15:01:28 <mgoddard> #startmeeting kolla 15:01:29 <openstack> Meeting started Wed Nov 27 15:01:28 2019 UTC and is due to finish in 60 minutes. The chair is mgoddard. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:30 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:32 <openstack> The meeting name has been set to 'kolla' 15:01:33 <mgoddard> #topic rollcall 15:01:36 <mgoddard> \o 15:01:41 <yoctozepto> o/ 15:01:45 <hrw> o/ 15:01:49 <scottsol> o/ 15:04:19 <mgoddard> #topic agenda 15:04:27 <mgoddard> * Roll-call 15:04:29 <mgoddard> * Announcements 15:04:31 <mgoddard> ** Ussuri priorities now listed on whiteboard https://etherpad.openstack.org/p/KollaWhiteBoard 15:04:33 <mgoddard> ** PTG summary on openstack-discuss: http://lists.openstack.org/pipermail/openstack-discuss/2019-November/011084.html 15:04:35 <mgoddard> ** Infra images discussion on openstack-discuss: http://lists.openstack.org/pipermail/openstack-discuss/2019-November/011120.html 15:04:37 <mgoddard> ** CentOS 8 host migration discussion on openstack-discuss: http://lists.openstack.org/pipermail/openstack-discuss/2019-November/011130.html 15:04:39 <mgoddard> * Review action items from last meeting 15:04:41 <mgoddard> * CI status 15:04:43 <mgoddard> * Train release planning 15:04:45 <mgoddard> * Ussuri release planning 15:04:47 <mgoddard> * Review priorities 15:04:49 <mgoddard> ** Ceph orchestrator 15:04:51 <mgoddard> #topic announcements 15:04:53 <mgoddard> #info Ussuri priorities now listed on whiteboard 15:04:55 <mgoddard> #link https://etherpad.openstack.org/p/KollaWhiteBoard 15:05:05 <mgoddard> #info PTG summary on openstack-discuss 15:05:11 <mgoddard> #link http://lists.openstack.org/pipermail/openstack-discuss/2019-November/011084.html 15:05:17 <mgoddard> #info Infra images discussion on openstack-discuss 15:05:22 <mgoddard> #link http://lists.openstack.org/pipermail/openstack-discuss/2019-November/011120.html 15:05:29 <mgoddard> #info CentOS 8 host migration discussion on openstack-discuss 15:05:35 <mgoddard> #link http://lists.openstack.org/pipermail/openstack-discuss/2019-November/011130.html 15:06:02 <mgoddard> Please join in these discussions as we try to firm up some of the more squishy features 15:06:07 <mgoddard> Any others? 15:06:49 <mgoddard> #topic Review action items from last meeting 15:07:05 <mgoddard> mgoddard to tidy up ussuri priorities and add to whiteboard 15:07:07 <mgoddard> check 15:07:13 <mgoddard> #topic CI status 15:07:24 <mgoddard> "kolla-toolbox fails to build on Ubuntu due to libssl older than libssl-dev" 15:07:34 <mgoddard> that one is 'fixed' now right? 15:07:46 <hrw> CI passed through it 15:08:42 <mgoddard> ok, reverted whiteboard to green 15:08:50 <mgoddard> #topic Train release planning 15:09:00 <mgoddard> I think we're getting close 15:09:17 <mgoddard> https://launchpad.net/kolla/+milestone/9.0.0 15:09:19 <yoctozepto> pending is the rmq issue 15:09:36 <yoctozepto> osmanlicilegi to work on that but not finished yet 15:09:37 <mgoddard> yeah 15:09:43 <mgoddard> is it release blocking? 15:09:52 <hrw> I would release with 3.7.10 15:09:55 <yoctozepto> other one is https://review.opendev.org/695804 15:09:56 <patchbot> patch 695804 - kolla-ansible - [Train] Fix Tacker config for the local Glance store - 3 patch sets 15:10:03 <yoctozepto> just waiting for you 15:10:16 <yoctozepto> mgoddard, hrw: if we don't mind mid-cycle upgrade 15:10:33 <yoctozepto> could work, not breaking completely obviously 15:10:37 <hrw> yoctozepto: at least it would be tested in ussuri. 15:10:38 <mgoddard> yoctozepto: I +2 tacker 15:10:54 <yoctozepto> mgoddard: I know, and send my thanks 15:10:56 <hrw> yoctozepto: now it is 'we have train built/tested, lets add fresh rmq and release' 15:11:22 <mgoddard> hrw: we're not talking about an rmq upgrade, "only" erlang 15:11:29 <hrw> yoctozepto: we completely lack !x86 support for rmq/erlang 15:11:29 <yoctozepto> hrw: I don't mind too much 15:11:38 <mgoddard> eitehr way it's still risky 15:11:47 <yoctozepto> mgoddard: yeah, only erlang and rmq is patch level 15:12:05 <yoctozepto> yeah, because it is erlang dribing most of networking there 15:12:09 <yoctozepto> driving* 15:14:22 <mgoddard> hmm, actually it is a rmq patch upgrade, missed that 15:14:43 <yoctozepto> mgoddard: no point in leaving bugs since they were blocked by erlang 15:14:50 <yoctozepto> though still it's erlang more worrying 15:15:01 <yoctozepto> the plus fact is that it's a configuration "tested" by rmq upstream 15:15:18 <yoctozepto> so we are possibly getting them to consider our bug reports as worthy ;p 15:16:10 <mgoddard> scottsol: I think this is your issue: https://bugs.launchpad.net/kolla/+bug/1814233 15:16:10 <openstack> Launchpad bug 1814233 in kolla "rabbitmq image fails to build" [Critical,Fix released] 15:16:46 <scottsol> yeh still stuck with that 15:17:07 <scottsol> erlang-hipe and the pinning of rmq wont work 15:17:20 <yoctozepto> well, not really stuck, it remains to patch tripleo and provide aarch64 repos to kolla 15:17:51 <mgoddard> how close are we to aarch64 repos? 15:18:15 <yoctozepto> I think hrw has finished them 15:18:19 <hrw> not quite 15:18:20 <yoctozepto> at least for some 15:18:23 <yoctozepto> ah ok 15:18:31 <yoctozepto> still, not in patch 15:18:36 <yoctozepto> osmanlicilegi not around hmm 15:19:15 <mgoddard> ok. Let's see how it progresses over the next few days 15:19:31 <openstackgerrit> Mark Goddard proposed openstack/kolla-ansible stable/train: Use valid_interfaces instead of os_interface for placement https://review.opendev.org/696323 15:19:46 <mgoddard> https://launchpad.net/kolla-ansible/+milestone/9.0.0 15:20:01 <mgoddard> the open bug list does seem to be going down 15:20:48 <yoctozepto> osmanlicilegi: ping me by tomorrow - otherwise I assume you are not able to make it so I can pick it up 15:20:58 <yoctozepto> thanks 15:21:17 <mgoddard> are we planning to get enforce_scope to work for train? 15:21:27 <mgoddard> seems like a big change 15:21:41 <yoctozepto> I think he was around during the day 15:22:09 <yoctozepto> mgoddard: yeah, I'm afraid it could break in funny ways 15:22:16 <yoctozepto> I would like to but scared 15:22:18 <mgoddard> I'm tempted to remove the 9.0.0 milestone for that one 15:22:32 <mgoddard> it's a big enough patch that it looks like a feature 15:22:47 <yoctozepto> it's kind of to support a feature in train 15:23:02 <yoctozepto> "fix kolla-ansible to enable that feature" thing 15:23:08 <yoctozepto> borderline 15:23:47 <mgoddard> should we try to get https://bugs.launchpad.net/bugs/1844349 in ? 15:23:47 <openstack> Launchpad bug 1844349 in kolla-ansible ussuri "Neighbour table overflow problem with Neutron" [Medium,In progress] - Assigned to Pierre Riteau (priteau) 15:24:29 <priteau> That would be great. I am OK with removing the conntrack bits if that helps 15:24:32 <mgoddard> probably not urgent but osmanlicilegi said it helped 15:25:01 <priteau> I think I only took them from the charm code, it wasn't actually required on the deployment from where this fix originated 15:25:05 <yoctozepto> mgoddard, priteau: I am fine with either way, I just missed that there was any response on this, sorry ;-/ 15:25:07 <mgoddard> the patch is W-1 currently priteau 15:25:25 <yoctozepto> taking unnecessary bits off is better as always 15:25:28 * priteau checking why it is like this 15:25:34 <mgoddard> W-1 Need to investigate if this is really relevant only for neutron-l3-agent. 15:25:37 <yoctozepto> if we all agree that less is more 15:25:40 <yoctozepto> (man less) 15:25:55 <priteau> Right, I wanted to check if it was needed only for neutron-l3-agent hosts 15:26:13 <yoctozepto> at least most relevant 15:26:22 <priteau> The problem is that the only environment where I can easily reproduce this is a production cloud :D 15:26:23 <yoctozepto> maybe get mnasiadka to ask neutron friends 15:26:32 <yoctozepto> priteau: :D 15:27:26 <hrw> priteau: perfect testing environment 15:27:40 <yoctozepto> closest to real world scenario 15:27:45 <hrw> priteau: at least when it fails lot of people tell you about it 15:27:59 <hrw> not that you will miss it etc 15:28:09 <yoctozepto> only pluses 15:28:47 <mgoddard> ok, if priteau can tidy up, we can try to get it into train 15:29:04 <mgoddard> I'll do another check of backports 15:29:05 <priteau> I am on it 15:29:08 <mgoddard> thanks 15:30:04 <mgoddard> I think rmq/erlang is the only thing holding the release at this point. Let's try to get an answer for that and get our GA out 15:30:28 <mgoddard> given the CentOS 8 migration we'll probably need some big changes to train down the line anyway 15:30:46 <hrw> once I sort out building erlang packages I will update patch with repos 15:30:53 <mgoddard> thanks gr 15:30:57 <mgoddard> hrw: 15:31:09 <mgoddard> let's move on to kayobe train release 15:31:39 <mgoddard> jovial has offered to do the release handle cranking 15:32:02 <mgoddard> we're hoping to be not too far behind kolla here 15:32:27 <mgoddard> we have a bunch of features listed on the whiteboard 15:32:49 <mgoddard> we don't really have time to do most of them though, so it's probably more of a list for Ussuri at this point 15:33:12 <openstackgerrit> Pierre Riteau proposed openstack/kolla-ansible master: [neutron] Adjust neighbour table thresholds https://review.opendev.org/682664 15:33:41 <mgoddard> we have a train release etherpad: https://etherpad.openstack.org/p/kayobe-train-release 15:34:06 <mgoddard> #action mgoddard to update kayobe train release etherpad 15:34:35 <mgoddard> anything to add priteau? 15:35:25 <priteau> There are many patches that we can still merge, but you're right, the priorities will have to switch to Ussuri 15:35:36 <yoctozepto> so kayobe is pretty much mgoddard, priteau, jovial and dougsz? 15:36:13 <mgoddard> mostly. we have contributions from others at times including scottsol & co 15:36:39 <yoctozepto> I see 15:36:41 <mgoddard> those are the main active cores 15:36:56 <mgoddard> room for more if anyone is interested :) 15:37:23 <mgoddard> #topic Ussuri release planning 15:37:41 <mgoddard> Feels like we've made a good start on some features now 15:37:53 <mgoddard> centos8/py3/drop py2 in particular 15:38:04 <hrw> mgoddard: thanks for sorting out that in k-a 15:38:11 <mgoddard> no problem 15:38:43 <mgoddard> any changes in the centos8 container patch? 15:38:52 <hrw> I plan to submit something like https://review.opendev.org/695796 for k-a. 15:38:52 <patchbot> patch 695796 - kolla - test-requirements: bump pyflakes to get f-strings ... (MERGED) - 1 patch set 15:39:03 <hrw> mgoddard: no extra content yet 15:39:12 <hrw> mgoddard: centos team busy with 8.1 builds now 15:39:43 <mgoddard> hrw: what is failing in centos/source currently? 15:39:56 <mgoddard> that should be smoother than binary, in theory 15:40:27 <hrw> mgoddard: no ceph, openvswitch etc. 15:40:50 <mgoddard> hmm, OVS is an issue 15:41:08 <hrw> mgoddard: in general: check CI result. it shows nicely what fails 15:41:29 <mgoddard> yes. easier to ask you though, and we all find out :) 15:41:58 <hrw> mgoddard: last update was to just rebase it and let CI check how it goes 15:42:04 <mgoddard> I'm just thinking, could we get a single node ansible job to pass? 15:42:22 <hrw> aio? 15:42:24 <mgoddard> possibly with some fudging 15:42:26 <mgoddard> yes 15:42:35 <hrw> no idea to be honest 15:43:30 <mgoddard> https://review.opendev.org/#/c/692368/ 15:43:30 <patchbot> patch 692368 - kolla - WIP: move to CentOS 8 in containers - 21 patch sets 15:43:36 <yoctozepto> someone mentioned testing in CI https://review.opendev.org/695192 15:43:36 <patchbot> patch 695192 - kolla-ansible - CI: Refactor a lot - 23 patch sets 15:43:45 <yoctozepto> testing, testing 15:43:48 <mgoddard> https://storage.bhs1.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_30c/692368/21/check/kolla-build-centos-source/30c7c33/kolla/build/ 15:44:03 <yoctozepto> I have the card "test swift" 15:44:09 <yoctozepto> because we have a bug pending on it 15:44:37 <hrw> yoctozepto: 'a lot' is very wrong commit message 15:44:53 <yoctozepto> hrwL very accurate though 15:45:14 <mgoddard> aio blockers: fluentd, glance-base, keystone-base, kolla-toolbox, neutron-base, nova-base, nova-libvirt, openvswitch-base, rabbitmq 15:45:44 <hrw> yoctozepto: can it get split into single small changes? 15:46:30 <hrw> yoctozepto: as now it looks like a way to sneak 7 separate changes in one commit 15:46:37 <yoctozepto> hrw: it is 15:46:43 <yoctozepto> it was meant to be testing qinling 15:46:52 <yoctozepto> but it was a pain so it got refactors in 15:46:58 <yoctozepto> now they are split 15:47:28 <hrw> yoctozepto: I mean to split that one 'split from qinling' patch 15:47:36 <yoctozepto> I know 15:48:08 <yoctozepto> extra work for not much gain, what can I say 15:48:43 <mgoddard> I would have preferred it to be split, but I have reviewed it now 15:49:03 <mgoddard> at least qinling is factored out now 15:50:20 <mgoddard> #topic Review priorities 15:50:42 <mgoddard> Same as usual really, train targeted bugs first, ussuri priorities second 15:50:59 <mgoddard> #topic Ceph orchestrator 15:51:04 <mgoddard> mnasiadka: you there? 15:51:17 <mnasiadka> yeah, just got to the keyboard 15:51:20 <mnasiadka> sorry for being late 15:51:28 <yoctozepto> (mgoddard, not seeing your review) 15:51:45 <mgoddard> (I reviewed before, not since your update) 15:51:47 <yoctozepto> mnasiadka: you had to rest from ml2 conf :-) 15:51:55 <yoctozepto> (I see) 15:51:58 <mgoddard> mnasiadka: Ceph orchestrator. Go! 15:52:06 <mnasiadka> Golang? ugh 15:52:20 <mnasiadka> So - cutting the story short - ceph-ansible's future is a bit not bright and shiny 15:52:28 <mnasiadka> What Ceph is going towards - is Ceph Orchestrator 15:52:50 <mnasiadka> so an abstraction layer between the operator/deployer and some deployment backends 15:53:06 <mnasiadka> in Octopus it will be the default deployment tool replacing ceph-deploy (and in future ceph-ansible) 15:53:32 <mgoddard> do you know which deployment backend will be recommendeD? 15:53:36 <hrw> march 2020, right? 15:53:39 <mnasiadka> So from my perspective - investing time in doing ceph-ansible integration with Kolla-Ansible is useless 15:53:55 <mnasiadka> mgoddard: the 'ssh' backend will be recommended - so it streams python scripts through ssh 15:54:09 <mnasiadka> hrw: march 2020 is the plan to release Octopus, yes 15:54:18 <mnasiadka> so it more or less fits into Ussuri timelines 15:54:21 <mgoddard> is the ssh backend different from ceph-deploy? 15:54:43 <mnasiadka> mgoddard: yeah, it doesn't require you to do all the pre-req jobs 15:54:51 <mnasiadka> the workflow is more or less: 15:54:52 <mgoddard> ok 15:55:00 <mnasiadka> 1) deploy 1 node ceph cluster (mon + mgr) 15:55:07 <mnasiadka> 2) enable ceph orchestrator module in mgr 15:55:14 <mnasiadka> 3) generate ssh keys and distribute them to nodes 15:55:24 <mnasiadka> 4) deploy more mons/mgrs for fully HA cluster 15:55:32 <mnasiadka> 5) scan for disks, deploy osds on them 15:55:43 <mnasiadka> (and add mds or rgw if you need them) 15:55:50 <mgoddard> 6) profit 15:56:05 <mnasiadka> yeah 15:56:17 <mnasiadka> so now the question is - what do we want to achieve from kolla-ansible perspective going forward? 15:56:32 <mnasiadka> I would say it would make sense to polish the external ceph 15:56:48 <mnasiadka> to remove current quirks and weirdness 15:57:29 <mgoddard> +1, and remain agnostic 15:57:36 <mnasiadka> I don't know if it makes any sense to integrate k-a with ceph orchestrator - it's should be up to the user to deploy it as he wants 15:57:42 <yoctozepto> +1 mgoddard 15:57:56 <mgoddard> we don't want to get hit by changing winds again 15:58:23 <mnasiadka> so from my perspective - maybe it's time to start moving towards removing kolla-ceph - or extracting it to separate project if there are maintainers interested in continuing 15:58:36 <mgoddard> we can always add support down the line if ceph orchestrator looks like it's here to stay, but it's hard to remove it 15:59:08 <mgoddard> mnasiadka: +1 for starting that conversation at least 15:59:09 <mnasiadka> mgoddard: I would probably prefer to write good docs instead of going that path... but there's still CI 15:59:23 <mnasiadka> I think we want to maintain a ceph-related CI, for all external ceph features 15:59:28 <mgoddard> yes 15:59:38 <mgoddard> I was going to ask what state your ceph-ansible patch is in 15:59:42 <mgoddard> could it be used for CI? 15:59:55 <mnasiadka> yeah, it could 16:00:04 <mnasiadka> the other option is to write a script leveraging the ceph orchestrator 16:00:25 <mnasiadka> or rework the ceph-ansible one - to be more frank 16:00:29 <mgoddard> we'd need to modify it to move the playbook to tests/ 16:00:29 <yoctozepto> mgoddard, mnasiadka: regarding splitting - it seems to correlate with the general extensibility idea 16:00:48 <mgoddard> it would be a good test case for extensibility 16:00:54 <yoctozepto> indeed 16:01:07 <mgoddard> although I see the ansible side as being a separate CLI + site.yml rather than hooking in 16:01:11 <mnasiadka> mgoddard: that's not a problem - but I think I would prefer to remove the playbook (which is just an import from ceph-ansible) and add ceph orchestrator based deployment to the bash script 16:01:26 <mgoddard> mnasiadka: sounds like moar work 16:01:42 <mgoddard> it's taken quite a while to get ceph-ansible working, why not use it? 16:02:14 <mgoddard> and ceph orchestrator is still in development by the sound of it 16:02:16 <openstackgerrit> Merged openstack/kolla-ansible stable/train: Fix kolla_ansible python module check https://review.opendev.org/696281 16:02:17 <openstackgerrit> Merged openstack/kolla-ansible master: Fix 'nova_cephx_raw_key' is undefined https://review.opendev.org/695933 16:02:30 <mnasiadka> mgoddard: well, let's assume we can use it for now, I still have some $job plan to investigate ceph orchestrator - so we can move at feasible time to use it instead of ceph-ansible 16:03:02 <mgoddard> doesn't ceph orchestrator require octopus 16:03:04 <mgoddard> ? 16:03:05 <mnasiadka> mgoddard: so 1) adapt current ceph-ansible change to use as CI only 2) look in future to change it to use ceph-orchestrator 16:03:20 <mgoddard> ok, makes sense 16:03:26 <mnasiadka> mgoddard: it does, but RDO will build Octopus either way in Ussuri :) 16:03:39 <mgoddard> eventually yes 16:03:48 <mgoddard> anyway, we're out of time 16:04:07 <mgoddard> mnasiadka: can you kick off the conversation with openstack-discuss about ceph removal? 16:04:16 <mnasiadka> yes, will send a mail tomorrow 16:04:30 <mnasiadka> asking for contributors for extracting kolla-ceph 16:04:31 <mgoddard> maybe try to find some past committers & known users e.g. wangwei 16:04:43 <mnasiadka> should we give ourselves some deadline for answers? 16:04:47 <mnasiadka> Like end of December? 16:04:54 <mgoddard> yeah 16:05:01 <mnasiadka> ok, sounds good 16:05:08 <mgoddard> let's try to be a bit active in seeking people out though. This is a big move 16:05:28 <mnasiadka> It is, but we have sort of announced it in the beginning of Train cycle 16:05:50 <mgoddard> well we announced a supported migration path to ceph-ansible. This is not that 16:06:02 <mnasiadka> yeah 16:06:17 <mnasiadka> I'll rework the BP to deliver a CI-only implementation of ceph-ansible integration 16:06:25 <mgoddard> ok 16:06:39 <mnasiadka> we could raise a BP around ceph orchestrator and taking over a kolla-ceph deployment 16:07:00 <mnasiadka> but I doubt I will have time to work on it in a reasonable velocity, given my $job priorities 16:07:25 <mnasiadka> I can ask in the same mail to ML if there are volunteers :) 16:07:40 <mgoddard> that would be good. 16:07:47 <mgoddard> let's call it a day 16:07:51 <mgoddard> thanks all 16:07:54 <mgoddard> #endmeeting