15:59:20 <inc0> #startmeeting kolla
15:59:21 <openstack> Meeting started Wed Apr 26 15:59:20 2017 UTC and is due to finish in 60 minutes.  The chair is inc0. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:59:22 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:59:25 <openstack> The meeting name has been set to 'kolla'
15:59:26 <hrw> o/
15:59:27 <spsurya_> 0/
15:59:32 <rwellum> o/
15:59:37 <inc0> #topic w00t for Kolla!
15:59:38 <duonghq> o/ wo0t
15:59:39 <egonzalez> w00t o/
15:59:42 <berendt> o/ woot
15:59:42 <sbezverk> o/
15:59:50 <inc0> loudly please
15:59:55 <inc0> WOOT!
16:00:02 <duonghq> hrw, spsurya_ rwellum are very eager members :P
16:00:02 <spsurya_> woot!
16:00:06 <jascott1> woot
16:00:27 <Jeffrey4l> woot
16:00:28 <kfox1111> o/
16:00:30 * hrw is used to o/ form
16:00:41 <blallau> o/
16:01:28 <inc0> let's give few more minutes
16:01:28 <rwellum> ^o^
16:01:43 <duonghq> today we have time :P
16:01:48 <inc0> yeah
16:01:57 <inc0> also attendance is somewhat small
16:02:00 <inc0> wonder why
16:02:05 <inc0> this whole week is quiet
16:02:20 <hrw> quiet? I got many patches merged ;D
16:02:32 <duonghq> so, next week may be even more quite
16:02:33 <inc0> irc-wise
16:02:34 <kfox1111> some folks preping for the summit. some other busy things going on.
16:02:53 <inc0> yeah, it's this pre-summit time, we might cancel next 2 meetings
16:02:58 <spsurya_> inc0: may be people are little busy for summit
16:03:21 <hrw> inc0: next one is summit time so +1 for cancel
16:03:29 <inc0> ok let's move on
16:03:36 <inc0> #topic Announcements
16:03:44 <inc0> nothign from me, anything from community?
16:04:09 <duonghq> who will go to Boston :P
16:04:11 <inc0> well, one thing, again will need mentors on workshops in summit
16:04:28 <Serlex> I will be there
16:04:31 <hrw> duonghq: I won't
16:04:46 <rwellum> Going too.
16:04:47 <hrw> May is quite busy with family things
16:05:22 <inc0> ok, moving on
16:05:28 <egonzalez> inc0 i'm not going to boston, but if need help remotelly for anything let me know
16:05:40 <inc0> #topic Mixing binary and source image (daidv)
16:06:06 <inc0> egonzalez: wouldn't od that to you, it'll be pretty late in Europe
16:06:17 <inc0> anyway duonghq I guess you're taking this one?
16:06:22 <duonghq> thank inc0, I cannot contact daidv, so I'll take over the topic,
16:06:43 <blallau> I can't be there
16:06:44 <duonghq> currently, we only allow users deploy whole system using binary or source images
16:06:55 <inc0> duonghq: not really
16:07:05 <inc0> you can always override images per service
16:07:12 <inc0> nova_compute_image_full
16:07:19 <inc0> less convenient, but doable
16:07:52 <duonghq> ah, yeah, so I just discover it last week, but forgot that I found it :P, quite cumbersome but ok
16:08:18 <inc0> I agree, but on the other hand it's not exactly common use case
16:08:27 <duonghq> it's better if we can give user option to make all service (e.g. nova) use one install_type
16:08:35 <Jeffrey4l> there are some sevice not. for example horizon. the configuration depends on install_type,
16:08:58 <Jeffrey4l> this can be treated as a bug.
16:09:01 <kfox1111> kolla-kubernetes can override from binary <-> source for any image or any service.
16:09:01 <inc0> good point Jeffrey4l
16:09:19 <duonghq> kfox1111, kolla-k8s has many cool features over ansible :P
16:09:19 <inc0> well it's very easy patchset too
16:09:35 <hrw> for some distros there are no binary images while sources ones exist
16:09:44 <Jeffrey4l> kolla-ansible is using meta/requirements feature. which stop change install_type.
16:09:45 <inc0> nova_install_type = {{ install_type }} in defaults
16:10:16 <Jeffrey4l> but, anyway, this is not a big issue. There will be a solution. like inc0 ^
16:10:31 <duonghq> sure, just enhancement
16:10:36 <Jeffrey4l> another solution is: remove the install_type in configuration files.
16:10:50 <kfox1111> just saying, if you wanted a place to test src/bin hybrids, it would be very easy to test there.
16:11:04 <inc0> Jeffrey4l: I'd avoid change config files
16:11:21 <inc0> because of backwards compatibility
16:12:02 <sbezverk> kfox1111: care should be taken when mixing source and binary. I hit some unpleasant issues when config was generqated for binary but image was used for source ..
16:12:04 <Jeffrey4l> end-user should use the same kolla and kolla-ansible, then it do not have backwards issue.
16:12:23 <Jeffrey4l> the root cause is: https://github.com/openstack/kolla-ansible/blob/master/ansible/roles/horizon/templates/horizon.conf.j2#L1
16:12:53 <inc0> anyway, there are ways
16:13:11 <Jeffrey4l> we can use some magic here. for example, if /var/lib/kolla/venv/lib/python2.7 exist, use it. otherise, use /usr/share/openstack-dashboard.
16:13:11 <inc0> I'd guess (besides horizon maybe) it's just matter of switching image_full
16:13:19 <kfox1111> sbezverk: interesting.
16:13:28 <Jeffrey4l> yep. i guess so.
16:13:29 <kfox1111> there's stuff hardcoded to care one way or the other?
16:13:33 <inc0> Jeffrey4l: I'd rather play with defaults
16:13:51 <inc0> duonghq: anything else on this one?
16:13:52 <sbezverk> kfox1111: all wsgi containers will have issues is config type and image type is not consistent
16:13:55 <kfox1111> would be nice to fix those sorts of issues in the image rather then the config if possible.
16:13:58 <duonghq> inc0, no atm
16:14:14 <kfox1111> sbezverk: do they point to different places in the container?
16:14:14 <inc0> #topic Tarball URLs (daidv)
16:14:16 <duonghq> I think other issues can be fix one-by-one
16:14:25 <sbezverk> kfox1111: yep
16:15:16 <duonghq> sbezverk, kfox1111 may I move to the current topic?
16:15:23 <Jeffrey4l> i think this topic is talked.  daidv asked me that why we use tag tarball rather than branch tarball for kolla stable branch.
16:15:23 <kfox1111> sbezverk: I wonder if we can just get some stretegically placed symlinks into the images.
16:15:29 <kfox1111> duonghq: sure.
16:15:53 <duonghq> Jeffrey4l, hmm, so how about the conclusion?
16:16:02 <Jeffrey4l> no.
16:16:21 <hrw> Jeffrey4l: tag tarball is always the same. branch can get new commits so new tarball so no reproducible builds?
16:16:22 <Jeffrey4l> nova stable/ocata is still develop branch.
16:16:29 <Jeffrey4l> it is unstable.
16:16:51 <inc0> well it's stable but less tested than tag
16:17:05 <inc0> but yeah, stable will still gets backported bugfixes
16:17:25 <inc0> which might or might not be required, depends on your comfort zone
16:17:26 <Jeffrey4l> unless we implement kolla tag with change all stable-tarball into tag tarball.
16:17:30 <hrw> then stable can do tag with changelog and we can move to new tag
16:17:36 <inc0> security updates are best served fresh
16:18:13 <Jeffrey4l> now, we update kolla stable tag regularly.
16:18:16 <inc0> on that note, it would be great to have kolla tag to deploy stable tags of services
16:18:24 <Jeffrey4l> hope we can do this automitically.
16:18:42 <inc0> while kolla stable/ocata branch point to stable/ocata branch in services
16:19:14 <inc0> not sure how to do it with openstack release
16:19:16 <inc0> process
16:19:39 <inc0> besides pushing patch with new versions, tagging, reverting patch to tip of branch
16:20:59 <inc0> ideas about that?
16:21:32 <Jeffrey4l> update openstack service tag automitically? or?
16:21:52 <hrw> Jeffrey4l: add a hook in openstack release which would ping kolla bot to bump?
16:21:53 <inc0> tag automation shouldn't be hard python script
16:22:09 <inc0> but having tip of the stable branch on our stable branch
16:22:10 <Jeffrey4l> hrw, should be possible.
16:22:12 <inc0> all the time
16:22:58 <Jeffrey4l> inc0, this is more hard. no idea on this .
16:23:08 <inc0> yeah
16:23:15 <inc0> well, something to think about tho
16:23:33 <hrw> Jeffrey4l: Fedora infrastructure uses fedmsg to announce whatever happens so each project can listen and react to data in own way
16:23:42 <Jeffrey4l> there should no pre-release patch concept, i think,which is pushed before real release patch.
16:24:08 <inc0> maybe something for Q tho
16:24:23 <inc0> we can start designing this
16:24:39 <inc0> I hope to have lots of stuff to design after summit
16:25:15 <inc0> ok, anything else?
16:25:30 <Jeffrey4l> hrw, i think there is a job,which run after each project is released. if so, we can inject some script to push a patch.
16:25:40 <hrw> Jeffrey4l: cool
16:25:54 <Jeffrey4l> just thoughts. not sure ;)
16:26:01 <Jeffrey4l> please move on .
16:26:16 <inc0> or just overhaul genconfig
16:26:32 <inc0> to do something like genconfig --tag or genconfig --stable-branch
16:27:07 <duonghq> inc0, +1 quick and simple idea
16:27:24 <Jeffrey4l> genconfig just generate a kolla-build.conf.sample file. all the real value is fixed in code.
16:27:40 <Jeffrey4l> sample file is full of commented.
16:27:42 <inc0> well you can specify it in kolla-build.conf
16:27:48 <inc0> ah
16:27:51 <inc0> true
16:27:52 <inc0> well
16:28:06 <inc0> but you get my idea - some script to generate either one or another
16:28:19 <Jeffrey4l> maybe ./build.py --openstack tag/branch
16:28:38 <Jeffrey4l> maybe ./build.py --openstack-release [tag/branch]      lol
16:28:39 <inc0> maybe that
16:29:10 <inc0> well, I think we need to have this discussion
16:29:26 <inc0> with broader community
16:29:51 <inc0> I'll buy a beer for anyone who will prototype this mechanism:)
16:30:38 <inc0> ok guys, this meeting is not very busy;) let's move on
16:30:42 <hrw> ;D
16:30:43 <inc0> and wrap up earlier
16:30:55 <hrw> I have to quit in 15 minutes
16:30:57 <inc0> to have some life back of our days;)
16:31:06 <inc0> #topic non-x86 q&a (hrw)
16:31:24 <hrw> ok, anyone has questions about kolla on non-x86 platforms?
16:31:35 <hrw> non-x86 == aarch64 (arm64) and ppc64le
16:31:43 <inc0> why just not use Intel? :P
16:32:01 <hrw> inc0: it is easier for me to get aarch64 machines than intel ones
16:32:11 <hrw> inc0: ppc64le I do as a favour ;D
16:32:16 <inc0> lol
16:32:22 <duonghq> because cointainer should be floating on the ocean (j/k)
16:32:37 <hrw> next question
16:33:00 <duonghq> do you have any demo at Boston?
16:33:29 <hrw> duonghq: I am not there. ask Gema will she show something
16:33:53 <duonghq> ah, hoped that I can meet you
16:34:03 <hrw> May is busy time for me family wise
16:34:14 <duonghq> got it
16:34:32 <hrw> hope to go to PTG but it would mean 3 weeks of jetlag or money loss to stay for a week in USA
16:34:55 <duonghq> aww, jet lag :(
16:34:58 <inc0> hrw: usually week is enough to heal up jetlag
16:35:08 <hrw> anyway catch Gema during summit - she knows out plan for 3rdparty CI etc
16:35:17 <inc0> cool
16:35:26 <hrw> inc0: europe -> USA for PTG -> Europe -> USA for Linaro Connect -> europe
16:35:30 <duonghq> I get a week to recovery from VN->US jet lag, and another for US->VN
16:35:57 <hrw> inc0: jetlag in us, then in europe, then in us, then in europe - even 4 weekd :(
16:36:17 <inc0> cool:)
16:36:25 <inc0> ok, can we wrap up?
16:36:30 <duonghq> mins
16:36:32 <duonghq> egonzalez, ping
16:36:33 <hrw> no other questions then yes
16:36:41 <inc0> #topic open discussion
16:37:00 <duonghq> o/
16:37:11 <egonzalez> duonghq sup
16:37:37 <duonghq> egonzalez, I think we can raise our ps about ks 0-downtime upgrade here? https://review.openstack.org/#/c/425446/
16:38:05 <vhosakot> o/ (sorry I'm late as I was at an appointment)
16:38:15 <duonghq> vhosakot, hi
16:38:19 <vhosakot> duonghq:
16:38:53 <egonzalez> I was working today to fit keystone 0-downtime with serial change some weeks ago
16:39:22 <spsurya_> duonghq: egonzalez as per our last discussion it will land in pike
16:39:30 <vhosakot> inc0: sdake: all, I'm giving a presentation about kolla and kolla-kubernetes at Red Hat summit next week.  would like to discuss about it for 2 mins at the end :)
16:39:30 <egonzalez> the issue is that current serial implementation does not allow overrides in playbooks, needed for zero-dowtime upgrade of neutron and keystone
16:39:31 <spsurya_> not in master IIRC
16:39:47 * hrw off
16:39:58 <egonzalez> spsurya_ master will be pike ;)
16:40:13 <duonghq> egonzalez, iirc, can we take idea from ceph?
16:40:53 <spsurya_> egonzalez: yes
16:42:20 <inc0> vhosakot: can we do it in #kolla channel?
16:42:34 <vhosakot> inc0: cool, that is good :)
16:42:50 <inc0> soo... duonghq anything else?
16:42:59 <duonghq> inc0, no,
16:43:01 <duonghq> thank youn
16:43:03 <duonghq> you
16:43:10 <inc0> ok, thank you all for coming:)
16:43:25 <inc0> hope to see at least some of you in Boston
16:43:34 <inc0> #endmeeting kolla