15:00:09 <srwilkers> #startmeeting openstack-helm
15:00:10 <openstack> Meeting started Tue Aug 15 15:00:09 2017 UTC and is due to finish in 60 minutes.  The chair is srwilkers. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:11 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:14 <openstack> The meeting name has been set to 'openstack_helm'
15:00:20 <jayahn> o/
15:00:24 <alraddarla> o/
15:00:27 <srwilkers> #topic roll call
15:00:30 <srwilkers> o/
15:00:39 <raymaika> o/
15:00:40 <lrensing> o/
15:01:01 <portdirect> \o
15:01:12 <srwilkers> hello everyone :)
15:01:21 <srwilkers> here's the agenda for today: https://etherpad.openstack.org/p/openstack-helm-meeting-2017-08-15
15:01:33 <jayahn> today was holiday here. :) it seems lots of things happening over there cross the pacific ;) long meeting note.
15:01:52 <srwilkers> got quite a bit of information on it already, but we can take a few minutes to add anything else
15:02:04 <v1k0d3n> o/ (sorry late)
15:02:13 <srwilkers> nice jayahn -- hope you enjoyed the holiday :)
15:02:45 <mattmceuen> o/ GM all
15:03:25 <srwilkers> hey mattmceuen
15:03:31 <srwilkers> welcome to the party
15:03:39 <srwilkers> let's go ahead and get started
15:03:43 <srwilkers> #topic PTG reminder
15:03:58 <lamt> o/
15:04:15 <srwilkers> just a friendly reminder -- the PTG's coming up quickly, so wanted to call attention back to the etherpad we had running for discussions: https://etherpad.openstack.org/p/openstack-helm-denver-ptg-topics
15:04:48 <jayahn> is there any schedule for ptg meeting, especially when we have meetings during five days of PTG.
15:04:52 <srwilkers> also, since we're a hosted project, we wont have an officially scheduled time.  i do know myself, portdirect, lamt and alraddarla will be there throughout the entire week
15:05:16 <portdirect> think alanmeadows should be around too
15:05:18 <srwilkers> we'll be responsible for identifying our own time to meet unfortunately, as we cant be put on the schedule officially
15:05:26 <srwilkers> its hard to see him in that black fleece
15:05:30 <srwilkers> he melds into everything
15:05:48 <jayahn> I would like to leave early if possible, probably on Thursday. will it be possible to have all openstack-helm meetings between Mon - Wendesday? just asking. :)
15:05:58 <srwilkers> jayahn: absolutely
15:06:11 <jayahn> great. then, i will finalize my flight. :)
15:06:29 <srwilkers> jayahn: awesome.  look forward to seeing and chatting with you :)
15:06:30 <portdirect> jayahn: but then we miss you on friday night :(
15:06:44 <jayahn> yeah.. that is true.
15:06:51 <alraddarla> i'm leaving friday night too :(
15:07:29 <srwilkers> #topic Libvirt/OVS breakout
15:07:30 * portdirect can come out for 1 (or 8) on thurs - but has work to do on friday :(
15:07:42 <srwilkers> portdirect: floors yours
15:08:23 <portdirect> hey, so I've started work on pulling libvirt and ovs out of their parent charts
15:08:37 <portdirect> they were odd-ducks in osh as they included the supporting infra
15:08:46 <portdirect> making it hard to plug in other backends
15:08:57 <portdirect> and when they were - they seemed like 2nd class citizens
15:09:20 <portdirect> this was also causeing us some trouble when we wanted to use theservice (gasp) without openstack
15:09:31 <v1k0d3n> this is really good to see.
15:09:37 <v1k0d3n> is there a blueprint?
15:09:44 <portdirect> the two ps's are up in gerrit - and I hope to have fininshed by eod
15:09:49 <portdirect> though reviews welcome
15:09:55 <portdirect> v1k0d3n: https://blueprints.launchpad.net/openstack-helm/+spec/split-nova-and-neutron-infra
15:10:13 <v1k0d3n> great! thanks portdirect
15:10:52 <portdirect> this will also make the neutron work a lot eaier i think for onos and other sdn backends to come in
15:10:56 <v1k0d3n> that makes things much more flexible. was wondering how the team was going to implement this, although it was pretty obvious this is probably the best path.
15:10:57 <portdirect> *easier
15:11:27 <jayahn> okay. +1 on breaking out these two.
15:12:17 <srwilkers> nice.  any other comments on this topic?
15:12:30 <v1k0d3n> docs are going to be included as well portdirect ?
15:12:37 <portdirect> check the ps :)
15:13:17 <v1k0d3n> i saw that. but other than just copy paste...i mean explanation on why, how this open things up to be more flexible.
15:15:22 <srwilkers> i think a mention of the flexibility it adds would be sufficient in my eyes -- the goals always been to be pluggable, and this really brings the charts in line with the message we've been trying to sell
15:15:39 <portdirect> Its fits the mission statement of osh i think?
15:16:49 <srwilkers> i think so.  i think the addition of varying backends would require a bit of documentation covering any of the specifics that might shed light on what they offer, but i'd like to see those added as the additional backends are brought in
15:16:52 <v1k0d3n> this is probably really easy to see for developers or those close to the project....
15:17:21 <portdirect> https://github.com/openstack/openstack-helm#mission
15:17:46 <v1k0d3n> however, i think it's a little lost that users struggle with OSH, and we're really trying to win over new operators and contributors. expanding on ps like these are the "ah-ha" moments that operators need.
15:18:08 <v1k0d3n> referring back to the mission doesn't really explain how each knob works.
15:18:53 <srwilkers> i think we need to balance which knobs need explaining -- there's two extremes i think.  too little explanation and too much.  i'm really fond of having self-documenting code and having it structured to represent that
15:19:33 <portdirect> +2
15:20:42 <portdirect> perhaps this is a perfect moment to lead into the next topic?
15:21:27 <v1k0d3n> i agree with balancing knobs. but only documenting cut/paste deployment commands shouldn't really be the answer.
15:21:30 <srwilkers> there's a few items listed under the documentation topic -- v1k0d3n portdirect, floors yours
15:21:42 <srwilkers> #topic Documentation
15:21:45 <v1k0d3n> portdirect: you could just reach back out to the community to see if someone could take over the ps and document some things for you.
15:21:46 <portdirect> v1k0d3n: take it away :)
15:22:05 <v1k0d3n> this is a great thing for a lead to do...and then that could be a way to engage the community better, and share work.
15:22:22 <portdirect> v1k0d3n: a follow on ps would be great
15:22:44 <portdirect> though not opposed to anyone adding docs to it :)
15:23:07 <v1k0d3n> ok, so to these points...i think we (collectively) could open opportunities for new contributors, and also be a bit more welcoming by asking for help with docs.
15:23:39 <portdirect> v1k0d3n: would be  great place for you to get involved again
15:23:47 <v1k0d3n> it would be nice to follow _almost_ a pair programming model where a lead contributes a large ps, and then works with a community member for providing docs.
15:23:49 <portdirect> as we've missed having you push work :)
15:23:58 <v1k0d3n> i'm somewhat back :)
15:24:53 <v1k0d3n> and as srwilkers knows...our group is starting to get more engaged.
15:24:58 <srwilkers> im open to getting new contributors working on the docs - it'd be a great place to get started.
15:25:33 <v1k0d3n> well, and if we have folks like portdirect srwilkers or others work with new contributors...people will feel like they're part of it.
15:26:13 <v1k0d3n> and i think we'd gain the momentum we want/need.
15:26:23 <v1k0d3n> along with solving some of the missing docs issues.
15:26:33 <v1k0d3n> so srwilkers can you come up with a list of documentation gaps?
15:26:42 <v1k0d3n> i can have our folks work with you to help fill in some of those...
15:27:15 <srwilkers> i think it'd be great for people identifying them to submit bugs about the shortcomings.  unfortunately its the feedback from those who can see something missing or that isnt clear that really helps shed light on what we need
15:27:43 <v1k0d3n> but more communication on IRC would be good too. blueprints would be good. in fact moving to storyboard would help a lot...so some load would be distributed more with the community.
15:27:58 <v1k0d3n> i want it to _feel_ like a community project...not just coming from one source.
15:28:35 <portdirect> v1k0d3n: could you make a list of the docs gaps and get your team on it? would be a great way to change the fear you have
15:28:47 <v1k0d3n> i don't have time portdirect
15:28:53 <v1k0d3n> sorry
15:29:09 <portdirect> :( I fear we are all in the same boat then
15:29:10 <v1k0d3n> plus, if i remember correctly...i thought were were supposed to accept ps with documentation.
15:29:29 <v1k0d3n> so this just executes on that plan
15:29:59 <v1k0d3n> and still leaves the hard parts to good devs like yourself. and pairs you up with someone who could learn from you from the community.
15:30:39 <v1k0d3n> if it's not a good idea...i'm fine with dropping it. i just think i'm seeing a bit more 3D coming to a new company, and hearing the questions directly from non-OSH devs.
15:30:54 <mattmceuen> There's already documentation in the PS.  I think there's a good opportunity for follow-on documentation that describes overall design philosophy.
15:30:59 <v1k0d3n> i gotta run to a meeting. srwilkers let me know what your thoughts are.
15:31:14 <srwilkers> cheers v1k0d3n -- grab coffee on the way
15:31:21 <v1k0d3n> so a one liner in the docs is enough for libvirt moving to a new chart?
15:31:22 <v1k0d3n> https://review.openstack.org/#/c/493738/5/doc/source/install/developer/all-in-one.rst
15:31:29 <v1k0d3n> ok...i bow out then :)
15:32:00 <srwilkers> let's move on
15:32:27 <srwilkers> #topic Images
15:32:33 <srwilkers> portdirect: floors yours
15:32:40 <portdirect> lol
15:32:50 <portdirect> 2 sec
15:33:20 <portdirect> think a copy/paste from the etherpad is relivant here:
15:33:27 <portdirect> The kolla 3.0.3 images have done us pretty well - though have shortcomings:
15:33:38 <portdirect> many severe issues are in the docker.io pushed images - and kolla will not update them.
15:33:38 <portdirect> - the lack of httpd/uwsgi in the containers prior to the 5.0.x series makes forward looking dev hard - and also is limiting when wanting to use internal tls (unless we go down the sidecar container route - yuk)
15:33:38 <portdirect> - This also make graceful handover of services on rolling upgrades a bit more janky than it has to be - apache is much better than eventlet here.
15:34:00 <portdirect> As a result would making and pushing new set of images make sense? we used to use the stackenetes images - as they encountered similar issues - and I'm thinking it may be time to do the same so that we can have a set of similary non-production images to use for dev and demo purposes.I
15:34:58 <srwilkers> ive been back and forth on this one
15:35:12 <srwilkers> but the more we've used the kolla images, the more issues we've found with them
15:35:22 <jayahn> new set of images, could you be more specific?
15:35:54 <portdirect> I'm thinking of a set of images, built from kolla-build 5.0.0 for ubuntu source
15:36:16 <portdirect> that would contain all the current fixes to openstack that are not in the images kolla pushed to dockerhub
15:36:28 <portdirect> and also have apache/uwsgi for apis
15:37:34 <jayahn> ah.. i got that.
15:37:39 <portdirect> in addtion to the reasons above - one of the main advantages of moving to a kolla-build 5.x derived image is fixed uid/gid for service users
15:37:54 <portdirect> which is a super big win for security
15:38:29 <lamt> ++
15:38:54 <srwilkers> i assume these'll get built and live in quay, and those become our reference?
15:38:56 <jayahn> i am totally okay with building a set of images from kolla-build, not using docker.io one.
15:39:13 <portdirect> srwilkers: thats what I think wuld be appropriate
15:39:46 <portdirect> we could include the refs to making them in gh - so they were totally reproduceable
15:39:57 <lrensing> i'm in support of that, build new and host on quay
15:40:06 <portdirect> but host and donate them to the community
15:40:12 <lrensing> yes
15:40:12 <portdirect> nice, cheers lrensing
15:40:38 <srwilkers> okay, lets give it a go and try it out
15:40:43 <jayahn> if i understand correctly (just for confirmaiton on my understanding)
15:40:54 * portdirect awats input
15:41:02 <srwilkers> its something ive been thinking about as we've found other services we need to build images for
15:41:03 <jayahn> we start building a set of openstack images with kolla-build tool 5.0.0
15:41:12 <jayahn> host those images in quay
15:41:23 <jayahn> use that image as reference in openstack-helm project.
15:41:28 <jayahn> am i getting right?
15:41:32 <portdirect> yes
15:42:13 <portdirect> I can host the images while a WIP - and then once things move towards mergable push to quay
15:42:46 <srwilkers> seems we have consensus for the most part?
15:43:08 <portdirect> srwilkers, lrensing, jayahn?
15:43:11 <srwilkers> portdirect: yeah, sounds good
15:43:31 <jayahn> yes. just one quick question. who will be responsible for making sure we are building good ones? :)
15:43:44 * portdirect looks other way
15:43:49 <jayahn> will we be putting some minial check routin on image buidling process?
15:44:00 <portdirect> the gate in osh
15:44:15 <portdirect> but to confirm - these are *NOT* ment to be production images
15:44:23 <jayahn> yeap. i got that. :)
15:45:14 <jayahn> i can let you or stacey to talk with Wil, to get some basic bats test we are using internally to check images from kolla-build.
15:45:30 <portdirect> jayahn: that would be awesome
15:45:33 <jayahn> although we are not using 5.0.0...
15:45:35 <jayahn> :)
15:45:37 <srwilkers> lrensing, lamt:  you on board?
15:46:22 <lrensing> i am in support of this
15:46:33 <srwilkers> awesome
15:46:48 <srwilkers> let's go ahead and move on then
15:46:59 <srwilkers> #topic neutron decomposition
15:47:03 <srwilkers> think this was you jayahn :)
15:47:05 <srwilkers> floors yours
15:47:20 <jayahn> ah. nothing special.
15:47:47 <jayahn> just want to check "what urgency do you guys have in mind to make it merge". :)
15:47:59 <portdirect> ok - s one bit of fb i have is here: https://review.openstack.org/#/c/481225/
15:48:20 <portdirect> I'm really not comfortable with the logic being inserted into the configmaps
15:48:28 <jayahn> i saw new review on the neutron decomposition ps, which i have to make siri to think more.
15:48:34 <jayahn> yeah. about that.
15:49:07 <jayahn> i will have internal discussion tomorrow about that, and will back with our newer idea.
15:49:15 <portdirect> nice - cheers jayahn
15:49:30 <portdirect> the decomposotion model is now in all openstack charts
15:49:37 <portdirect> I'm fine with other knobs being added
15:49:43 <jayahn> just, FYI, we are kind of urgent to build an ci pipeline around SONA integration.
15:49:47 <lamt> srwilkers : yes, sorry was distracted by other convo
15:49:54 <portdirect> but am worred that if we go to far then the templates will become a mess
15:50:21 <portdirect> and my gut says a 'drop-in' replacement would be better
15:50:39 <portdirect> than adding a lot of logic to the current template
15:50:47 <jayahn> so wanted to make it work from upstream, but will get you input and will think more about that.. after all, it is about making it better and simple to use.
15:50:58 <portdirect> +2 cheers jayahn
15:52:30 <srwilkers> awesome
15:52:32 <srwilkers> anything else here?
15:54:07 <srwilkers> #topic open discussion
15:54:19 <srwilkers> we've got a few minutes left.  any other items anyone wants to chat about?
15:54:51 * portdirect has this terrible pain in all the diodes down his left side
15:55:28 <srwilkers> alright, ill go ahead and close it out.  see you all in the channel
15:55:35 <srwilkers> #endmeeting