelova> eglynn, as now you need to go to 4 different places withough any stevedore, etc. to add new resource 15:20:31 <jd__> cdent: I'm not blaming anyone, don't read me wrong 15:20:39 <eglynn> "more openstacky"? 15:20:48 <DinaBelova> eglynn, do you think it's a nice something to be done for gnocchi? 15:21:00 <cdent> Oh, I don't think you are, I'm saying that as a group we need to agitate up the hierarchy to inform them of the need. 15:21:03 <eglynn> "more openstacky" == "fitting in with the usual openstack idioms"? 15:21:09 <jd__> cdent: I'm actually not expecting ceilometer-core to review gnocchi, I think I would have better change to bring people from the outside of OpenStack right now 15:21:09 <cdent> So that the need can be seen to be relevant 15:21:36 <DinaBelova> eglynn, more openstacky == using logical things like stevedore, etc. 15:21:49 <mesterm> Jd i plan on spending alot more time in the review of gnocchi -> ceil 15:21:56 <eglynn> jd__: do you mean domain experts in metrics stroage? 15:22:02 <eglynn> jd__: ... how about trying to get pauldix involved? 15:22:29 <jd__> eglynn: pretty sure pauldix has other cats to whip, but I'd love to 15:22:44 <jd__> eglynn: no I mean people who would start using Gnocchi to store metrics and who would have interest in it in general 15:23:00 <jd__> you know, users who become developers, that's how FOSS work most of the time ;) 15:23:00 <DinaBelova> eglynn, do you thing some gnocchi resource adding logic improvement like I've mentioned will be nice thing for Nelly to work with? 15:23:01 <eglynn> jd__: ok 15:23:17 <gordc> mesterm: sounds good. thanks for reviewing 15:23:40 <eglynn> DinaBelova: it would certainly be useful, but it there enough work for a 13-week internship? 15:24:10 <DinaBelova> eglynn, nope I suppose.... Although I don't have better ideas now 15:24:25 <eglynn> nellysmitt: how far thru' your 13 weeks are you now? 15:24:42 <nellysmitt> 1 month + 1 week 15:24:51 <DinaBelova> eglynn, it you have some it'll be really cool 15:25:00 <eglynn> so 9 weeks left, ok 15:25:17 <nellysmitt> yeah, but I lost a lot of time solving issues 15:25:27 <nellysmitt> which were caused by hardware :( 15:25:49 <eglynn> that's unfortunate 15:25:54 <eglynn> BTW the final patches do not have to be landed within the internship period, as long as there's been good progress 15:26:39 <eglynn> the previous OPW intern for ceilo continued chipping away at her patches for a few months and finally got most of it landed 15:26:40 <DinaBelova> eglynn, I had also some idea about implementing driver for https://github.com/kairosdb/kairosdb 15:26:57 <DinaBelova> that's tsdb over cassandra 15:27:18 <DinaBelova> eglynn, althoguh that'll be going kind of really tricky 15:27:40 <eglynn> ok, that could be interesting, though we'd have to be careful about proliferation of drivers 15:28:01 <DinaBelova> eglynn, yeah, that's also the issue 15:28:11 <eglynn> DinaBelova: is there any more contained feature of the OpenTSDB driver that could be split off? 15:28:33 <fabiog> eglynn: I am working on some API and datastructure tweaks to improve batching greatly 15:28:33 <eglynn> DinaBelova: ... e.g. the kind of async expiry/downsampling logic you spoke about before? 15:28:41 <eglynn> fabiog: cool 15:28:57 <DinaBelova> eglynn, python code is almost ok now, as for the retention/downsampling I was already working on it - and that's pure Java 15:28:57 <fabiog> eglynn: I am planning to do a Google Hangout sometimes soon to check the approach is viable 15:29:40 <eglynn> DinaBelova: a-ha, ok 15:29:53 <eglynn> fabiog: sounds interesting 15:29:58 <fabiog> eglynn: and jd__ retention policies are per project/user combination 15:30:21 <fabiog> is that really necessary? what is the usecase of retaining a resource longer for a specific user within the same project? 15:30:58 <jd__> fabiog: API tweaks on Gnocchi? 15:30:58 <eglynn> fabiog: you've lost me ... there's a single ArchivePolicy per Metric, no? 15:31:29 <DinaBelova> eglynn, well, may you please spend some time on if KairosDB driver will be the nice task? 15:31:33 <fabiog> eglynn: yes, but metric is linked to project/user combination through resources, isn't it the case? 15:31:51 <DinaBelova> eglynn they have python client as well 15:32:05 <fabiog> jd__: yes I want to decouple measures from metrics so we can batch them as they come without having to group them by metric 15:32:33 <jd__> fabiog: you mean sending measures for several metrics in one HTTP request? 15:32:38 <eglynn> fabiog: so you meant retaining data for the same Metric name (e.g. cpu_util) for different resources of the same type (e.g. instances) for different periods? 15:32:39 <fabiog> yes 15:32:52 <fabiog> jd__: yes 15:33:12 <jd__> fabiog: ok, I'm curious as I'm not sure it'd be an improvement, but I'd be glad to discuss if you've a patch or proto or something 15:33:35 <jd__> eglynn: feel free to continue the agenda if there's any btw :) 15:33:46 <fabiog> jd__: give me a week or so and I will prepare a presentation 15:34:04 <eglynn> fabiog: ok, that retention flexiblity is provided by the gnocchi API ... but in practive I suspect ceilo will not usually be using that degree-of-freedom 15:34:10 <fabiog> jd__: I was meant to do that at the mid-cycle ... 15:34:24 <eglynn> fabiog: ... unless for example some subset of resources are consider higher priority than others 15:34:43 <eglynn> fabiog: ... e.g. running a retail site, as opposed to build farm 15:34:53 * nealph thinks that's an interesting concept 15:34:53 <fabiog> eglynn: right, but this is the cause of the batching restrictions IMHO 15:35:23 <eglynn> fabiog: a-ha, so you don't want to batch across metrics with different ArchivePolicies? 15:35:57 <eglynn> OK, this sounds like an interesting subject that we could debate at length on a hangout or a specs review or some-such 15:36:02 <fabiog> eglynn: wait ... I want to batch measures independently from its metrics 15:36:25 <eglynn> fabiog: OK, I'm missing the relevance of the variable retention policy in that case 15:37:03 <fabiog> eglynn: it is better I prepare my presentation and schedule a meeting. It is hard to explain it in IRC. I need pictures, data structures and so on :-) 15:37:14 <eglynn> fabiog: coolness, that makes sense 15:37:27 <eglynn> ... I look forward to getting an invite :) 15:37:54 <eglynn> k, let's move on, a few housekeeping items to cover 15:37:57 <eglynn> #topic Miscellaneous housekeeping 15:38:24 <eglynn> for those who didn't hear, we decided to cancel the midcycle meetup 15:38:45 <eglynn> ... as we didn't reach a quorum of contributors committed to attend 15:39:16 <eglynn> ... but let's try to cover the topics via remote collaboration instead over the coming weeks 15:40:17 <eglynn> there was a thread on the ML from ttx about changes to the L* design summit format 15:40:21 <eglynn> #link http://lists.openstack.org/pipermail/openstack-dev/2015-January/054122.html 15:40:49 <eglynn> so we, as a project, will need to express some preferences for how we'd like to accomodated at summit 15:42:11 <eglynn> IIRc basically a mix of ... 15:42:19 <eglynn> (a) fishbowl-type sessions in larger spaces 15:42:29 <eglynn> (b) more focussed working sessions in smaller spaces 15:42:37 <eglynn> ... and (c) free-flowing contributor meetup style sessions 15:42:58 <eglynn> I'm thinking (a) is more appropriate for the large projects e.g. nova & neutron 15:43:21 <eglynn> which leaves us with (b) and possibly (c) as the best option for ceilometer? 15:43:29 <llu-laptop> eglynn: agreed 15:43:46 <DinaBelova> eglynn, I think so as well 15:43:56 <llu-laptop> I don't think we need 100 to 200 people for fishbowl 15:44:03 <eglynn> llu-laptop: I agree 15:44:06 <ildikov> eglynn: yeap, I don't feel myself as a fish either :) 15:44:30 <eglynn> so any objections to requesting mainly (b) for the ceilometer track? 15:44:44 <mesterm> How many sessions on ceil are we thinking? 15:44:52 <llu-laptop> just a question, if we don't apply for fishbowl, do we get enough timeslot for the working sessions we requires? 15:46:04 <mesterm> I am hoping there are some that are somewhat ops/implementation focused to address existing issues 15:46:08 <gordc> cool with me 15:46:34 <gordc> ^ both eglynn and mesterm suggestions 15:46:43 <eglynn> mesterm: historically we've gotten around 8-10 sessions ... but I suspect the competition for space will be more intense in the L* summit 15:47:01 <jd__> (b) sounds fine 15:47:23 <ildikov> (b) looks good to me too 15:47:26 <mesterm> eglynn: sounds good 15:47:30 <eglynn> llu-laptop: I think applying for fishbowl time would only be appropriate if we really feel we can fill the space 15:47:56 <eglynn> BTW we had been jokingly calling the L* cycle "Lemming" 15:48:13 <eglynn> ... apparently though that's been removed from the list of candidate names 15:48:33 <eglynn> ... too easy for the haters to make jokes at our expense, apparently 15:49:25 <eglynn> Love, London, Liberty & Lizard on the list instead 15:49:38 <eglynn> my money is on Lizard 15:49:40 <cdent> Lizard 15:49:47 <eglynn> snap :) 15:49:53 <gordc> eglynn: lemmings are adorable now that i've googled it 15:49:58 <llu-laptop> love seems a little bit weird 15:50:06 <mesterm> what about liver.... everyone loves liver 15:50:15 <ildikov> gordc: LOL :) 15:50:17 * nealph grimaces 15:50:33 <eglynn> mesterm: ... preferably chopped :) 15:51:02 <fabiog> eglynn: but are we moving away from being something from the place, e.g. steet, neighborough etc ... 15:51:31 <fabiog> fabiog: I am wondering if London is a street in Vancouver 15:51:50 <eglynn> fabiog: I think it's still primarly placename-based, maybe loosening the definition a bit though 15:52:02 <eglynn> last item of housekeeping ... Use of spec proposal and feature proposal deadlines during kilo-3? 15:52:19 <eglynn> Thierry wants to have a lits of the deadlines each project is aiming for 15:52:52 <fabiog> eglynn: can you post the schedule again, please? 15:53:16 <llu-laptop> feb 5th kilo-2, 15:53:18 <eglynn> #link https://wiki.openstack.org/wiki/Kilo_Release_Schedule 15:53:25 <fabiog> eglynn: thanks 15:53:26 <llu-laptop> mar 19th, kilo-3 15:53:34 <eglynn> how about ... feature proposal freeze two weeks out from kilo-2, spec proposal freeze 4 weeks out? 15:54:39 <llu-laptop> feb 19th and mar 5th? 15:54:54 <gordc> works for me... 15:54:57 <eglynn> yeah, that sounds right 15:54:59 <eglynn> cool 15:55:06 <llu-laptop> good to me 15:55:15 <ildikov> looks ok 15:55:50 <eglynn> #topic open discussion 15:56:12 <eglynn> anyone want to raise anything in the remaining few minutes? 15:57:05 <mesterm> working remote during what was the mid-cycle will be done maybe via etherpad? 15:57:48 <eglynn> mesterm: etherpads are one way, also fabiog is proposing a g+ hangout for his topic 15:58:07 <fabiog> mesterm: I think the person that proposed the topic will be in charge of organizing something 15:58:15 <eglynn> fabiog: agreed 15:58:28 <mesterm> i just dont want to miss anything :) 15:59:09 <eglynn> cool 15:59:10 <fabiog> mesterm: I will put my presentation schedule as topic for next meeting 15:59:28 <mesterm> fabiog: thanks 16:00:02 <eglynn> ok folks, outta time, let's call this a wrap ... thanks for your time! 16:00:11 <eglynn> #endmeeting ceilometer