02:00:19 <dangtrinhnt> #startmeeting openstack telemetry
02:00:19 <rafaelweingartne> Hello guys
02:00:20 <openstack> Meeting started Thu May  9 02:00:19 2019 UTC and is due to finish in 60 minutes.  The chair is dangtrinhnt. Information about MeetBot at http://wiki.debian.org/MeetBot.
02:00:21 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
02:00:23 <openstack> The meeting name has been set to 'openstack_telemetry'
02:00:23 <engel75> hi all
02:00:40 <dangtrinhnt> okie, we have 5 people, it's good :)
02:00:51 <lxkong> such a big team
02:01:09 <dangtrinhnt> So please follow the meeting agenda using this etherpad #link https://etherpad.openstack.org/p/telemetry-meeting-agenda
02:01:54 <dangtrinhnt> #topic Finalize the core team & active contributors
02:03:00 <zhurong_phone> shaoman is on our internel meeting, so he can't attend
02:03:29 <dangtrinhnt> So now besides lxkong, zhurong_phone, we have Adrian Turjak is interested in being the core and contribute
02:03:45 <dangtrinhnt> I may send an nomination email after this meeting.
02:04:32 <dangtrinhnt> So rafaelweingartne, engel75, I know you guys just involve in OpenStack recently (?), I may ask would you like to contribute to Telemetry?
02:04:34 <lxkong> adrian will join shortly
02:04:35 <adriant> o/
02:04:38 <dangtrinhnt> adriant hi
02:04:44 <lxkong> ah, so quick, adriant
02:05:00 <rafaelweingartne> dangtrinhnt: yes, we would love to
02:05:28 <dangtrinhnt> rafaelweingartne, engel75 you don't have to take the core team but your input and code contribution would be precious for us.
02:05:28 <rafaelweingartne> I mean, that is what we intent to, and that is why we asked about the event system deprecation
02:05:38 <engel75> I am not a developer at all but rafael will contribute in the near future
02:05:51 <dangtrinhnt> That's awesome. Thanks rafaelweingartne
02:05:55 <dangtrinhnt> engel75, np
02:06:23 <rafaelweingartne> engel75 is an operator, so I see his inputs as very valuable providing the use cases that we design and implement
02:06:50 <dangtrinhnt> engel75, rafaelweingartne, awesome, we absolutely need use cases!!! :)
02:07:08 <engel75> yes I will try to provide input from our priduction experiences
02:07:17 <engel75> yes I will try to provide input from our production experiences
02:07:19 <dangtrinhnt> Thanks!
02:08:02 <dangtrinhnt> ok, I guess we can move to the next topic, any questions/idea?
02:08:08 <dangtrinhnt> lxkong, zhurong_phone
02:08:09 <dangtrinhnt> ?
02:08:22 <lxkong> go ahead
02:08:26 <dangtrinhnt> cool
02:08:43 <dangtrinhnt> #topic Meeting time for non-APAC contributors
02:10:00 <dangtrinhnt> In our next poll about the time for team meeting, we missed out the non-APAC contributors. Part of it was the polling time is during the holiday and summit so those guys didnot have a chance to vote.
02:10:45 <dangtrinhnt> So, I suggest we made a time change so the developers/operators in the US/EU region could join.
02:10:48 <dangtrinhnt> Any suggestion?
02:11:32 <dangtrinhnt> rafaelweingartne and engel75 are also suffered. It's 4AM their time.
02:11:45 <zhurong_phone> ohh...
02:11:53 <engel75> no rafael is from Brazil
02:11:53 <dangtrinhnt> Right, rafaelweingartne, engel75?
02:11:55 <rafaelweingartne> actually, 23:00 PM on my time
02:11:59 <lxkong> i am going to ask do we already know if there are people from non-APAC intereated in contributing ceilometer...
02:12:07 <engel75> My time zone is GMT+2
02:12:07 <dangtrinhnt> yes, like mrunge
02:12:17 <engel75> so 4:12 now :)
02:12:20 <dangtrinhnt> a couple developers emailed me
02:12:22 <dangtrinhnt> lxkong
02:12:34 <lxkong> ok, that's great
02:12:47 <dangtrinhnt> engel75, :)
02:13:20 <dangtrinhnt> so, ok, I will send out an email asking for time change after this meeting.
02:13:27 <engel75> but I will not be able to attend that often anyway
02:13:28 <zhurong_phone> can we set another poll for this
02:13:37 <dangtrinhnt> zhurong_phone, sure
02:14:00 <dangtrinhnt> engel75, no worries, there are still a couple US-based guys
02:14:23 <dangtrinhnt> ok, should we move on?
02:14:40 <zhurong_phone> I think so
02:14:45 <dangtrinhnt> #topic Moving telemetry bug tracker to Storyboard: https://review.opendev.org/#/c/651963/
02:15:20 <dangtrinhnt> It supposed to finish the last couple weeks but due to the summit, it's delayed.
02:15:27 <dangtrinhnt> So I'm trying to push it.
02:15:38 <lxkong> dangtrinhnt: thanks for that
02:15:47 <engel75> yes thank you
02:16:07 <dangtrinhnt> After the migration is done, we could go through the stories and maybe set aside sometime for bug triage, I guess.
02:16:22 <dangtrinhnt> I think there's not much to talk about this. so let's move to the next topic
02:16:34 <engel75> ok
02:16:52 <dangtrinhnt> #topic Train-1 milestone targets (Jun 03 - Jun 07)
02:17:07 <dangtrinhnt> #link https://releases.openstack.org/train/schedule.html
02:17:16 <dangtrinhnt> FYI, we will reach Train-1 soon :)
02:17:39 <dangtrinhnt> I would love to set some goals even small ones
02:18:00 <dangtrinhnt> So what do you think you would want to accomplish at that milestone?
02:18:35 <dangtrinhnt> lxkong, zhurong_phone, rafaelweingartne, engel75
02:18:45 <lxkong> dangtrinhnt: on aodh side, i will try to finish the auto-healing before T-1
02:18:45 <engel75> would a statement about the event_subsystem be a valid candidate?
02:18:57 <dangtrinhnt> lxkong, awesome.
02:19:06 <rafaelweingartne> What about a community decision regarding the event subsystem deprecation? And maybe some new poller or something (if we decide to keep the system)?
02:19:09 <lxkong> the last part is the notifier
02:19:31 <dangtrinhnt> engel75, given the amount of time left, I would put that into Train-2
02:19:42 <lxkong> do we have other options for now to replace the event_subsystem?
02:20:12 <engel75> may I ask why it has to be replaced?
02:20:12 <zhurong_phone> container metrics and switch oslo.privileps
02:20:14 <dangtrinhnt> rafaelweingartne, but if you think you can make it on time, we could consider that
02:20:38 <lxkong> i don't know the history of it's deprecation, if there is not good reason in the current situation, we could get it back
02:20:48 <dangtrinhnt> engel75, I think what lxkong means is an alternative
02:21:04 <lxkong> we are also using the notification for billing
02:21:07 <lxkong> adriant ^^
02:21:23 <engel75> as we do
02:21:40 <engel75> together with the pollers
02:21:46 <dangtrinhnt> engel75, part of it was the feature is never finished
02:21:49 <adriant> the notification samples are the only way to get accurate samples
02:22:08 <adriant> for when a resource is created or deleted
02:22:14 <lxkong> agree
02:22:27 <adriant> polling is useful, but it will always be behind
02:22:35 <engel75> it will
02:22:45 <adriant> and even means if a resource is only active within the polling period, we can't bill it
02:22:46 <engel75> a combination is the way to go imho
02:22:52 <adriant> since the pollster will miss it
02:22:59 <lxkong> so if there is no good candidate option, we could considering bringing that back and remove the deprecation msg.
02:23:08 <rafaelweingartne> lxkong: the reason we got was that people wanted to shrink the code. However, there was not a clear picture of the benefits and good alternatives in place.
02:23:21 <dangtrinhnt> engel75, rafaelweingartne, adriant, lxkong, https://github.com/openstack/ceilometer/commit/8a0245a5b3e1357d35ad6653be37ca01176577e4
02:23:35 <dangtrinhnt> so I guess the next topic will be discussed now
02:23:50 <rafaelweingartne> yes
02:24:16 <dangtrinhnt> ok, so again, I would like to set the targets for Train-1 milestone
02:24:33 <lxkong> rafaelweingartne: does the current behavior satisfy your requirement?
02:25:13 <lxkong> if not, could you please create a story after we have storyboard, and we could discuss how to improve there
02:25:24 <engel75> in our environment some nofitications are "missed" that's why we use the pollsters to make sure to get all resources
02:25:29 <rafaelweingartne> yes it does, but we will need to add/extend a few things to fully deliver what we need. engel75 is more suitable to present the use cases.
02:25:55 <engel75> by "our" I mean rafaelweingartne and mine
02:25:58 <rafaelweingartne> There are these issues for instance, but the idea is to work to fix them
02:26:02 <dangtrinhnt> lxkong, rafaelweingartne, engel75 so you want us to have it in Train-1?
02:26:30 <engel75> would be great - at least a statement?
02:26:34 <rafaelweingartne> Do you mean a statement regarding the deprecation?
02:26:53 <lxkong> i'm not sure things will be clear before T-1, but let's be pessimistic and make it for T-2
02:26:55 <engel75> to show it is maintained again
02:27:02 <dangtrinhnt> lxkong, +1
02:27:12 <rafaelweingartne> statement, or a decision regarding the deprecation
02:27:13 <engel75> ok
02:27:14 <dangtrinhnt> engel75, it could be right after this meeting
02:27:27 <engel75> ok great
02:27:53 <dangtrinhnt> engel75, rafaelweingartne, like lxkong suggested, finish the feature in T-2, but announce it at T-1, ok?
02:28:01 <engel75> perfect
02:28:14 <engel75> +1
02:28:22 <dangtrinhnt> awesome, anything else?\
02:28:29 <dangtrinhnt> adriant, zhurong_phone, lxkong?
02:28:40 <lxkong> dangtrinhnt: not sure if you missed the task zhurong_phone mentioned just now
02:28:57 <dangtrinhnt> put it on the etherpad already
02:29:00 <dangtrinhnt> https://etherpad.openstack.org/p/telemetry-meeting-agenda
02:29:02 <lxkong> good :-)
02:29:04 <dangtrinhnt> :)
02:29:20 <adriant> dangtrinhnt: there was some issues around a cumulative cpu usage feature lxkong was using that is now gone
02:29:20 <rafaelweingartne> I would just like to clarify what you mean by feature, dangtrinhnt. Do you mean a statement regarding the event system? Or, a proposal of an actual new feature in Ceilometer regarding the event system?
02:29:26 <zhurong_phone> thanks dangtrinhnt
02:29:54 <dangtrinhnt> adriant, cpu_util?
02:30:09 <adriant> yeah, based on the transformer
02:30:11 <adriant> which was removed
02:30:13 <dangtrinhnt> rafaelweingartne, I mean making it work
02:30:14 <lxkong> yeah, cpu_util, there are some people asking for that
02:30:16 <adriant> it is needed for autoscaling
02:30:22 <lxkong> or for an alternative
02:30:24 <adriant> ^
02:30:39 <rafaelweingartne> dangtrinhnt, ah. Sure, got it.
02:30:53 <dangtrinhnt> rafaelweingartne, :)
02:30:59 <zhurong_phone> now in function test, using cpu :()
02:31:28 <adriant> the feature was removed because it was meant to be able to be queried from gnocchi but lxkong has said he can't figure out how to get the same type of numbers/data
02:32:11 <lxkong> and i failed to manage to get any useful suggestion from Gnocchi team.
02:32:38 <dangtrinhnt> adriant, lxkong, zhurong_phone, gnocchi is out of our hands so I guess we can either bring it back to OpenStack and maintain it ourselves
02:32:43 <lxkong> i have a feeling we may bring that back if there are a lot more people asking for that
02:32:53 <lxkong> as a short term plan
02:32:59 <engel75> yes we should try that!
02:33:01 <dangtrinhnt> lxkong, let's create a discussion on the ML
02:33:22 <zhurong_phone> we also want to bring ceilometer api back
02:33:31 <lxkong> i hope there could be some people jump out and say: it's very easy to achieve that in Gnocchi :-)
02:33:46 <rafaelweingartne> but, this process(Gnocchi) would take some bureaucratic steps to re-introduce the project to OpenStack foundation, right?
02:33:55 <adriant> zhurong_phone: potentially, maybe even as just a wrapper around gnocchi
02:34:04 <dangtrinhnt> okie guys, way back into love :)
02:34:11 <rafaelweingartne> which means that it is something for a medium/long term plan
02:34:17 <dangtrinhnt> rafaelweingartne, sure
02:34:25 <dangtrinhnt> that could be possible
02:34:43 <dangtrinhnt> if we can find some champion taking care of it
02:35:07 <dangtrinhnt> otherwise it could properly be a burden in the future
02:35:53 <lxkong> dangtrinhnt: let's discuss in mailing list first, see if there is an easy answer
02:35:59 <engel75> I don't see any alternative to gnocchi right now
02:36:05 <dangtrinhnt> lxkong, +1
02:36:12 <lxkong> engel75: are you using Gnocchi now?
02:36:18 <engel75> lxkong, +1
02:36:21 <engel75> yes
02:36:22 <lxkong> or plan to use that in the near future
02:36:24 <lxkong> ok
02:36:34 <engel75> in production
02:36:44 <dangtrinhnt> engel75, awesome
02:36:47 <engel75> it scales very well
02:36:51 <engel75> on SSDs...
02:36:59 <dangtrinhnt> engel75, you can help lxkong on the cpu_util
02:37:04 <dangtrinhnt> :)
02:37:20 <engel75> I need to get it explained but yes I can try to
02:37:40 <engel75> please send me an email after the meeting
02:37:42 <lxkong> engel75: thanks, please reply the email after you see that
02:37:53 <engel75> sure
02:38:27 <dangtrinhnt> engel75, if you can, please cc the openstack-discuss ML as well so others can follow
02:38:38 <engel75> ok will do so
02:38:41 <dangtrinhnt> with [telemetry] prefix :)
02:38:46 <dangtrinhnt> engel75, great thanks
02:38:56 <dangtrinhnt> ok, anything else?
02:39:06 <dangtrinhnt> zhurong_phone, adriant?
02:39:15 <zhurong_phone> nup
02:39:38 <adriant> all good :)
02:39:43 <dangtrinhnt> ok
02:40:07 <dangtrinhnt> #topic Telemetry roadmap cont. https://etherpad.openstack.org/p/telemetry-train-roadmap
02:40:15 <dangtrinhnt> ok, I guess we have Train-1 targets
02:40:39 <dangtrinhnt> but I also would love to discuss our big plan for Train, AKA, the Roadmap
02:41:01 <adriant> dangtrinhnt: sorry just remembered,  there may be some missing pollsters we can add, such as those for Magnum and I think there are some missing ones for Octavia
02:41:15 <dangtrinhnt> ah, okie
02:41:25 <rafaelweingartne> adriant: we have been discussing that
02:41:33 <rafaelweingartne> and will probably implement it
02:41:40 <adriant> would be good to do a general overview of what services are missing pollsters/event watching
02:41:42 <dangtrinhnt> adriant, so when do you think we can have those features?
02:41:42 <engel75> yes we are happy to help with those pollsters
02:41:59 <dangtrinhnt> rafaelweingartne, enge75, great
02:42:50 <lxkong> i suggest we begin with the ones we care about, e.g. for us the Magnum and Octavia
02:43:12 <dangtrinhnt> lxkong, +1
02:43:20 <rafaelweingartne> lxkong, +1
02:43:40 <dangtrinhnt> adriant, lxkong, what timeline do you think is doable for those?
02:43:40 <zhurong_phone> +1
02:43:57 <lxkong> T-2 or T-3, adriant?
02:44:07 <adriant> probably yeah
02:44:07 <lxkong> should be an easy one because we already have that in our production
02:44:33 <adriant> I don't know how much I can dedicate to telemetry for code work, but would gladly code review
02:45:20 <dangtrinhnt> adriant, lxkong, how were those pieces of code licenced? I would avoid any commercial licenced code
02:45:36 <dangtrinhnt> adriant, it's great as well.
02:45:43 <rafaelweingartne> lxkong: also, do you have it in Stein or Rocky release?
02:46:04 <lxkong> dangtrinhnt: don't worry, all apache 2.0 :-)
02:46:14 <lxkong> rafaelweingartne: no, we are using a very old version ceilometer
02:46:14 <dangtrinhnt> lxkong, awesome!
02:46:14 <engel75> lxkong, would you mind sharing your extensions regarding those pollsters?
02:46:56 * adriant is looking at our patch for magnum pollster now
02:46:58 <dangtrinhnt> rafaelweingartne, engel75, we have 14m left, I guess lxkong and adriant can propose some patches for review
02:46:59 <engel75> we have to introduce a Octavia pollsters the next couple of weeks
02:47:01 <rafaelweingartne> lxkong: so there would be an effort to push forward the code (if the structure/design of the code has changed over time)
02:47:04 <lxkong> i will see if i could submit those code in T for review
02:47:10 <adriant> not very much code, but yes, easy to share and get working for newer versions of ceilometer
02:47:18 <rafaelweingartne> cool
02:47:25 <dangtrinhnt> guys, awesome!
02:47:55 <lxkong> dangtrinhnt: please make it for T-3 and allocate magnum pollster task to me
02:48:05 <engel75> so regarding the roadmap in the long term - we should align with cloudkitty?
02:48:10 <dangtrinhnt> lxkong, adriant, engel75, rafaelweingartne, \m/ rock!
02:48:46 <engel75> minute based billing :)
02:49:06 <dangtrinhnt> engel75, could be possible
02:49:30 <dangtrinhnt> for that, I would send an discussion email to the ML and let figure that out, I will put that as an option
02:49:48 <engel75> we just have to make sure not event is missed or if one is missed there is some sort of pollster to get it after a couple of seconds
02:49:57 <engel75> ok
02:49:57 <dangtrinhnt> +1
02:50:04 <dangtrinhnt> ok, 10m left
02:50:43 <dangtrinhnt> so for the roadmap, do you think puting back some features that have use cases a possible option?
02:51:02 <rafaelweingartne> I would say so
02:51:07 <dangtrinhnt> lxkong, zhurong_phone, adriant, rafaelweingartne, engel75
02:51:40 <lxkong> yep, if that's the best option we have at the moment
02:51:42 <rafaelweingartne> I would also say that we need a clear perspective on how we handle things. For instance, what can be deprecated and removed? broken features, features with a clear and better alternatives, and so on
02:52:02 <adriant> dangtrinhnt: yes, for cpu_util maybe, but the issue why it was removed: "Transformers cannot work correctly on multiple nodes without workload partitioning, which has been removed."
02:52:09 <adriant> https://review.opendev.org/#/c/580709/
02:52:21 <adriant> so that may end up a rabbit hole
02:52:30 <dangtrinhnt> rafaelweingartne, +1
02:52:38 <adriant> we need something 'like' it, but maybe bringing it back isn't the answer
02:52:51 <dangtrinhnt> adriant, makes sense
02:53:39 <rafaelweingartne> adriant: maybe, we could bring back the feature, but with a new design/implementation
02:53:48 <dangtrinhnt> or alternatives. We will have to put more effort investigating that.
02:53:49 <adriant> rafaelweingartne: that's pretty much what I mean
02:53:49 <lxkong> i prefer we deprecate something unless we have a feature parity alternative, something like 'not work', 'covered by other projects' are not good reasons
02:54:09 <dangtrinhnt> rafaelweingartne, agree
02:54:11 <dangtrinhnt> lxkong, +1
02:54:17 <rafaelweingartne> lxkong, +1
02:54:32 <dangtrinhnt> ok good
02:54:34 <dangtrinhnt> anything else?
02:54:43 <dangtrinhnt> 5m left
02:54:58 <lxkong> not from me
02:55:18 <engel75> what a bout the documentation
02:55:27 <dangtrinhnt> engel75, what do you mean?
02:55:30 <engel75> it is not the most "complete" one
02:55:38 <dangtrinhnt> missing, not correct?
02:55:45 <dangtrinhnt> ah ok
02:56:12 <dangtrinhnt> we absolutely need to correct and improve the documentations
02:56:20 <dangtrinhnt> I will put it on the todos
02:56:25 <engel75> I will try to find missing parts
02:56:29 <dangtrinhnt> Stories I meant :)
02:56:34 <dangtrinhnt> thanks
02:56:52 <dangtrinhnt> I guess we can complete it in T-1
02:57:19 <dangtrinhnt> 3m left
02:57:23 <dangtrinhnt> anything else?
02:57:26 <engel75> I never contributed to some Openstack documentation but I could try to improve it from a operator perspective
02:57:44 <adriant> engel75: from an operators perspective is sorely needed in the docs
02:57:45 <engel75> I mean the ceilometer part
02:58:05 <dangtrinhnt> engel75, that is awesome.
02:58:08 <lxkong> engel75: that will be very helpful
02:58:19 <rafaelweingartne> adriant, +1
02:58:20 <dangtrinhnt> if you need any guideline for that, let me know
02:58:24 <zhurong_phone> engel75: thanks
02:58:39 <dangtrinhnt> oke, 2m left
02:58:57 <engel75> dangtrinhnt, ok I will send you an email to get help with it
02:59:05 <dangtrinhnt> engel75 +1
02:59:38 <dangtrinhnt> Ok, I think we're good for now. Anything else you would like to discuss, please send an email to the ML
02:59:52 <engel75> ok
03:00:13 <dangtrinhnt> Thanks everyone for participating in this great meeting.
03:00:24 <dangtrinhnt> #endmeeting