16:00:13 #startmeeting openstack_ansible_meeting 16:00:14 Meeting started Tue Mar 26 16:00:13 2019 UTC and is due to finish in 60 minutes. The chair is mnaser. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:16 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:18 The meeting name has been set to 'openstack_ansible_meeting' 16:00:31 few things to start us off 16:00:40 welcome noonedeadpunk to core, and late welcome to guilhermesp too :) 16:00:51 thank you! 16:01:22 hope I'm aggregating to the team 16:01:36 ptg document is here, please add topics: https://etherpad.openstack.org/p/osa-train-ptg 16:02:31 guilhermesp: jrosser logan- spotz please add topics if there is anything you find we should discuss 16:02:59 something that we discussed last week was about var clean ups across roles, would be something we aim to this cycle or? 16:03:19 \o/ 16:03:49 o/ 16:04:30 guilhermesp: you can add that as a discussion item 16:04:53 o/ 16:04:58 just really leaving it up for anyone to add things if they need 16:05:09 we need to discuss release things 16:05:25 I think we need to land python_venv_build stuff this week 16:05:29 to avoid a huge amount of backports 16:06:02 most of the odyssey4me 's patches are merged I think. Now the focus would be adapt other roles to use pythn_venv_build? 16:07:00 I think the pattern to adapt it would be those two: https://review.openstack.org/#/c/637240/ and https://review.openstack.org/#/c/620340/ 16:08:16 yes agreed.. I think it wont be a massive work to do that 16:09:33 o/ 16:10:17 odyssey4me: is it accurate that is what has to be done? 16:10:58 Yep, pretty much - naturally a single patch can implement https://review.openstack.org/#/c/637240/ for everything in the integrated repo, which will save having to do it multiple times. 16:11:18 yeah that's what I thought we could do as well 16:11:32 then it's down to implementing it in the multiple roles - and once those are done we can clean up the integrated repo of the legacy things 16:11:42 o/ late as always 16:12:23 perhaps we should etherpad the work to be done so that we can volunteer to do what needs doing? 16:12:36 o/ 16:12:55 I think it is reasonable odyssey4me 16:13:16 yeah. that way we can maybe put links on the review that does the change 16:13:20 ok, I'll put an etherpad together 16:13:24 and for the integrated repo change it can be one change only 16:13:38 I want to add collect-logs from tripleo in OSA to aggregrate logs at one places from different node during train cycle if feasible? 16:14:11 chandankumar what does collect-logs do? 16:15:01 if we can setup like a deadline to replicate python_venv_build stuff I can try to volunteer myself to do it 16:15:14 I think we need to get it done before Friday, I can help out with some of the roles 16:15:33 I may take several as well 16:15:36 cool, then the etherpad would be handy 16:15:44 odyssey4me: https://github.com/openstack/tripleo-quickstart-extras/tree/master/roles/collect-logs -> it collect logs from different nodes at one place 16:16:05 chandankumar that sounds useful for CI, but not really for production? 16:16:17 isn't it done with syslog? or graylog? 16:16:30 and also feeds the info to the openstack logstash 16:16:48 ok, sounds nice for CI then - but not really for production 16:16:53 it also provides a way to generate docs from ansible roles from shell script template 16:17:26 yes for CI, it is very much useful 16:17:35 for production case, we have not looked into it 16:18:38 perhaps it'd be useful to have it as a zuul role then, and our CI and devstack CI could just use it in zuul? 16:19:02 here is the current used https://git.openstack.org/cgit/openstack-infra/tripleo-ci/tree/toci-quickstart/config/collect-logs.yml 16:19:22 we are creating it as an independent role 16:19:28 I have started the seperation today 16:20:10 ok, it might be useful as a zuul role in openstack ci which any project can use and provide vars for 16:20:53 * chandankumar will pass the message to the tripleo team 16:21:32 tomorrow 16:22:12 can I add this info in ptg planning? weshay_pto will be there during PTG 16:22:17 sure 16:22:48 thanks@ 16:22:52 now on the other hand 16:23:01 I think we need to branch off by saturday 16:23:07 even if we're unable to land all the work 16:23:19 its not ideal.. but yeah 16:23:21 thoughts? 16:23:27 Merged openstack/openstack-ansible-os_tempest master: Use tempest_keystone_interface_insecure var for ssl verification https://review.openstack.org/644816 16:24:26 i'm ok with that since we minimize the # of backports of the remaining roles 16:24:27 mnaser perhaps we should divide the work into two parts - essential to get done, and non-essential? 16:24:47 odyssey4me: I think so, I think right now the biggest reason that I feel stops us is to avoid excessive amount of backports 16:25:04 so, for example - nova/neutron/keystone/cinder/glance will have to be done ... what other services are commonly used? 16:25:21 odyssey4me: so perhaps what you mention makes sense. I think anything inside setup-openstack.yaml should be done first 16:25:24 horizon? 16:25:26 odyssey4me: mnaser can we make os_tempest role branchless from this release? 16:25:43 as tempest is branchless and it supports previous 3 releases 16:25:57 chandankumar I don't think so - we rely on the branch for catering to ansible funnies 16:26:18 yes ansible release have another fun stuff! 16:26:18 yeah. we can look into it _eventually_ but it's a bit too last minute to look into that 16:28:53 horizon already uses python_venv_build but needs some cleanup 16:29:48 it needs that final state similar to https://review.openstack.org/#/c/620340/ 16:30:16 yep 16:30:17 yeah I agree the part 1 would be port all core-services to use python_venv_build.. but the thing is we would reduce just a few roles to backport. I think we have a lot of non-core roles 16:30:36 #link https://etherpad.openstack.org/p/osa-python-build-simplification 16:33:37 perhaps we could start the work and start populating this etherpad with PR odyssey4me ? :) 16:33:48 ok that looks really good odyssey4me. 16:34:19 I think redoing the roles might be a days work 16:34:35 guilhermesp I'd suggest volunteering for some of the roles first to show that you're going to do it. 16:34:40 Put your name next to it. 16:34:59 ok sounds reasonable 16:38:23 I will pick some up today 16:38:41 also tempest is not really branchless according to our past :) 16:38:54 (sorry for the lag) 16:39:28 I think I'll be able to pick up some of those soon as well 16:39:49 I'll see how my time looks like and add more 16:41:09 so, it's time to grab roles from etherpad? 16:41:42 sure 16:41:49 pick at it! I think it would be good to grab like 3-4 at a time 16:41:56 that way if something gets you busy it doesn't block someone else 16:42:04 just saw no changes and didn't want to be the first) 16:42:25 noonedeadpunk: aha, yeah, ill mark myself up for a few I know 16:43:21 evrardjp: yes, correct, I know the pain I need to pin each release to a particular version in RDO otherwise it breaks lots of tests and stuff, adding extra pain to backport! 16:43:24 I added my name on a few and we'll try to work on those 16:43:40 yeah I will pick some today, 2 or 3 I think 16:45:05 great, thanks for the ether pad odyssey4me 16:45:55 odyssey4me: you have things you need to take care of, but if you have any spare time towards the removal/cleanup, I think that would probably where we'd find your help the most useful 16:46:13 _IF_ you have any spare time, if not, I plan to try and learn and figure it out as well 16:46:33 mnaser yep - I think if we can get the mainline AIO bits done, I'll start working on the clean up immediately 16:46:56 ok ill try to push on those as fast as I can 16:48:21 I've volunteered for nova/neutron - I should be able to work on those tomorrow, if not later this evening. 16:48:38 ok great, ill hack as much as I can, ill try to move things around 16:48:43 to push this through 16:50:16 anyone want to volunteer to get item 1 done - every other patch will depend on it 16:50:21 does item 1 on the pad need to be done per role, or is that one big patch to fix up the rest 16:50:24 ^ that 16:50:49 jrosser one big one makes more sense - no need to waste everyone's time on individual patches 16:51:08 I can do one big patch today 16:55:33 remeber we have a few roles that are totally broken becasue of systemd-networkd things 16:55:45 like master branch of swift and a couple of others 16:56:30 ergh 16:56:36 can we just move them to integrated tests? 16:57:03 i think thats the answer yes 16:57:51 Mohammed Naser proposed openstack/openstack-ansible master: Include package pins for OpenStack services https://review.openstack.org/647813 16:57:54 it might just be easier for us to do that 16:58:01 odyssey4me: ^ 16:58:21 if you see this openstack-ansible-functional-ubuntu-bionic finger://ze03.openstack.org/6c020bdae7f945e5856e815f87738be5 : RETRY_LIMIT in 11m 46s 16:58:41 then it's the connection to zuul been dropped at the point networkd gets reloaded 16:58:52 ugh, that's annoying we don't actually have the real output 16:59:15 well, if we're happy to do so, we can kill the standard role 'functional' tests and just switch them all to use an applicable aio test instead 16:59:31 that's what I think we should do as well 16:59:34 the infrastructure is all there 16:59:46 the problem is defining where we run the jobs, im thinking we should actually define all the jobs in openstack-ansible 16:59:52 ok, let me push some of those patches up tonight - I have some time to get on it 16:59:54 and then reference them all in the roles 17:00:01 that'll give everyone else room to do the other bits 17:00:10 because if we define the jobs in every role, it can be a bit hard to maintain 17:00:25 so I was thinking 'borrowing' the tripleo scenario list 17:00:32 mnaser well, that makes maintenance harder, actually - because to change a test we have to change it in two places 17:00:32 or the one from puppet-openstack-integration 17:00:58 right now it's fairly simple - the test definition is in the role in question 17:01:14 yeah, that's fair, im indifferent, its not like we change those much 17:01:23 ...they take a while to change if history is proof of anything 17:01:23 :P 17:01:48 just to close the office hour out, I think we got good discussions 17:01:54 having tried to change up the integrated repo definitions many times, it's a real pain due to constant merge conflicts because they're all in one file 17:01:54 we'll bother evrardjp when we've done our mess and he can help with releases 17:02:04 ya I remember how long it took to land that 17:02:13 odyssey4me: do you wanna rm tests/ as part of your patches too ? 17:02:30 if we're gonna drop the role tests might as well get rid of them I guess 17:02:31 mnaser not yet, because we still have to figure out how to change the linters 17:02:39 ah riht 17:02:41 right* 17:02:53 good thought 17:02:54 #endmeeting