15:00:18 <srwilkers> #startmeeting openstack-helm
15:00:19 <openstack> Meeting started Tue Dec 11 15:00:18 2018 UTC and is due to finish in 60 minutes.  The chair is srwilkers. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:20 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:22 <openstack> The meeting name has been set to 'openstack_helm'
15:00:36 <srwilkers> heya -- portdirect is running a bit late, so i'm going to kick things off for him
15:00:39 <srwilkers> #topic rollcall
15:00:41 <srwilkers> o/
15:00:56 <evrardjp> o/ (because I can't really \o/)
15:01:07 <mattmceuen> o/
15:01:20 <srwilkers> if you turn in a circle really quick evrardjp, a o/ can become a \o/
15:01:26 <srwilkers> but you have to be really quick
15:01:47 <srwilkers> lets give it a few minutes to let people trickle in -- also, here's the etherpad for today: https://etherpad.openstack.org/p/openstack-helm-meeting-2018-12-11
15:01:52 <srwilkers> seeing some good stuff on here
15:02:29 <evrardjp> srwilkers: :)
15:03:17 <howell> o/
15:04:03 <jayahn> o/
15:05:02 <srwilkers> alright, let's get started
15:05:10 <srwilkers> #topic gate refactor
15:05:53 <srwilkers> first thing -- we've talked about getting periodic gates going in the past.  we're at the point we should get that started
15:06:15 <srwilkers> i've admittedly not set one up before, and i think evrardjp had some great ideas about what these would look like back in Denver
15:06:20 <srwilkers> have any thoughts here evrardjp?
15:06:54 <evrardjp> well it's quite simple, dailies are a simple job that runs on a specific pipeline
15:07:00 <evrardjp> named "periodic"
15:07:24 <evrardjp> those jobs can be extended in terms of length.
15:07:37 <evrardjp> (I mean runtime)
15:07:53 <evrardjp> There two things to figure out there:
15:08:16 <evrardjp> 1) What will be the content of those (extending the current job to increase coverage is good!)
15:08:25 <evrardjp> 2) Who will be monitoring those
15:08:39 <evrardjp> The second point is by far more important:
15:08:52 <evrardjp> If we have jobs that nobody monitors, there is almost 0 point of doing so.
15:09:14 <evrardjp> it's easy for people to forget those periodics.
15:09:44 <srwilkers> yeah, that's what i was thinking as well.  ive typically got zuul.openstack.org up on my screen at all times (at home and at work), but that's just me.  we'd certainly need to treat these as first class and not ignore them
15:10:18 <evrardjp> There is no pipeline yet for weeklies. We could have a job that skips when not on the week-end for example. But I will work with infra to see if it's possible to create a week periodic pipeline.
15:10:30 <srwilkers> that'd be awesome
15:10:37 <evrardjp> However, I don't think we should make use of such pipeline, as daily seems enough for OSH cases
15:11:09 <evrardjp> srwilkers: do you have a specific idea of things that should run every week but not every day?
15:11:17 <srwilkers> yeah -- i was thinking weekly would probably be good for something like the old armada job we ran
15:11:22 <srwilkers> or some other sort of big bang integration run
15:12:17 <evrardjp> k
15:12:32 <srwilkers> but daily should really be enough for what we need
15:12:48 <srwilkers> and i think pete's comment about the "big and fat" old jobs could possibly be a good start there
15:12:52 <srwilkers> or some subset of those
15:13:07 <jayahn> +1 on armada job - probably good candidate for it
15:14:26 <evrardjp> srwilkers: could you clarify what those are?
15:14:52 <evrardjp> Is that the regular folder's jobs? (running everything in multinode for example)
15:14:59 <srwilkers> the old multinode jobs we ran previously (the five node jobs)
15:15:00 <srwilkers> yeah
15:15:45 <evrardjp> I see.
15:16:22 <srwilkers> that could at least get us started, and we could modify those as needed/required moving forward
15:17:06 <evrardjp> I have to check what was done/merged recently though, to ensure I am not re-inventing the wheel.
15:17:32 <srwilkers> evrardjp: of course :)
15:17:34 <evrardjp> (when implementing those new jobs)
15:17:48 <evrardjp> I will probably dropped my old efforts then :'(
15:18:14 <evrardjp> I don't have anything to add
15:19:06 <evrardjp> srwilkers: do you plan to talk about the refactor and its potential issues, if I am not mistaken?
15:19:16 <srwilkers> another item possibly related to the refactor is a failure in part of the script we used to exercise the stack
15:19:21 <srwilkers> evrardjp: took the words out of my mouth
15:19:27 <srwilkers> ie: https://review.openstack.org/#/c/623218/
15:20:02 <srwilkers> the latest failure for the compute-kit gate ran fine, up until we attempted to ssh into the vm we launch and curl the keystone endpoint
15:20:09 <srwilkers> http://logs.openstack.org/18/623218/6/check/openstack-helm-compute-kit/f1143fa/job-output.txt.gz#_2018-12-07_17_42_36_020311
15:20:25 <srwilkers> i had planned to do some digging here today, but that's the only issue i've noticed bubble up so far
15:22:05 <evrardjp> thanks srwilkers for the work, the debugging, and the fix ! :)
15:23:08 <srwilkers> evrardjp: don't thank me.  this was just a by-product of the conversations you helped drive in Denver :)
15:23:18 <srwilkers> that's all i had to talk about wrt the gates.  anybody else have anything?
15:25:22 <mattmceuen> Not I!
15:25:40 <evrardjp> I have nothing
15:26:01 <srwilkers> #topic image builds
15:26:48 <srwilkers> i'll just quickly touch on the point portdirect left here, then let jayahn and evrardjp chime in
15:27:02 <evrardjp> I guess two big things need to happen there
15:27:34 <srwilkers> there was some chat in the openstack-helm channel today about portdirect getting things squared away for us to be able to push images, so there's a path forward there
15:27:41 <srwilkers> evrardjp: yep :) take it away
15:28:11 <evrardjp> 1) Work on the new repos -- 2) Clean the old ones
15:28:34 <evrardjp> I put numbers there, because I'd hope we start 2) only after 1) is fully done
15:28:54 <evrardjp> To talk about 1) -- A few elements need happening
15:29:24 <evrardjp> Include LOCI's image building into this new repo (WIP, only a few minor details are remaining)
15:29:45 <evrardjp> Include proper logging of the building
15:29:58 <evrardjp> Publish images
15:30:37 <evrardjp> Be able to release images (so in other words... be able to "version" images in a nicely fashion, for 1.0 for example)
15:31:10 <evrardjp> Then last step would be be able to consume said images into the other repos' charts.
15:31:16 <evrardjp> Does that sound correct?
15:32:00 <srwilkers> that seems sane to me
15:32:10 <evrardjp> A stretch goal would be to be able to test images separately.
15:32:32 <evrardjp> For all those above I am looking for help, and I hope a few SUSE people will step in the process :)
15:33:02 <evrardjp> The last one (test images separately) is the trickiest for me, as I have no idea on where to start.
15:33:13 <srwilkers> evrardjp: i'll be happy to jump in and help there.  i'll take a look to see where i can start getting my hands dirty
15:33:23 <evrardjp> Call for help is made, I leave the mic to other people now :)
15:33:37 <evrardjp> srwilkers: we have docs :)
15:33:56 <srwilkers> jayahn: i see you've got quite a few items here.  floors yours
15:34:14 <evrardjp> I think all of those patches should be into the same topic (I don't have the topic name yet, but will think about it)
15:35:23 <jayahn> situation #1. skt has been deveoping charts to enable lbaas and fwaas, with kolla image. To upstream these charts, skt knows charts should pass gates using loci images. That loci part has been a big blocker for us.
15:35:23 <jayahn> situation #2. skt ci pipeline currently detects any changes in values (regarding images, any tag changes or new image). skt has been replacing loci-based image values to kolla-based image values.
15:35:23 <jayahn> Additionally, skt knows osh proejct will have loci-based gates only. that is totally understandable.
15:35:23 <jayahn> based on situation #1, situation #2, skt need osh cores guide, opinion, or discussion,.
15:35:23 <jayahn> skt can provide information on how to make osh charts work with kolla-image for anyone who wants to use kolla-images. it won't have gates, but we can always provides "any type of documents" for reference information. >> will it be valuable for osh project?
15:36:21 <jayahn> in contrary, skt really needs help on loci image. how we can properly set / build loci image so that we can uptream our charts. can anyone help us to set up loci images?
15:37:51 <srwilkers> i'll admit, i'm by no means an expert with the LOCI build process.  but we've added added required packages to the LOCI images before (things like the fluent handler for oslo.log come to mind)
15:37:53 <jayahn> or we can even publish kolla-images we are building internally. >> will it provide any value for osh project?
15:39:10 <evrardjp> I am no core but please allow me to give my opinion
15:39:15 <srwilkers> evrardjp: of course
15:39:19 <jayahn> sure
15:39:56 <srwilkers> jayahn: for what its worth though, we've got other charts that use kolla images currently (ceilometer comes to mind)
15:40:34 <evrardjp> I don't think it's a good idea to extend too much the amount of different image build process that will be included in osh repositories. So I'd say, do not focus on adding more images if there are any existing images that match your purpose
15:41:08 <evrardjp> Second, I'd say -- if you want to upstream a chart based on image x, you should be able to allow any deployer to use image y.
15:41:30 <evrardjp> should (s)he want to be alone, let him/her/it be :)
15:42:31 <evrardjp> portdirect: should talk about the necessity of building with LOCI to get an upstream chart in, but I don't believe it must be a necessity
15:42:34 <evrardjp> myself
15:42:42 <evrardjp> (that's for situation 1)
15:43:28 <evrardjp> for situation 2: I would strongly hope that changes in values would be documented throught the help of anything (including release notes). This is now the "API" of the project, according to me
15:44:18 <evrardjp> point 3) Documenting what needs to be done for using kolla images vs LOCI images is good -- I think that's a valid "user story"
15:44:41 <evrardjp> should someone want to run with its own images, he could, by following the documentation
15:45:12 <jayahn> FYI, replacing loci base images into kolla-based one is not 1:1 mapping. (as you probably knows). for loci, it is neutron, for kolla, it will be multiple images, neutron-api, neutron-server, .... sometimes can be tricky. :)
15:45:15 <evrardjp> and last point -- I am sure the people in #openstack-loci can help you with the building of the images :D
15:45:39 <jayahn> evrardjp: #openstack-loci it is. :)
15:46:07 <srwilkers> i think engaging individuals in LOCI would be the best path forward
15:47:11 <jayahn> for us, any effort we are putting on loci, is purely for upstream contribution. we are willing to do it, but really needs some help to shorten our trial and errors. :)
15:47:13 <evrardjp> I think the OSH should keep its DNA of being "image independant", but still provide "reference" images that help do the things with minimum value overrides basically
15:48:07 <evrardjp> there is already an octavia image btw :)
15:48:10 <jayahn> so one additional document entry: reference information when you want to use kolla images.
15:48:31 <jayahn> evrardjp: we found out that, we are currently testing with octavia loci image. :)
15:48:35 <evrardjp> if you don't mind I would prefer a different title
15:48:43 <evrardjp> "How to bring your own images"
15:48:58 <jayahn> that can work.
15:49:12 <evrardjp> assuming you're not using kolla, how would you do? (for example you build your own with Dockerfiles, or ansible, or dib...)
15:49:35 <evrardjp> I think a practical example with Kolla would be great though, as you have the content
15:50:04 <jayahn> evrardjp: you need to completely modify "images section in value.yaml" to meet with your images.
15:50:50 <jayahn> currently, images keys are compatible with kolla-images, that is why, it is easy for us to set kolla-based images.
15:51:16 <jayahn> if you build your own stuff, you need to changes everything, not only values, but also keys
15:51:41 <evrardjp> all of it is worth documenting, IMO :)
15:51:47 <jayahn> yeap.
15:52:25 <srwilkers> agreed
15:52:35 <srwilkers> anything else on this topic?  we've got about 7 minutes left
15:52:43 <evrardjp> none
15:53:09 <srwilkers> jayahn?
15:53:12 <evrardjp> "help needed"
15:53:19 <jayahn> one additional question, if we would like to build loci images for osh chart, would we contribute it to osh-images-repo?
15:53:32 <evrardjp> yes
15:53:36 <portdirect> jayahn: yup
15:53:39 <srwilkers> jayahn: yeah, that makes sense
15:53:45 <jayahn> okay
15:54:18 <evrardjp> seems portdirect srwilkers and I are aligned -- doomsday is near
15:54:28 <srwilkers> :o
15:54:40 <srwilkers> #topic storyboard
15:54:48 <jayahn> only three aligned,  no doomsday yet
15:55:19 <evrardjp> :)
15:55:25 <srwilkers> just a quick touch here -- i know portdirect is still working on reviewing whats in storyboard
15:55:56 <portdirect> but whats there to date - should be good to get going on as its leading towards the 1.0.0 release
15:56:12 <evrardjp> great!
15:56:15 <portdirect> in particular we really need to get the docs and helm driven tests in shape
15:56:34 <portdirect> (sorry ive not caught up on scrollback - so may be rehashing things)
15:56:39 <evrardjp> (maybe worth giving a link for the record?)
15:56:55 <srwilkers> https://storyboard.openstack.org/#!/worklist/341
15:57:37 <portdirect> I'd say for people looking to get invovled this would be a great place to start
15:57:57 <portdirect> evrardjp: it would be great to see some helm tests for services that dont have coverage atm
15:58:11 <portdirect> as it would really help validate multi distro support amongst other things
15:58:48 <portdirect> jayahn: i know that we need to give more attention to docs than we have recently
15:59:11 <portdirect> do you have the bandwidth to enage there atm?
15:59:15 <portdirect> *engage
15:59:17 <evrardjp> portdirect: agreed on the helm tests.
15:59:36 <srwilkers> we've got less than a minute
15:59:40 <srwilkers> let's take this to #openstack-helm
15:59:47 <evrardjp> I will although focus on the image building steps to get all of this out of the door. Maybe I should all of those into storyboard too?
15:59:49 <srwilkers> thanks for coming everyone
15:59:53 <evrardjp> thanks srwilkers !
15:59:58 <srwilkers> #endmeeting