15:00:38 <gordc> #startmeeting telemetry 15:00:38 <openstack> Meeting started Thu Jan 28 15:00:38 2016 UTC and is due to finish in 60 minutes. The chair is gordc. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:39 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:42 <openstack> The meeting name has been set to 'telemetry' 15:00:56 <jd__> o/ 15:01:18 <idegtiarov> o/ 15:01:35 <gordc> ... very quick meeting? 15:01:56 <llu-laptop> o/ 15:01:59 <pradk> o/ 15:02:23 <gordc> let's blame through this 15:02:24 <_nadya_> o/ 15:02:27 <ildikov> o/ 15:02:37 <sileht> o/ 15:02:41 <gordc> #topic roadmap items (new/old/blockers) https://wiki.openstack.org/wiki/Telemetry/RoadMap 15:02:42 <ityaptin> o/ 15:02:53 <gordc> ok. so we are in final stretch 15:03:11 <gordc> i have list of Roadmap items listed... if someone still has open time 15:03:20 <gordc> but i have nothing really to add. any blockers? 15:04:11 <gordc> cool. nothing 15:04:18 <gordc> #topic aodh topics 15:04:23 <gordc> anything for aodh? 15:04:41 <_nadya_> yep! I have a question about mandatory limits 15:04:48 <gordc> _nadya_: kk 15:05:20 <_nadya_> so, I don't think we need it in aodh. Also, it affects QAs and customers very much 15:05:27 <gordc> #topic aodh mandatory limits 15:06:22 <_nadya_> the main purpose of mandatory limits in Ceilo was not to show too many resources and not create a load on dbs 15:06:25 <gordc> _nadya_: qa because they need to update default return setting? 15:06:58 <gordc> _nadya_: so i'm pretty indifferent to mandatory limits... are we against adding limit completely though? 15:07:24 <_nadya_> gordc: yes. so you have a test with expectation that 10001 resource should be returned, but it starts return only 10000 15:07:24 <llu-laptop> So the question is that how many alarms typically would return in a query for production? 15:07:38 <liusheng> o/ 15:07:51 <gordc> _nadya_: well we can have a limit but make it optional? 15:08:03 <gordc> liusheng: do you know how many alarms we have? 15:08:16 <gordc> i guess per resource it becomes a lot? 15:08:21 <_nadya_> gordc: yep, I don't mind to have -l option, but not mandatory 15:08:53 <ildikov> I think it's better to keep the option, but I agree to skip the mandatory part 15:09:03 <jd__> there's a standard in OpenStack, let's use it 15:09:04 <liusheng> gordc: don't have a exact num 15:09:17 <gordc> jd__: what's the standard? 15:09:31 <gordc> jd__: this seems like taht pagination craziness again. 15:09:38 <jd__> the one we use in Gnocchi AFAIK, that is backed by oslo.db features 15:09:40 <llu-laptop> jd__: the standard for limit is optional or mandtary? 15:10:03 <liusheng> gordc: but seems about several thousands 15:10:16 <jd__> gordc: that's that, but it's standardized AFAIK now – at least for sql 15:10:41 <gordc> jd__: is pagination in oslo.db? 15:10:56 <jd__> yes AFAIK 15:10:58 <jd__> sileht knows 15:11:08 <gordc> the problem is we aren't killing mongo/hbase for aodh just yet 15:11:43 <gordc> so... maybe we should officially mark them deprecated 15:11:57 <jd__> yes, we merged a spec for that 15:11:59 <gordc> and then start applying all the oslo.db pagination easy stuff afterwards. 15:12:02 <jd__> and just not implement the feature there 15:12:35 <ildikov> gordc: I will add docco if we deprecate Mongo and HBase now 15:13:08 <gordc> jd__: i'm ok with adopting oslo.db pagination. we just need to formally say: 'mongo/hbase aodh is what it is' 15:13:17 <gordc> and maybe have a v2.1 api or something 15:13:27 <jd__> no need AFAIK since it's just addition 15:13:47 <gordc> something to consider. 15:14:08 <gordc> so we all ok with prioritising deprecation spec 15:14:24 <_nadya_> I'm not sure I wanted this.... 15:14:28 <gordc> and we can apply oslo.db rules for api limits in aodh? 15:14:39 <gordc> _nadya_: the deprecation? 15:15:00 <_nadya_> gordc: yep :) So we will have only one db option for Aodh? 15:15:14 <gordc> _nadya_: yes. 15:15:28 <gordc> let's be honest here, our resources are stretch too thin as is. 15:15:45 <_nadya_> gordc: agreed 15:15:49 <gordc> unless someone yells "i'm using mongo/hbase and i will work on it" 15:15:52 <liusheng> gordc: the pagination in oslo.db can surppot limit, marker, sort, don't have a default limit confg option 15:15:55 <gordc> i don't see how we can maintain it. 15:16:27 <gordc> liusheng: so it could solve our concerns still? 15:17:02 <gordc> _nadya_: i think i consider it like zmq or qpid in oslo.db. 15:17:07 <sileht> gordc, you just have to set it like this: https://github.com/openstack/gnocchi/blob/master/gnocchi/indexer/sqlalchemy.py#L526 15:17:10 <_nadya_> gordc: people usually use the same backend for alarming as for metering. but probably, it is not a blocker. It just a tradition 15:17:13 <gordc> it's there. but no one gives a damn about it so probably don't use it. 15:17:34 <gordc> or at least for qpid 15:17:48 <gordc> sileht: ah cool. 15:17:49 <gordc> nice. 15:18:20 <gordc> sileht: qpid still exists right? in oslo.messaging? 15:18:27 <gordc> it just not tested or worked on? 15:18:35 <sileht> gordc, no it have been removed 15:18:38 <ildikov> _nadya_: I assume it's the easiest to handle, but it's prolly not a blocker 15:18:39 <gordc> oh.lol 15:18:48 <sileht> the 1 cycle deprecation have been done 15:18:54 <_nadya_> ildikov: yep, tend to agree 15:18:57 <gordc> yeah. so i guess we do the same. 15:19:09 <gordc> we deprecate and if no one steps up. it's gone? 15:19:24 <gordc> for me i'd rather have one working rather than 3 broken options. 15:19:53 <ildikov> _nadya_: although if we go into this direction I assume we will hear complainings from ops if it wasn't a that good idea 15:20:38 <gordc> ildikov: yep. and then they can go to their boss and let them know backend support don't grow on trees 15:20:46 <gordc> and then we get more resources :) 15:20:48 <jd__> gordc++ 15:20:49 <llu-laptop> if we want to deprecate mongodb/hbase, maybe we should write an tool to help migrate them into sql? 15:21:03 <jd__> llu-laptop: no, same answer than gordc just gave 15:21:15 <ildikov> gordc: lol :) agreed 15:21:32 <ildikov> gordc: also the code is still up on github it can be picked up later... 15:21:44 <gordc> llu-laptop: yeah. if there's a request... we're not at point where we can give false hope. :) 15:21:52 <_nadya_> jd__: why not? It should be pretty easy 15:22:06 <gordc> ildikov: ++ 15:22:06 <_nadya_> false hope, ok 15:22:15 <liusheng> llu-laptop: I have tried that, but I am in busy for other things these days :( 15:22:16 <gordc> _nadya_: yeah, all about resources. 15:22:19 <jd__> _nadya_: sure, you'll do it? 15:22:40 <gordc> _nadya_: we can backlog it. 15:22:50 <_nadya_> jd__: ok, I understood :) Anyway, I think I will 15:23:08 <gordc> i should create a section that says: 'pay someone, tell them to do this, and it will happen' 15:23:10 <jd__> _nadya_: taking your word for it :) 15:23:39 <gordc> although, that's basically the enitre backlog 15:23:48 <_nadya_> jd__: my name can be translated in english as "hope" :) 15:23:57 <gordc> :) we all good? 15:23:59 <llu-laptop> gordc: yeah, let's backlog that and encourage new comer to work on that. 15:24:19 <gordc> #action backlog mongo/hbase to sql tool 15:24:39 <gordc> #action prioritise db deprecation 15:24:53 <gordc> #action adopt oslo.db for pagination/limits 15:24:57 <jd__> _nadya_: good to know lol 15:25:13 <gordc> cool. anything else? 15:25:33 <jd__> llu-laptop: all those shitty tasks in the backlog that nobody wants to work on, you call that encouragement. remind me to never work for you ;) 15:26:56 <gordc> jd__: intel has a lot of money... you may eat your words 15:27:06 <gordc> #topic ceilometer topics 15:27:59 <gordc> i'm going to assume nothign for this? 15:28:01 <_nadya_> can I merge this nova polling improvements https://review.openstack.org/#/c/209799/9? 15:28:13 <jd__> gordc: don't confuse the man and the company :) 15:29:58 <gordc> _nadya_: sorry, was readnig liushengs comment 15:30:01 <gordc> i'm ok merging. 15:31:12 <gordc> _nadya_: i'm going to edit to make liushengs comment address my comment but i'll push merge or fix later 15:31:29 <gordc> _nadya_: feel free to merge now if you like 15:31:36 <_nadya_> gordc: ok! 15:31:39 <gordc> #topic gnocchi topics 15:32:05 <jd__> I just released 1.3.4 with a bunch of interesting fixes 15:32:05 <gordc> jd__: anything? or just check openstack-telemetry logs to see all the comments? 15:32:33 <gordc> jd__: we haven't released the split timeserie patch yet right? 15:32:38 <jd__> now I'm planning on 2.0 in the next weeks, once dust settles 15:32:43 <jd__> gordc: no that'd be in 2.0 15:32:49 <gordc> jd__: kk. 15:32:52 <jd__> whereas my plan to release it in the upcoming weeks 15:33:04 <llu-laptop> jd__: i'm not sure stevelle has pinged you or not about the ansible support problem of gnocchi with swift 15:33:10 <gordc> good to see what bugs we get from nicodemus_ 15:33:13 <jd__> next we'd release 2.1 with sileht work on dynamic resource type, probably around mitaka I guess 15:33:19 <jd__> llu-laptop: he didn't 15:33:25 <gordc> jd__: make sense. 15:33:41 <jd__> then the dynamic resource type support will have to got in Ceilometer for N I guess :) 15:33:58 <jd__> other than that, project is in really good shape so far 15:35:50 <gordc> jd__: cool cool. 15:36:05 <llu-laptop> jd__: the problem is that if goncchi saves data into swift, the ceilometermiddleware will send out notifications which are turned into metrics to be published into gnocchi 15:36:11 <llu-laptop> a loop 15:36:15 <gordc> jd__: hoping lab is ready next week and i can start etsting in a few weeks. 15:36:23 <gordc> llu-laptop: yeah. 15:36:29 <gordc> you can filter it in ceilometermiddleware 15:36:38 <gordc> or just disable ceilometermiddleware. 15:36:40 <sileht> llu-laptop, a configuration flags exists to filterout the gnocchi account 15:36:50 <gordc> or create another proxy-server? 15:37:02 <jd__> what sileht said 15:37:09 <llu-laptop> sileht: configuration flag in ceilometermiddleware? 15:37:17 <jd__> in Gnocchi 15:37:31 <gordc> sileht: i think it needs to happen at proxy-server rather than dispatcher. 15:37:44 <sileht> llu-laptop, https://github.com/openstack/ceilometermiddleware/blob/master/ceilometermiddleware/swift.py#L39 15:38:00 <sileht> llu-laptop, or https://github.com/openstack/ceilometer/blob/master/ceilometer/dispatcher/gnocchi.py#L47 15:38:09 <sileht> depending of your needs 15:38:45 <gordc> llu-laptop: i'd recommend the ceilometermiddleware. the dispatcher filter will still hammer queue. but yeah, you have options 15:40:09 <gordc> any other items? 15:40:46 <gordc> #topic open discussion 15:41:13 <gordc> just a heads up, feb 1 is last day to submit speaker talks for summit (if you plan to do antyhing) 15:41:19 <gordc> that's all i have. 15:42:51 <gordc> let's call it. 15:42:54 <gordc> thanks folks. 15:42:57 <gordc> #endmeeting