15:02:55 <peschk_l> #startmeeting cloudkitty
15:02:55 <openstack> Meeting started Fri Sep 13 15:02:55 2019 UTC and is due to finish in 60 minutes.  The chair is peschk_l. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:02:56 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:02:58 <openstack> The meeting name has been set to 'cloudkitty'
15:03:42 <peschk_l> Hi everybody, welcome to the september cloudkitty meeting
15:03:52 <peschk_l> Today's agenda can be found here: https://etherpad.openstack.org/p/cloudkitty-meeting-topics
15:04:14 <peschk_l> as usual, feel free to add any topic you'd like to discuss
15:04:29 <peschk_l> and there will be a Q&A session at the end
15:04:40 <peschk_l> #topic feature freeze
15:05:08 <peschk_l> as you know, the feature freeze for the Trains cycle happened yesterday
15:05:49 <peschk_l> I'm quite happy with what we achieved this release, as we have  achieved more than what was planned
15:06:32 <peschk_l> and the last feature patch will be merged today, meaning we're only one day late, which is a pretty good score for openstack projects :)
15:06:52 <peschk_l> anyway the complete list of release notes can be found here: https://docs.openstack.org/releasenotes/cloudkitty/unreleased.html
15:07:26 <peschk_l> and you'll have to add an experimental v2 storage driver for elasticsearch to that list
15:07:39 <peschk_l> (the patch is currently being gate tested/ merged by zuul)
15:08:50 <peschk_l> A lot of new features and hardening happened during this cycle, and jferrieu and I will spend the feature freeze expanding the tempest plugin
15:09:02 <peschk_l> (If we find no bug, of course :) )
15:10:11 <peschk_l> concerning the community goals: we've been one of the first projects to implement full python3 support, and I'd like to drop python2 support ASAP
15:10:30 <peschk_l> it'll probably be one of the first things in the U cycle
15:10:46 <jferrieu> that would be a good thing indeed
15:11:19 <peschk_l> well as soon as the TC gives a green light of course
15:11:48 <peschk_l> Ionce support for py2 has been dropped, I'd like to progressively introduce typ[e annotations
15:11:54 <peschk_l> with eventual mypy checking
15:12:36 <peschk_l> they've already been some proposal for that in openstack, but the answer was "meh, let's wait for py3, we don't want two types of annotations"
15:13:05 <peschk_l> anyway, adding some typing would really help with hardening IMO
15:13:19 <jferrieu> I agree
15:13:31 <peschk_l> concerning the second community goal, which is PDF generation, we haven't been able to implement it yet
15:14:06 <peschk_l> the biggest problem I encoutered being that tables and API reference generation is not supported by latex
15:14:19 <peschk_l> so only a small part of the doc can be generated
15:14:38 <peschk_l> I'm waiting for help of the docs team on this
15:15:38 <peschk_l> somebody proposed a patch on the dashboard. We'll have to review it, but frome what I've seen, only a "project"official" badge is visible in the final pdf
15:15:40 <peschk_l> #link https://review.opendev.org/#/c/681703/
15:16:09 <peschk_l> any questions about the feature freeze / community goals ?
15:16:32 <jferrieu> I'm good
15:16:54 <peschk_l> all right, next topic then
15:17:03 <peschk_l> #topic meeting schedule
15:17:58 <peschk_l> we've gained some part-time contributors on the project. They'll be working from monday to wednesday
15:18:52 <peschk_l> Also, huats lives in montreal now, so I'd like to propose a new meeting time that's at the beginning of the week, and that is compatible with the montreal and Paris timezones
15:19:27 <peschk_l> I'd also like to make the meeting happen twice a month, and publish a recap on the openstack-discuss mailing list
15:20:03 <peschk_l> would the 1st and 3rd monday of each month be okay for you ?
15:20:18 <peschk_l> at 14h UTC
15:20:26 <jferrieu> I am ok with this, the recap will be a nice plus also imo
15:20:34 <peschk_l> great
15:21:22 <peschk_l> Of course, if we gain contributors from other timezones, we'd be pleased to have at least one of both meetings on a timeslots that suits them
15:22:11 <peschk_l> okay then, I'll propose a patch in order to change the meeting schedule
15:22:57 <peschk_l> and I guess that was it concerning the meeting schedule
15:23:21 <peschk_l> #topic storyboard
15:23:38 <peschk_l> new release cycle, new resolutions
15:24:36 <peschk_l> during the last release cycle, we've been writing specs for each new feature, and we keep doing it, which is a nice thing. I'd like to push storyboard usage now
15:25:08 <peschk_l> It will need some celaning (Ill take care of it), but once it is done, I'd like to have every bug to be registered there
15:25:26 <peschk_l> and not just the major, like we did until now
15:26:31 <peschk_l> one of the first we'll need to adress is that hashmap rules are reloaded way too often, which can lead to a real SQL database hammering when a short collect period is used
15:27:11 <peschk_l> we'll probably introduce some kind of TTL cache in the Hashmap module for this
15:28:09 <peschk_l> We'll also use storyboard to follow the v1 API porting status
15:28:33 <openstackgerrit> Merged openstack/cloudkitty master: Add an ElasticSearch v2 storage driver  https://review.opendev.org/679072
15:28:54 <peschk_l> (ah, this was our last feature patch :) )
15:29:07 <jferrieu> congrats
15:29:32 <priteau> Ideally the API would notify the processors to force refresh of their hashmap cache when rules are updated, otherwise this could lead to puzzling behaviour
15:29:51 <priteau> Oh and hello :)
15:30:31 <peschk_l> priteau: hi :) agreed
15:30:57 <peschk_l> The TTL cache would rather be a quick fix
15:31:58 <peschk_l> but we'll eventually have the same problem with the new rating module so yeah, the long-term fix would be a notification sent out to the processors
15:33:44 <peschk_l> I think that's it concerning storyboard
15:33:49 <peschk_l> #topic open discussion
15:34:13 <peschk_l> priteau: you were intersted in the elasticsearch storage driver, right ?
15:35:09 <priteau> Yes, particularly if it could help to charge based on events rather than metrics.
15:36:35 <peschk_l> priteau: well it doesn't change the way we charge, but It provides some features wich INfluxDB does not
15:36:47 <peschk_l> HA, etc
15:37:45 <peschk_l> priteau: But you're using the monasca collector right ? If monasca provides proper support for events, it would be a nice addition to the monasca collector
15:38:50 <priteau> Monasca event support is just a spec for now. Once there is code it might be a solution.
15:39:17 <peschk_l> I see
15:40:09 <peschk_l> what kind of events do you have in mind ? recource creation / deletion ?
15:40:13 <peschk_l> *resource
15:41:49 <priteau> Yes, there are notifications for creation/update/deletions of many types of resources in OpenStack already which could be used for rating.
15:42:50 <priteau> I suppose the difficulty is keeping track of what resources are allocated if there is no periodic notification.
15:44:44 <peschk_l> Yup. Given ck's design it would require to "make them look like a metric"
15:45:32 <peschk_l> Meaning having some kind of pollster somewhere checking if the resources still exist
15:45:56 <peschk_l> Which is more or less what ceilometer does
15:46:44 <peschk_l> priteau: i don't remember, did you use monasca alone for rating or did you also use ceilosca?
15:48:07 <priteau> Both actually. In the blog post already published I used Ceilosca, but I also configured rating using an openstack prometheus exporter sending data to Monasca (still need to write the blog post for this one)
15:49:53 <peschk_l> Okay
15:51:58 <priteau> I may be able to work on this again soon, I will be able to find out what works bes.
15:52:00 <priteau> best
15:53:35 <peschk_l> priteau: We'de definitely be interested in feedback
15:54:23 <peschk_l> a temporary solution for event-based rating might also be to push already rated dataframes into
15:54:30 <peschk_l> into cloudkitty
15:54:56 <peschk_l> jferrieu recently added a v2 API endpoint for that
15:55:37 <priteau> I'll look into it, thanks
15:58:22 <peschk_l> Are there other topics to be discussed ?
15:59:16 <jferrieu> I'm ok :)
16:01:13 <peschk_l> priteau, do you have more questions ?
16:01:31 <priteau> Not for now, probably in a few weeks when I have the chance to work with CloudKitty again ;-)
16:02:41 <peschk_l> priteau: all right, we're available on IRC, and not just on meeting times :)
16:02:49 <priteau> Thank you
16:02:54 <peschk_l> thanks guys!
16:03:13 <peschk_l> #endmeeting