16:01:02 #startmeeting kolla 16:01:02 hey 16:01:04 Meeting started Wed Jan 25 16:01:02 2017 UTC and is due to finish in 60 minutes. The chair is inc0. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:06 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:09 The meeting name has been set to 'kolla' 16:01:18 #topic w00t for kolla 16:01:21 Hi 16:01:23 w00t people 16:01:24 ;) 16:01:25 o/ 16:01:28 o/ 16:01:28 o/ 16:01:30 o/ 16:01:30 Hai 16:01:31 oops 16:01:31 o/ 16:01:33 hi 16:01:40 w00t 16:01:41 o/ 16:01:42 good afternoon 16:01:43 o/ 16:01:51 * portdirect w00t 16:01:54 good afternoon 16:02:10 :is all out of woots to give 16:02:29 and it's only 8am? your day is looking bright srwilkers 16:02:43 woot o/ 16:02:58 o/ 16:03:08 o/ 16:03:15 o/ 16:03:58 #topic agenda 16:04:06 * Roll-call 16:04:06 * Announcements 16:04:06 * Documentation (pbourke) 16:04:06 * Ocata kolla deliverable (inc0} 16:04:06 * Ocata kolla-ansible deliverable (inc0) 16:04:07 * kolla-kubernetes 0.5.0 (sdake) timebox [10 minutes] 16:04:09 * deprecation of Debian images (berendt) 16:04:11 * Debian images for ARM architecture (gema, berendt) 16:04:13 * Finalize Brainstorming of PTG topics (we have 2 days of PTG space) timebox [10 minutes] 16:04:15 * Open Discussion 16:04:18 #topic announcements 16:04:35 We came into agreement how to add new deliverables 16:05:04 it will be change to this file https://review.openstack.org/#/c/424784/ 16:05:22 and merge as usual with caveat of PTL +2 present 16:05:44 any community announcements? 16:06:12 guess not:) let's get to business 16:06:16 #topic Documentation 16:06:19 pbourke, you're up 16:06:41 ok, simple enough really 16:06:41 o/ 16:06:42 o/ 16:07:03 I had a short note on the ML but the gist is, I think some people are working in the background on revamping the kolla docs 16:07:14 I think a lot of people have grand visions of what needs to happen 16:07:28 till then though, Id like to suggest we take some immeidate steps to remove duplication 16:08:00 do we have a doc liason for the docs team? 16:08:03 by that I mean strip down the kolla section to just building info and info related to the image themselves, along with hyperlinks to kolla deliverables that can deploy the images 16:08:21 \o 16:08:31 if people like this idea I can submit a change before end of the week 16:08:40 pbourke: i like it :) 16:08:50 pbourke: +1 16:09:05 berendt, officialy no, but we have people talking to them 16:09:07 i have a problem, the document we have to refer to what to do 16:09:07 i think we should think about a central starting page 16:09:22 and some guide which tool to choose 16:09:30 yes 16:09:40 pbourke, +1 but Kolla need including some common info too 16:09:48 sounds like no major concerns with the above so that's it from me :) 16:09:48 wo should need some guide and tools 16:10:09 pbourke, how about QSG? 16:10:18 generic information in kolla docs? 16:10:36 berendt, sure 16:10:38 duonghq: quick start guide is related to ansible/salt/k8s 16:10:47 every deliverable needs a qsq 16:11:18 +1 to qsg 16:11:25 im agree with pbourke. 16:11:26 so we need one landing page in Kolla and point to deliverables qsg? 16:11:55 kolla QSQ just solve image build. kolla-ansible QSQ just explain how to deploy images. 16:11:57 duonghq: i think this is the best way 16:12:12 agreed with berendt 16:12:17 again, Id rather not go into all the various improvements needed. My main goal is just to remove the current duplication 16:12:21 Jeffrey4l_: +1 16:12:46 for QSG, we should use images from docker hub (IMO) 16:12:48 pbourke: then i think the best is to start with the split of the qsg and to remove common content from kolla-ansible 16:12:50 +1 pbourke 16:13:02 duonghq: +1 but we also need a qsg how to build images 16:13:20 i agree with pbourke the first step is seeing what we have, kolla-k8s in particular has a lot of very out of date docs, that need removed or reworked before we promote them 16:13:23 well we do have good doc on it 16:13:28 just not called qsg 16:13:31 berendt, not really qsg 16:13:33 duonghq: +1 about building images 16:13:41 we can link to it from kolla-ansible qsg 16:13:43 yes, maybe qsg is the wrong wording for it 16:14:03 also, qsg itself can ask people to use dockerhub images 16:14:07 to make stuff even easier 16:14:10 berendt, something like Basic building instruction (w/ some advanced instructions...) 16:14:24 we need qsg, event though current "qsg" is too complicate. 16:14:39 does it makes sense the pbourke prepares a list for the next meeting? 16:14:40 berendt: this doc will be the starting point. right ? 16:14:41 yeah, iirc, sdake is working on that 16:14:42 deploy vs building qsg - have a switch at the starting page 16:14:48 trinaths: yes 16:15:23 berendt: then let it be qsg. it gives a good meaning too 16:16:04 i think we need a outline which explain how the doc will looks like in the future in overview. pbourke 16:16:19 honestly Id rather not commit to that 16:16:21 at least right now 16:16:35 +1 Jeffrey4l 16:16:37 yeah we have release 16:16:44 to reiterate, all I want to achieve is that when you go to http://docs.openstack.org/developer/kolla-ansible/ vs http://docs.openstack.org/developer/kolla/ they actually look different 16:16:52 include it in the ptg? 16:17:00 right now we have patches going into both, it makes no sense 16:17:16 Jeffrey4l_: oultine mean future plans? 16:17:19 it should have happened long ago 16:17:23 let's freeze kolla-ansible/docs until we start working on it 16:17:52 trinaths, yep. bird's eye. 16:17:52 sorry guys but the improved docs stuff should go in Ocata or Pike? 16:18:16 hello 16:18:18 i think in pike 16:18:23 meeting started ? 16:18:27 zioproto: yes 16:18:35 ok, lets move on 16:18:37 Jeffrey4l_: i believe it to have a roadmap line. and doc t more concentrate on bringing up the environment 16:18:43 zioproto: yes 16:18:46 so PTG would be better place for further discussion? 16:19:00 (I'm going to cannibalize kolla-k8s from this meeting, sorry, we're in release week( 16:19:08 #topic ocata-3 16:19:12 duonghq: i think so 16:19:16 https://launchpad.net/kolla-ansible/+milestone/ocata-3 16:19:45 https://launchpad.net/kolla/+milestone/ocata-3 16:20:17 the list is long and full of not implemented 16:20:39 inc0, which day we have to release m3? 16:20:43 Jeffrey4l_, we'll tag ocata-3 at 27th right? 16:20:47 this week 16:20:53 27th is last day 16:21:02 yep. 27th is the last day. 16:21:36 which means I'll be bumping blueprints to Pike 16:21:43 which aren't finished by then 16:21:46 and there is not assignee and the there is good progress. 16:23:17 zhubingbing, whats the status of fluentd? 16:23:20 is it finished? 16:23:26 inc0 yep. 16:23:28 yes 16:24:01 anyone knows anything about keystone upgrade? 16:24:03 does it include an upgrade from heka to fluentd? 16:24:29 keka replaced fluentd's work has been completed 16:24:30 berendt, yep. heka will be removed during upgrade from newton to ocata 16:24:44 changed to implemented 16:24:46 good job! 16:24:53 Jeffrey4l_: good :) 16:25:01 https://blueprints.launchpad.net/kolla/+spec/ks-rolling-upgrade-role <- this needs to happen 16:25:33 sp___, are you there? 16:25:33 this should be moved to P cycle ;( 16:26:06 without this KS might break upgrade, and have right to do so 16:26:46 inc0: will that also handel change such as enableing fernet tokens correctly on upgrade? 16:26:52 seems it is not started. i am afraid this can not be done before release. 16:27:14 sean-k-mooney, i do not think so 16:27:16 Jeffrey4l_, it need some rework due to recent ansible refactoring 16:27:16 sean-k-mooney, negative, this will just make sure we'll handle db changes correctly 16:27:39 yup, is under a bug_id instead of the bp https://review.openstack.org/#/c/388544/ 16:27:41 https://review.openstack.org/#/c/405215/2 there is some code for it 16:28:07 duonghq, thoughts? you wrote most of the code 16:28:23 think you can get it done? 16:29:26 i can do this work ;)if duonghq have free time;) 16:29:26 duonghq, if tomorrow I can get it done than it can be done, or not, :( 16:29:38 ok, try it 16:29:43 roger 16:29:58 let me know if you need help, I can take on it too 16:30:26 ok next one https://blueprints.launchpad.net/kolla/+spec/api-interface-bind-address-override 16:30:28 me too 16:30:32 I think it's no longer relevant 16:30:57 k8s do not reply on this? 16:31:18 depends on* 16:31:39 inc0, ^ 16:31:50 well no movement in this change and k8s seems to handle itself 16:31:53 kfox1111, comments? 16:32:42 whether k8s depends on orchestration_engine variable? kfox1111 inc0 16:33:05 inc0 that was well before my time - but looks like it was working on the assumpption of running in the hosts network namespace? 16:33:42 yeah, and we're not really using host networking everywhere right? 16:33:47 not for apis anyhow 16:34:03 we are using it as little as possible 16:34:14 Jeffrey4l_, yeah this was really mitaka 16:34:21 I'm going to close it for now 16:34:23 so though nice to have I dont think its causing any issues at present 16:34:41 portdirect, would you be kind enough and double check? we can always reopen 16:35:01 yeah i'll ask kfox and co, np 16:35:37 https://blueprints.launchpad.net/kolla/+spec/apply-service-upgrade-procedure 16:35:49 this is related to keystone 16:35:51 inc0: is k8s using the docker proxy for apis if so that could be a scaleing problem though i understand why you might want to avoid net=host 16:36:13 sean-k-mooney, k8s doesn't really uses docker proxy afaik 16:36:22 it has it's own network overlay 16:37:08 it have some network option, the recommend is flannel 16:37:33 inc0: ack, it should be ok then. 16:37:43 o/ 16:37:50 but back to topic, duonghq I assume someone will have to take on neutron too right? 16:38:19 or I see you're working on this change 16:38:27 we support both external and host ports currently in kolla-kubernetes. 16:38:31 so let's do this, you focus on neutron 16:38:33 about the apply-service-upgrade-procedure, the neutron's ansible potion run into same problem as ks: need to be migrated to new ansible structure 16:38:38 and me and egonzalez will take keystone ok? 16:38:44 since ingress is starting to stabalize, we may support that at some point too. 16:38:48 inc0, nice, 16:38:53 thanks 16:39:25 just get neutron done:) egonzalez I'll start today and by your morning you should have something to work on:) 16:39:33 feel free to coauthor patch 16:39:44 thank inc0 and egonzalez 16:39:58 inc0: roger, i'll check what you got at the morning 16:40:38 https://blueprints.launchpad.net/kolla/+spec/support-python-35 this is doen? 16:41:16 inc0, it should be depend on code coverage improvement bp 16:41:59 inc0: Python 3.5 support in some openstack projects is still experimental. i assume this is for the kolla module rather then the containers? 16:42:10 sean-k-mooney, yup 16:42:15 duonghq: cool 16:42:18 sean-k-mooney: yes, should be for kolla itself 16:42:29 yeah, I'll bump it to pike 16:42:33 and all the rest in fact 16:42:35 for all Python code, indeed 16:42:40 since these are highs/essentials 16:42:55 #link https://launchpad.net/kolla-ansible/+milestone/ocata-3 16:43:23 inc0: would you like me to pre-emptively bump dpdk,odl and hugepage configuration to pike? i doubt they will be merged by the 27th and they are also not currently targeted to ocata-3 16:43:27 inc0, https://blueprints.launchpad.net/kolla/+spec/coverage-increment-for-kolla -> here, last bp should depend on this 16:43:35 sean-k-mooney, yes, please 16:43:40 sean-k-mooney: hugepage yes 16:43:52 sean-k-mooney: i do not have time to work on it at the moment and prio is low 16:44:02 duonghq, then i'll keep this one 16:44:08 since it might be merged 16:44:14 but please, focus on upgrades 16:44:32 sure 16:44:45 berendt: i have some patches for it up for bootstrap servers playbook. 16:44:48 I'll bump all meds/lows 16:45:21 kolla-ansible doesn't look bad 16:45:50 ok, let's move on 16:46:08 #topic Debian deprecation 16:46:12 forgot, the apply service upgrade should be done for all service that have rolling upgrade procedure, not just ks and neutron, but ks and neutron have most mature procedure, so in general, part of the bp should be done on Ocata, other in Pike or next... 16:46:15 gema: around? 16:46:17 berendt, gema we wanted to discuss 16:46:18 yep 16:46:23 your turn 16:46:38 gema, question, do you *need* debian containers for ARM? 16:46:46 or ubuntu would work? 16:46:53 inc0: our current release is based on Debian and CentOS 16:47:04 so we were planning to produce containers first for debian then for centos 16:47:04 ok, so CentOS would work? 16:47:18 inc0: yes, but we prefer debian 16:47:29 issue is, nobody else prefers debian;) 16:47:30 since it is a purely community driven project 16:47:39 inc0: all our members are keen on it 16:47:45 most :D 16:47:55 gema: are you willing to working on the debian images? 16:48:00 if you volunteer to be maintainers of it in kolla, we can discuss keeping it 16:48:00 berendt: yes 16:48:03 and please add job gate for debian ;) 16:48:12 yeah, gate job for debian is first step 16:48:20 we'll be happy to help you do it 16:48:22 first step is making the debian images working again 16:48:28 we will be working on debian ARM64, I am not sure how much work is involved in also making them work on x86_64 16:48:33 next step gates for debian 16:48:34 but we don't have hardware to test that 16:48:56 arm64 is the next topic, i think we should discuss the addition of a new platform independent of the debian deprecation 16:48:58 gema, it will have to work on both arm and x86 as gates are all x86 16:49:13 gema i would not expect there to be much if any difference between x86_64 and armv8 in this context 16:49:15 i prefer to add a non-voting debian gate, then fix all issue to make it green. 16:49:22 inc0: we will be adding a 3rd party CI for ARM64 16:49:23 Jeffrey4l_: +1 16:49:31 Jeffrey4l_: agreed 16:49:40 gema, fine, but we also want proper openstack CI 16:49:45 gema, re 3rd ci , cool. 16:49:49 berendt, inc0: plus I can only commit resources to do this for pike 16:49:49 and yeah, I don't expect debian to be voting anytime soon 16:50:02 ocata is done anyway 16:50:03 we'll be ramping up in the meantime 16:50:09 ok, good 16:50:43 when adding debian for arm64, can the current images be used or do you have to add it as a new distro? 16:50:46 so order of business would be: 16:50:50 oh. that reminds me. are we in feature freeze yet? 16:50:52 berendt: I don't know 16:50:53 at the moment the debian images are nearly the same like ubuntu 16:50:53 1. make sure debian works at all 16:50:58 berendt: it would have to be a new distro 16:50:59 kfox1111, k8s is not 16:51:07 ansible and kolla yeah 16:51:11 no, kollla. I'd still like to see https://review.openstack.org/#/c/422950/ go in. 16:51:14 berendt: the binaries in the current image would not work on arm 16:51:19 when it would to be a new distro we can deprecate current debian images and add debian arm64 as new images 16:51:24 sean-k-mooney: corrent 16:51:26 its an additional set of containers, so shoudl be very low risk. 16:51:26 correct 16:51:27 kfox1111, I'll review it in a moment 16:51:31 inc0: thx. 16:51:35 yeah, I'm not concerned 16:52:01 yeah berendt that's good point 16:52:06 we can have distro debian-arm 16:52:11 and deprecate x86 16:52:15 i think the best way is to deprecate current debian images in kolla, to not deprecate debian support in kolla-ansible 16:52:19 i would actully thik it would be best to add an arch flag not a distro 16:52:31 this way gema can re-add working debian arm64 images 16:52:36 with a spec 16:52:36 not sure level of effort is that much different for arm vs arm+x86. 16:52:41 the packages should all be the same. 16:52:44 inc0: do you have contact with stackenetes at all? they use kolla debian 16:52:59 Id expect most of the work to be arm-kvm related really. 16:53:08 I know, but they don't work of it at all any more 16:53:30 gema: does this work for you? deprecation of debian images this cycle? 16:53:40 this way we are able to remove them next cycle 16:53:41 I like idea treating debian-arm is pseudo distro 16:53:48 berendt: yes, as long as we are allowed to work on the ARM64 ones going forward, that works 16:54:00 gema: yes of course 16:54:01 we will likely also work on centos-arm 16:54:08 but I see that as a future thing 16:54:12 let's make it distros 16:54:23 but it makes no sense to keep the x86 images when nobody works on them and you are only willing to work on arm 16:54:24 since there will be 3 of us only working on this part time 16:54:25 we don't need to change build for these two 16:54:37 there is a PS to add powerservers support, already are changing to support different archs in kolla build IIRC 16:55:05 do we need an other vote when we change the scope of the deprecation? 16:55:09 gema, will we see you in Atlanta? 16:55:14 i think long run having an arch flag would be better then distos per arch 16:55:15 just to throw the question out there, 16:55:25 inc0: yes, I will try to make time from the interop meetings to come visit you guys 16:55:28 we will keep debian in kolla-ansible, we will deprecate current debian images 16:55:30 do we beleive in container space, arches are as much work as seperate distros? 16:55:34 berendt, let's discuss it on PTG 16:55:43 I really don't think they are. 16:55:48 ptg is too late for a deprecation? 16:55:53 egonzalez, power support https://review.openstack.org/423239 16:55:59 kfox1111: I am not sure of the implications of either 16:55:59 kfox1111, if you look at our debian 16:56:08 it's 90% of ubuntu with few exceptions 16:56:11 Jeffrey4l_: thanks 16:56:16 kfox1111: they should be minimal change unless except when building cross arch images e.g. arm images on windows 16:56:22 inc0: ah. so the debian's not debian. 16:56:37 inc0: thats still a seperate issue of, do we support a debian with x86/arm, or just arm. 16:56:38 well names of packages are mostly the same anyway 16:57:04 we narrowed it down from "nobody cares for debian" to "nobody cares for debian on x86" 16:57:06 I am pretty sure the instructions for building a debian based arm container once multiarch is in kolla's build scripts, will be essentially the same as for x86. 16:57:07 inc0: yes currently we dont use the full name 16:57:15 I think that basically comes for free. 16:57:24 inc0: we excluded the arch and that is determined by apt/yum 16:57:29 kfox1111: if that is the case, we will add it 16:57:33 I guess we can cross that bridge when we get to it though. 16:57:38 yeah. 16:57:39 3 minutes left.. how to proceed? 16:57:39 as long as someone can help validating 16:57:56 berendt, let's hold on with deprecation since gema is volunteering 16:58:19 inc0, +1 16:58:26 inc0: +1 16:58:31 yep +1 16:58:41 inc0: ok i will close the vote and send the decision to the list 16:58:48 and extend debian to pike, then gema can take on fixing it:) 16:58:54 if we support arm, it will be easy to support x86/power 16:58:54 +1! 16:58:58 we will need gates tho, I'm adamant about it 16:59:03 but we'll help 16:59:14 ok, we ran out of time 16:59:20 thank you all for coming! 16:59:26 thank you! 16:59:26 #endmeeting kolla