16:00:18 <Jeffrey4l> #startmeeting kolla 16:00:19 <openstack> Meeting started Wed Mar 28 16:00:18 2018 UTC and is due to finish in 60 minutes. The chair is Jeffrey4l. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:20 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:22 <openstack> The meeting name has been set to 'kolla' 16:00:50 <Jeffrey4l> #topic Roll-call 16:00:58 <caoyuan> hi guys 16:01:05 <egonzalez> o/ 16:01:06 <bmace> o/ 16:01:07 <ktibi> o/ 16:01:11 <caoyuan> o/ 16:01:21 <pbourke> o/ 16:01:27 <duongq> o/ 16:01:29 <gkadam> o/ 16:01:34 <mgiles> o/ 16:01:49 <chason> o/ 16:02:06 <Jeffrey4l> #topic Announcements 16:02:08 <spsurya__> 0/ 16:02:28 <Jeffrey4l> one from me: 6.0.0 and 5.0.2 are release this week. 16:02:58 <Jeffrey4l> thanks for all the contributor :D 16:03:02 <egonzalez> Nice 16:03:22 <Jeffrey4l> any thing else from community? 16:03:31 <bmace> the kolla-cli repo exists now, and i think jeffrey set up the cores last night 16:03:40 <Jeffrey4l> btw, 4.0.x will be released later this week 16:03:47 <bmace> still a lot of work to get it in shape to remove some of the Oracleisms out of it, but starting work on it 16:04:00 <Jeffrey4l> yes. the permission is well configured 16:05:07 <Jeffrey4l> now kolla-cli repo has its own core team, which including bmace, and one of his cowork, and all kolla-core group member. 16:05:41 <Jeffrey4l> and thanks bmace and oracle to contribe this project to community. 16:05:47 <egonzalez> Mibd sharing a repo link? 16:05:48 <duongq> nice 16:05:50 <egonzalez> Mind 16:05:57 <bmace> thanks for taking it! :) 16:06:05 <Jeffrey4l> #link https://github.com/openstack/kolla-cli 16:06:12 <duongq> so we will have some interesting work here 16:06:20 <duongq> thank guys 16:07:17 <Jeffrey4l> ok let us move on 16:07:26 <Jeffrey4l> #topic retire kolla-kubernetes 16:07:56 <Jeffrey4l> i sent a mail in ML just now. 16:08:04 <Jeffrey4l> #link http://lists.openstack.org/pipermail/openstack-dev/2018-March/128822.html 16:08:07 <gkadam> so the kolla-kubernetes will have seperate irc channel? 16:08:36 <Jeffrey4l> gkadam, no. we are saying sunset this project. 16:08:48 <gkadam> Jeffrey4l, oh 16:09:03 <Jeffrey4l> if you are still interested in this area, you should check openstack-helm project. 16:09:14 <gkadam> Jeffrey4l, sure 16:09:16 <Jeffrey4l> this is not happen, just a propose 16:09:54 <Jeffrey4l> so feel free to reply on ML if you have any concern. 16:09:57 <Jeffrey4l> so feel free to reply on ML if you have any concern. 16:10:24 <gkadam> Jeffrey4l, ack 16:10:24 <duongq> so sad 16:10:49 <Jeffrey4l> i checked these two projects recently. tbh, openstack-helm is more better. it also implemented the micro charts. 16:11:31 <ktibi> Jeffrey4l, helm can use kolla image or they have their own images ? 16:11:50 <Jeffrey4l> duongq, mee too. but there is no reason to keep two same projects in openstack. 16:11:51 <egonzalez> ktibi, they use kolla and loci images 16:12:00 <Jeffrey4l> egonzalez, yes. 16:12:24 <duongq> Jeffrey4l, kk OSH has much better sponsor so the contributors etc. 16:12:26 <spsurya__> egonzalez: +1 16:13:28 <ktibi> Jeffrey4l, ok indeed we have no reason to keep two same project, so between kolla and loci, what's the next step ? because for me, it's the same project no ? 16:14:05 <pbourke> ktibi: first step is to make kolla-ansible support loci images 16:14:07 <spsurya__> sup kfox1111 16:14:11 <kfox1111> hi 16:14:24 <Jeffrey4l> ktibi, for kolla-k8s and osh, they are both using helm for deploying openstack on k8s. 16:14:35 <pbourke> ktibi: that effort is for review at https://review.openstack.org/550958 16:14:41 <Jeffrey4l> for loci and kolla is different. 16:14:46 <kfox1111> yes, but with one very key difference. 16:15:00 <pbourke> in the long term kolla images could be removed but I dont see that any time soon 16:15:06 <Jeffrey4l> we are trying to suport loci. by pbourke's spec 16:15:16 <kfox1111> openstack-helm believes helm has enough orchestration suport in it to do it only with helm. 16:15:28 <kfox1111> kolla-kubernetes has positioned itself to use orchestration other then helm. 16:16:19 <Jeffrey4l> pbourke, yep. kolla images will exist for a long time as far as I know. 16:17:01 <Jeffrey4l> kfox1111, but both these project are using helm to handle this, right? 16:17:33 <Jeffrey4l> and no active contributor is another critial reason. 16:17:37 <kfox1111> Jeffrey4l: openstack-helm believes 'helm upgrade <chart>' will be good enough to be able to handle openstack's upgrade workflow. 16:17:51 <kfox1111> I have grave doubts that will ever work long term. 16:18:16 <kfox1111> I currently believe we need some helm kubernetes operators. 16:18:23 <kfox1111> helm driving, kubernetes operators. 16:18:33 <Jeffrey4l> kfox1111, they can control the charts by some variables. which could implement that "helm upgrade glance" just run db sync 16:18:39 <kfox1111> kind of like how rook behaves. 16:19:23 <kfox1111> I've upgraded openstack enough times to know that it will never be that simple. 16:20:09 <kfox1111> schema upgrades fail. you have to iterate over it a few times, sometimes applying patches half way, then retrying. 16:20:42 <kfox1111> openstack never upgrade tests over multiple versions worth of data in the database. odd things pop up all the time. :( 16:21:04 <Jeffrey4l> kolla-ansible run db sync only once ;( 16:21:05 <kfox1111> so, again, I think openstack-helm's focus is on day 1. 16:21:19 <kfox1111> kolla-kubernetes was written by focusing on day 2. 16:22:05 <kfox1111> Jeffrey4l: and, no affence intended, but I had relucatnce in deploying kolla-ansible too when I started working on kolla-kubernetes instad. 16:23:21 <kfox1111> as an operator, I need day2 stability. day one only happens once. 16:23:31 <kfox1111> if day one is hard, but day 2 is reasonable, then its not so bad to live with IMO. 16:23:33 <Jeffrey4l> for i can see is "variable controlled chart" is also the same with "multi small charts" 16:23:40 <ktibi> kfox1111, +1 16:24:13 <Jeffrey4l> agree with you on this point. 16:24:15 <kfox1111> Jeffrey4l: aggregating a few charts into one bigger chart with enables is not the root of the argument. 16:25:02 <kfox1111> its how things are layed out, and intended to be orchestrated is the core issue. 16:25:24 <kfox1111> the way things are now, orchestration is a completely seperate piece of the archetecture. 16:25:53 <kfox1111> you could install the microservice charts with ansible, with helm itself (service charts), with an operator, or something else. 16:26:36 <kfox1111> we have it orchestrated with bash in the gate, with a python script that rwellum wrote, and an ansible playbook others have written. 16:27:12 <kfox1111> kolla-kubernetes needs to focus on a final reference orchestrator to cover day1 and make it super easy to play with. 16:27:25 <pbourke> kfox1111: not disagreeing with any of this but... fact of the matter is noone is really working on this 16:27:29 <kfox1111> if kolla-kubernetes is to be successful. 16:27:39 <kfox1111> that is the problem yeah. 16:27:46 <pbourke> so its pointless to keep dragging it out imo 16:28:38 <kfox1111> yup. most folks are moving on from openstack. :/ 16:29:43 <Jeffrey4l> so contributor is another importance issue. There is only fewer patches in kk8s in recent 3 month. 16:30:09 <kfox1111> yeah. 16:30:45 <kfox1111> unfortunately I don't have much time to spend on it right now. I may in a couple months, but I don't know for sure. 16:32:26 <kfox1111> so, I guess the question is, do we just give up now, or do we wait a bit longer to see if anyone else is interested? 16:32:26 <Jeffrey4l> anyway, thaks kfox1111 16:32:53 <Jeffrey4l> kfox1111, we can wait for a bit longer. 16:33:23 <spsurya__> Jeffrey4l: kfox1111 +1 16:34:38 <Jeffrey4l> let us move on. if you have any concern about this, please reply in mail. 16:34:47 <Jeffrey4l> #topic Open Discussion 16:35:06 <Jeffrey4l> any volunteer? 16:35:14 <duongq> o/ 16:35:35 <Jeffrey4l> please duongq 16:35:35 <duongq> just short and simple announcement 16:35:44 <duongq> the bp 16:35:44 <duongq> https://blueprints.launchpad.net/kolla-ansible/+spec/ansible-specific-task-become 16:35:49 <duongq> is done 16:36:00 <duongq> please help me recheck if I need to do something more 16:36:05 <duongq> before we closed this 16:36:16 <duongq> after many cycles (it's my fault) 16:36:17 <ktibi> I have one question 16:36:31 <ktibi> As I had already spoken, I work on security aspects. I realized that the version of Docker was really obsolete on centos. 16:36:40 <Jeffrey4l> duongq, ok 16:37:06 <duongq> thank Jeffrey4l 16:37:19 <ktibi> I know redhat support just 1.12.6 but that version is very old and as you know, docker have a lot of security breach .... 16:38:08 <Jeffrey4l> ktibi, which version are you talking? 16:38:16 <ktibi> and the second pb is docker change the name of package docker-engine => docker-ce, so very hard to upgrade ^^ 16:38:34 <Jeffrey4l> techinally, kolla support all docker version after 1.12.0, including docker-ce 16:38:49 <Jeffrey4l> i use docker-ce 17.09, which is more stable. 16:39:13 <egonzalez> Yep, is working on all docker-ce versions, in gates we use latest iirc 16:39:19 <ktibi> Jeffrey4l, it's kolla which install docker ? 16:39:43 <ktibi> because the repo used by kolla is not update since 1 year 16:39:56 <Jeffrey4l> egonzalez, should be. but iirc, only one of kolla-ansible / kolla is using the latest. 16:40:48 <Jeffrey4l> ktibi, if there is any place force to use order docker, we should fix it. kolla works with the latest docker-ce 16:41:22 <ktibi> Jeffrey4l, ok because for now, kolla use 1.12.0 for centos 16:41:47 <ktibi> #link https://github.com/openstack/kolla-ansible/blob/master/ansible/roles/baremetal/defaults/main.yml#L34 16:42:35 <Jeffrey4l> yeah, we should fix this or make it configurable. 16:43:21 <ktibi> ok I can make review to be configurable :) 16:43:40 <Jeffrey4l> cool, thanks 16:44:12 <Jeffrey4l> any other topics? 16:44:14 <ktibi> the issue is with upgrade 16:44:27 <ktibi> docker-engine => docker-ce (need to remove old package) 16:44:54 <Jeffrey4l> ktibi, no. i tested upgrade from docker-engine 1.12.6 to docker-ce 17.9, which works perfect. 16:46:09 <ktibi> Jeffrey4l, ok great, when I tested, I had conflict with packages maybe wrong command :/ 16:46:38 <egonzalez> Changes the docker daemon name in the systemd, but package replacement is fine 16:47:01 <Jeffrey4l> egonzalez, +1 16:48:31 <Jeffrey4l> ok, let us end the meeting earlier. thanks guys for comming 16:48:36 <egonzalez> Another things which makes all containers restart is a pid default value 16:48:37 <Jeffrey4l> #endmeeting