15:01:08 <gordc> #startmeeting telemetry 15:01:09 <openstack> Meeting started Thu Dec 3 15:01:08 2015 UTC and is due to finish in 60 minutes. The chair is gordc. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:10 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:13 <openstack> The meeting name has been set to 'telemetry' 15:01:16 <nijaba> o/ 15:01:22 <ildikov> o/ 15:01:30 <llu> o/ 15:01:33 <r-mibu> o/ 15:01:52 <jd__> o/ 15:02:11 <cdent> o/ 15:02:13 <gordc> lets' start 15:02:22 <gordc> cdent: welcome back 15:02:29 <cdent> thanks 15:02:41 <gordc> #topic recurring: roadmap items (new/old/blockers) https://wiki.openstack.org/wiki/Telemetry/RoadMap 15:02:58 <gordc> just a quick update, Mitaka-1 is in processed of being tagged 15:03:24 <gordc> https://wiki.openstack.org/wiki/Mitaka_Release_Schedule 15:03:34 <gordc> this is your release cycle in case you're interested 15:03:59 <gordc> we should probably get started on reviewing specs so people can start working on them. 15:04:29 <gordc> make sense? 15:04:43 <r-mibu> ack 15:04:58 <ildikov> yeap 15:05:00 <jd__> k 15:05:12 <gordc> cool. feel free to mention any spec related items as we run through each project 15:05:28 <gordc> #topic aodh topics 15:05:39 <gordc> any aodh topics we want to address? 15:05:49 <gordc> i know there's a thread relating to vitrage 15:05:59 <gordc> i assume that's still being discussed but no issues? 15:06:01 <jd__> indeed 15:06:02 <pradk> o/ 15:06:08 <ildikov> I'm still reading it 15:06:25 <Guest51944> o/ 15:06:49 <r-mibu> yeah, it's interesting 15:07:13 <gordc> jd__: it's weird regarding 1 min alarm creation 15:07:25 <jd__> gordc: it's just the cache ttl 15:07:37 <ildikov> yeah, it's the time for creating an alarm 15:07:42 <gordc> ah, yeah. that makes sense. 15:07:56 <r-mibu> yeah, that's my bad explanation though 15:07:59 <gordc> i assume it was related to polling 15:08:27 <gordc> cool. i should mention aodh mitaka-1 release is also being tagged today 15:08:42 <r-mibu> ok 15:09:08 <gordc> did we have an owner for aodhclient? i can't remember if we had volunteer from summit 15:09:11 * jd__ sighs 15:09:47 <r-mibu> i will try after tempest 15:09:47 <idegtiarov> o/ 15:09:55 <gordc> jd__: we'll track it for now. start worrying in the new year 15:10:15 <gordc> any other aodh related topics? 15:10:42 <r-mibu> no, i guess 15:10:45 <gordc> cool 15:10:50 <gordc> #topic Monasca publisher blueprint discussion 15:11:00 <gordc> rohit_: i assume this is you :) 15:11:40 <rohit_> Thx gordc 15:11:56 <gordc> #link https://review.openstack.org/#/c/249406/ 15:12:08 <gordc> fo reference 15:13:15 <rohit_> Question on the v2 api - I assume it's not deprecated now? 15:14:07 <_nadya_> no, it's not 15:14:32 <gordc> v2 metering api is basically in maintenance. 15:14:34 <fabiog> _nadya_: and the process hasn't even started yet, right? 15:14:52 <_nadya_> fabiog: AFAIK, no, it's not 15:15:11 <gordc> i don't want to say deprecated because it seems people are using it (to what success i'm unsure) 15:15:35 <gordc> rohit_: were you looking to add somehting? 15:15:45 <_nadya_> rohit_: why you cannot improve the current kafka publisher? 15:15:51 <fabiog> gordc: there are companies like CloudCruizer and Talligent that can use the API 15:16:22 <fabiog> gordc: they wrote code against it 15:16:29 <ildikov> gordc: 'in maintenance' means no new features but fixing issues if any? 15:16:37 <gordc> fabiog: yep. "people are using it" 15:17:16 <gordc> so the general openstack decision on api is that they will basically never be removed. 15:17:24 <gordc> ^i don't have a link for that 15:17:45 <rohit_> Gordc: yes the api adapter for minasca api 15:17:56 <fabiog> _nadya_: we are publishing to the Monasca API, so sending to Kafka does not address the problem 15:18:04 <gordc> ildikov: that's basically what i imagine. 15:18:24 <ildikov> gordc: ok 15:18:43 <gordc> rohit_: does that require some sort of plugin? 15:18:47 <_nadya_> fabiog: so Monasca cannot read directly from its transporter? 15:18:55 <ildikov> gordc: so the v2 API and the corresponding backends are in maintenance 15:19:37 <rohit_> Gordc: yes, a storage driver plugin 15:19:41 <gordc> ildikov: yeah. people are welcome to fix them and improve them. but clearly the community is spending less and less time on it. 15:19:54 <ildikov> gordc: ok, cool, tnx 15:20:01 <fabiog> _nadya_: trsnsporter? It can read from Kafka bus, but you will loose the authentication of the source and the extra data the API adds before sending to Kafka. Moreover, is bad to expose the guts of a service like that 15:20:30 <gordc> rohit_: so it could be a plugin similar to sql/mongo 15:20:38 <gordc> rohit_: that's good 15:20:42 <jd__> a dispatcher 15:20:53 <fabiog> _nadya_: going through the API means keeping independence on the internal format and other potential chages 15:21:00 <gordc> rohit_: basically teh comments on the patch are that we should make it pluggable, and it should be maintained in it's own tree 15:21:07 <jd__> a dispatcher means RabbitMQ I don't think they want that 15:21:38 <fabiog> jd__: if you are going to an external source, why should you re-publsih to the rabbit? 15:21:40 <gordc> rohit_: as i mentioned in patch, it's not ideal ot have ceilometer devs (unfamiliar with a backend) to properly review something 15:21:52 <gordc> fabiog: that's the point he was making 15:22:13 <gordc> fabiog: the spec is to avoid the republishing 15:22:41 <gordc> rohit_: were there issues with setting up testing in it's own repo? 15:22:45 <jd__> fabiog: it's not always re-publish (polling does not read from rabbit) and rabbit offloads the work so if your external source is down, you can have a backlog 15:23:22 <jd__> gordc: so that's why it should be a publisher no? 15:23:54 <gordc> jd__: isn't that what the spec says? 15:24:10 <jd__> it is, but you talked about a dispatcher 15:24:17 <jd__> [16:20:30] <gordc> rohit_: so it could be a plugin similar to sql/mongo 15:24:18 <jd__> :p 15:24:42 <gordc> jd__: oh, that was because they have a translator from monasca -> v2api 15:25:01 <jd__> ah ok for reading more then 15:25:06 <gordc> right 15:25:35 <gordc> i actually don't mind either dispatcher or publisher 15:26:00 <gordc> reason we had dispatcher was to avoid having the write load affecting pipeline processing 15:26:39 <_nadya_> I'm ok with Monasca publisher as well 15:26:44 <fabiog> gordc: but Monasca has a Kafka bus that acts like the Ceilo topic, that why it would be redundant to have it as a dispatcher 15:27:00 <rohit_> gordc: we have some basic tests running in monasca-ceilometer now 15:27:05 <fabiog> gordc: and in this way we put less stress on the common rabbitmq 15:27:28 <gordc> fabiog: understood. i just wanted to point out reasoning for dispatcher vs publisher 15:27:59 <gordc> it was mainly for http dispatcher which is synchronous and could slow your processing quite a bit 15:28:20 <gordc> fabiog: i've been told writes to monasca are wicked fast so it's a non-issue 15:28:59 <gordc> rohit_: cool. 15:29:24 <fabiog> gordc: not sure I understand what you suggest. You want use to use the http dispatcher? 15:29:26 <gordc> rohit_: so is the discussion whether we need it in tree still? 15:29:39 <rohit_> gordc: but yes, it needs more funtional type tests 15:29:53 <gordc> fabiog: no. i'm just pointing out why there's dispatcher vs publisher. you can ignore this. 15:30:22 <gordc> fabiog: as i said, i don't really have an opinion on you using publisher or dispatcher 15:30:52 <gordc> rohit_: why does that require it to be in celiometer gate? 15:31:14 <gordc> rohit_: we need to address that gap i think 15:31:15 <rohit_> gordc: i was referring to including the storage driver as well 15:31:46 <gordc> rohit_: the idea is we want drivers to be managed appropriately (by correct people) and be tested properly 15:32:01 <gordc> rohit_: it seems to be doable since powervm is doing it in their own gate 15:32:39 <gordc> you just need to pull in ceilometer repo when testing monasca publisher+storage 15:32:43 <jd__> it definitely is 15:33:24 <gordc> rohit_: it's similar to how neutron tests it drivers. 15:34:02 <rohit_> gordc: sure, we wanted this to be in-tree as an endorsement/acceptance 15:34:02 <gordc> i don't think you want a bunch of people not familiar with a driver to review it. 15:34:11 <gordc> rohit_: ah i see 15:34:28 <gordc> rohit_: so that's what i suggested we do same thing as powervm 15:34:29 <rohit_> i understand your point on the maintenance issue with drivers residing in-tree 15:34:52 <gordc> it'll be officially under 'telemetry' team alongside ceilometer+aodh+gnocchi+powervm 15:35:03 <gordc> it'll just be managed by people that actually understand it 15:35:14 <rohit_> ok 15:35:42 <gordc> rohit_: i think the main reason i'm pushing for this path is right now i'm reviewing a bunch of libvirt code 15:35:57 <gordc> and full disclosure, i don't understand anything about libvirt 15:36:48 <gordc> rohit_: if there are testing issues, definitely raise them, and we can work with infra or change ceilometer code abit to accommodate 15:37:32 <rohit_> gordc: there a process to bring a repo under telemetry 15:37:41 <gordc> this will hopefully be faster for monasca in the long run. we already had a few drivers die in review stage because we didn't have appropriate reviewers 15:38:16 <gordc> rohit_: i think collectively we just need to sign off on it. 15:38:43 <rohit_> gordc: sounds good 15:38:45 <gordc> i'll dig up governance patch for powervm 15:39:12 <gordc> https://review.openstack.org/#/c/245894/ 15:39:19 <gordc> rohit_: ^ 15:39:59 <gordc> you just need ptl, but i've generally just polled everyone beforehand to make sure we have agreement 15:40:08 <rohit_> gordc: cool 15:40:36 <gordc> but it seems like we do? everyone cool with monasca interface project living under telemetry at some point? 15:40:48 <fabiog> rohit_: that should be easy because monasca-ceilometer is already under the big tent, so it is just a change of root project 15:41:12 <ildikov> gordc: +1 15:41:29 <gordc> fabiog: yep. i don't think it's much issue but you can ask throst from powervm team to see 15:42:02 <gordc> they bsaically asked me on a friday and then by the next week had it being reviewed 15:43:15 <gordc> rohit_: i'll work with you guys to make sure we can plug in appropriately 15:43:22 <rohit_> fabiog: +1 15:43:30 <gordc> should start soon i guess so we have time to make changes in ceilometer if needed 15:43:55 <gordc> sound good? 15:43:57 <rohit_> gordc: k, sounds good 15:44:03 <gordc> rohit_: :) 15:45:01 <gordc> #topic gnocchi topics 15:45:30 <gordc> jd__: anything? i assume we are not doing a release 15:45:34 <sileht> aodh and ceilometer patch to use gnocchiclient are ready to review 15:45:48 <gordc> sileht: right. will start looking 15:45:54 <sileht> (and depends on the keystoneauth1 review) 15:45:56 <jd__> not much 15:46:15 <gordc> jd__: cool cool. we'll give time to ildikov 15:46:20 <gordc> let's move on. 15:46:27 <gordc> #topic DocImpact flag updates (ildikov) 15:46:37 <ildikov> gordc: tnx, I don't need much 15:46:53 <ildikov> the docs team updated a bit the concept and mechanism 15:47:17 <ildikov> so when you add the DocImpact tag it has to have an explanation too 15:47:57 <ildikov> and the bug will land at the project as opposed to openstack-manuals 15:48:03 <ildikov> so we can track it better 15:48:27 <gordc> ildikov: i've never understood why it landed at openstack-manuals 15:48:31 <ildikov> I hope later on we can move back documentation parts too and the docs itself can live closer to the code also 15:48:46 <gordc> what's an example of the new tag? 15:48:50 <ildikov> gordc: for Nova and I think 4 more project it still does 15:49:07 <gordc> i guess they have dedicated docs people 15:49:16 <ildikov> yeah, most probably 15:50:05 <jd__> ildikov: cool 15:50:11 <jd__> that's going in the right direction 15:50:17 <ildikov> examples: #link http://specs.openstack.org/openstack/docs-specs/specs/mitaka/review-docimpact.html#examples 15:50:56 <ildikov> jd__: I totally agree 15:51:20 <gordc> ildikov: so docimpact <stuff that should already be in your commit message but people like me (and others) don't write proper commit messages> 15:51:38 <pradk> this takes effect for Mitaka? so if we fix something in stable/liberty this wont apply? 15:51:41 <ildikov> so basicly these are the changes, as much as I saw the patches are already merged 15:52:56 <gordc> ildikov: cool cool. i will continue to rely on you to yell 'docimpact!' :) 15:53:32 <ildikov> gordc: you can add pointers into the description like wiki page or etherpad or explicit notes like Install guide update needed, etc. 15:54:16 <ildikov> gordc: lol :) 15:54:38 <gordc> i see. 15:54:58 <gordc> cool. any other questions/comments relating to docs? 15:55:09 <ildikov> pradk: I think it's regardless of the branches, but I will check 15:56:07 <pradk> ildikov, k thx 15:56:12 <gordc> #topic open discussion 15:56:54 <gordc> i should mention if someone wants to be a cross project spec liaison, let me know 15:57:03 <gordc> you just read openstacks-specs apparently. 15:57:44 <gordc> ^ if that interested you, that may or may not exist soon (see last crossproject meeting log) 15:58:13 <gordc> i'll close this in a few, if no one hsa anything 15:59:02 <gordc> thanks folks. 15:59:05 <gordc> #endmeeting