16:00:58 #startmeeting OpenStack Ansible Meeting 16:00:58 Meeting started Thu Aug 20 16:00:58 2015 UTC and is due to finish in 60 minutes. The chair is odyssey4me. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:59 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:02 The meeting name has been set to 'openstack_ansible_meeting' 16:01:09 o/ 16:01:23 #topic Agenda & rollcall 16:01:28 hey 16:01:32 o/ 16:01:42 o/ 16:01:46 o/ 16:01:59 o/ 16:02:16 o/ 16:02:35 hi 16:04:24 o/ 16:04:39 hello 16:05:06 Howdy all - nice to see that we have a fair set of attendeed. 16:05:09 *attendees 16:05:29 #topic Review action items from last week 16:05:49 * odyssey4me sigmavirus24 and cloudnull to put together a POC for pip and epoch's 16:05:54 * sigmavirus24 sighs 16:06:05 maybe after all this ugprade nonsense is done 16:06:11 hehe, now that cloudnull is back, perhaps that can go ahead - we carry again? 16:06:19 please 16:06:24 #action sigmavirus24 and cloudnull to put together a POC for pip and epoch's 16:06:26 cloudnull isn't exactly back 100% 16:06:35 * odyssey4me odyssey4me to add upstream liberty milestones to the trunk series 16:07:02 this has been done - it's perhaps useful to target trunk bugs against liberty-3 and liberty-release 16:07:28 I just ask that any liberty-related bugs also be added as a related bug to the liberty blueprint please 16:07:33 #link https://blueprints.launchpad.net/openstack-ansible/+spec/liberty 16:07:47 * odyssey4me all to review blueprints, self-assign and target, then submit and review blueprints 16:08:02 hopefully some of that has been done, we'll go into that in a bit 16:08:46 before we dive into blueprints, I'd like to shuffle the agenda a little to get the rest out the way 16:08:51 are we happy with that? 16:09:21 ok 16:09:36 #topic Broken Gate 16:10:12 I think most people in the meeting are aware that gating for master has been broken for almost 3 weeks now 16:10:25 the issue has been tempest failing when testing nova/cinder 16:10:33 it only happens in hpcloud 16:10:48 today we hope that we've found the solution in https://review.openstack.org/215093 16:11:01 note that juno/kilo gating has been fine all this time 16:11:38 if the above review doesn't work, then https://review.openstack.org/215028 is a workaround to get dev moving again and allow more time to figure out the issue 16:11:56 any comments/questions about that? 16:13:10 None from me 16:13:14 none 16:13:29 alright 16:13:34 #topic Stackforge -> Openstack 16:13:35 if this isn't the solution, I'll try to have a closer look on friday with you. we need to have it working 16:13:46 thanks evrardjp :) 16:13:57 Another update 16:14:24 We are on the list for upcoming project renames, which will rename the project from os-ansible-deployment to openstack-ansible, and move the repo to the openstack namespace: https://wiki.openstack.org/wiki/Meetings/InfraTeamMeeting#Upcoming_Project_Renames 16:14:29 #link https://wiki.openstack.org/wiki/Meetings/InfraTeamMeeting#Upcoming_Project_Renames 16:14:59 I read it wasn't that easy, what's the impact for us? 16:15:18 a process for handling renames has been sent out on the mailing list - please read and provide feedback 16:15:23 #link http://lists.openstack.org/pipermail/openstack-dev/2015-August/071816.html 16:15:25 ok thanks 16:15:28 #link http://lists.openstack.org/pipermail/openstack-dev/2015-August/072140.html 16:15:40 git remote set-url many times 16:15:51 evrardjp so the repo name will change, and the organisation name will change 16:15:58 what hughsaunders said 16:16:18 and probably many docs changes for us too 16:16:21 that's what I understand, and the remotes, that's what seems logical to me at first, but will it impact other processes/gating/documentation? 16:16:29 ok 16:17:01 yeah, the gate stuff will be handled in the rename - but we'll have some in-repo changes which will need to be sorted out too 16:17:18 we'll deal with that when it happens - it's a pretty easy search and replace for most of it 16:18:01 any more thoughts/comments/questions? 16:18:26 none 16:18:46 #topic Upcoming Milestones 16:19:09 odyssey4me: Birthdays? 16:19:14 \o/ 16:19:32 lol, my birthday was last week hughsaunders :p 16:19:42 who else is having one? 16:19:50 happy birthday then, sorry for being late on this one! 16:19:58 odyssey4me: everybody is having one at some point. 16:20:06 Happy Birthday! 16:20:16 deep andymccr 16:20:16 andymccr ever the pragmatist 16:20:35 sarcastic I'd say 16:20:53 adding that dose of realism that was required 16:21:08 o/ 16:21:12 so https://launchpad.net/openstack-ansible/+milestone/11.2.0 is set for 11 Sep, but we needed to release 11.1.2 early due to some changes upstream and some stuff I broke in the requirements 16:21:17 * cloudnull late was in a meeting. 16:21:35 It appears that the ceph work is largely done and that the work is ready for backporting 16:22:05 I suggest that we bring forward the 11.2.0 date to Friday next week (28 Aug), then continue from there 16:22:12 any objections to that? 16:22:18 FYI, I'm using it in kilo right now, not tested the cinder-backup, but everything looks fine 16:22:24 easy backporting then 16:22:41 no objection 16:22:56 yup, there are some extra tweaks in flight which we can merge and backport easily once master is unblocked 16:23:03 welcome cloudnull :) 16:23:31 +1 on bringing 11.2.0 forward 16:23:36 hughsaunders andymccr are you guys happy to bring the 11.2.0 milestone forward? 16:23:51 :p 16:25:01 sure 16:25:17 #action odyssey4me to move 11.2.0 milestone to 28 Aug 16:25:29 #topic Blueprints 16:25:44 #link https://blueprints.launchpad.net/openstack-ansible 16:26:00 we have a few blueprints of note 16:26:15 liberty is coming up 16:26:18 #link https://blueprints.launchpad.net/openstack-ansible/+spec/liberty 16:26:48 master's development from the end of august will start to diverge from kilo somewhat as we focus a little more on liberty changes 16:27:03 please see the dependency tree at the bottom of the page 16:27:23 palendae & cloudnull there are blueprints that liberty's release depend on there 16:27:59 these aren't hard deps, except perhaps for the upgrade path - but I'd like to confirm that we think we can get these done in time? 16:28:10 When's the liberty release date? 16:28:17 cloudnull I know that you had much of the work for yours done already 16:28:42 liberty-release is on 19 Sep: https://launchpad.net/openstack-ansible/+milestone/liberty-release 16:28:48 I would say yes, unless people have problems with the upgrade path proposed 16:28:56 Then we need to actually write kilo->liberty upgrades 16:29:27 #link https://review.openstack.org/207713 16:29:39 that's the spec for the upgrade path - it's gone through some reviews already 16:29:55 we need more attention on that now as it's going to hit crunch time soon 16:29:57 I'll likely need some help, given the timeline; but I have an idea for a starting point if people aren't opposed to me starting a parallel code path 16:30:10 parallel code patch* 16:30:31 palendae I think it's fairly safe to say that master is available to start ripping the upgrade stuff apart and doing the thing 16:30:52 I'd suggest trying to break it into parts so that the patches are easier to test and review. 16:30:53 odyssey4me: Cool, then I will try to start on that tonight or tomorrow, at least to get a commit on for people to lob comments at 16:30:55 Right 16:31:23 sigmavirus24 cloudnull hughsaunders andymccr please review https://review.openstack.org/207713 asap 16:31:35 we need to finalise and approve that before we get any patches merged 16:32:01 needs manual rebase 16:32:08 #action cores to review the liberty upgrade spec https://review.openstack.org/207713 16:32:14 ^^ 16:32:29 My first patch is basically, for master, going to be ripping out everything and leaving a blank slate so we can start filling stuff in that's relevant 16:32:43 Maybe an error message if Juno config files are found letting the deployer know they should go to kilo 16:32:54 hughsaunders I think we need to revise gating in time for liberty too - will you be able to put https://blueprints.launchpad.net/openstack-ansible/+spec/split-aio-gates together before the next meeting? 16:33:01 That's where I'll start, and probably need to get started on the upgrade gate script very soon, too 16:33:52 hughsaunders I'm happy to work with you on it, but our gating issues are starting to cause major disruption and the tests are not trusted, resulting in delays for reviews 16:34:25 odyssey4me: cool, lets collaborate, I thought I saw you working on some sort of gating matrix. 16:34:32 For upgrade testing: are people ok with just verifying the upgrade runs then running tempest after the upgrade? 16:34:44 As in install (current-1), upgrade to current, run current's tempest 16:34:53 Should probably put that in the spec 16:35:06 deploy, tempest, upgrade, tempest 16:35:14 I think that is a great place to target for liberty, palendae 16:35:34 palendae I think tempest both times would be better - and I think that the upgrade tests should perhaps be periodic, rather than per commit - the upgrade test will be very long 16:35:39 otherwise you don't know whether hte initial deploy or the upgrade failed 16:35:51 hughsaunders: Why a tempest in the middle, out of curiosity? I'm leaving out because of (current -1) because (hopefully) its gate is doing that 16:35:59 odyssey4me: Yes, that needs to be periodic 16:36:09 I need to get it written before I ask infra about it, but yep 16:36:50 palendae we'll be doing work with infra and can help with that - figure out the bits in our repo and once you have a process that works then we can put a gate test in that uses it 16:36:57 Yep 16:36:59 palendae: tempest is a reasonably short amount of time out of a deploy-upgrade run, and running it after the initial deploy makes determining the source of a problem easier 16:37:08 hughsaunders: Fair enough 16:37:31 agreed with hughsaunders - if we don't know that the starting env was ok, then any fails in the upgrade are not determinable 16:38:40 Ok, that's fair 16:38:43 It's why I asked :) 16:39:02 beyond liberty, in the M cycle, we have these blueprints to cover 16:39:11 #link https://blueprints.launchpad.net/openstack-ansible/+spec/independent-role-repositories 16:39:25 #link https://blueprints.launchpad.net/openstack-ansible/+spec/implement-rally-support 16:39:49 #link https://blueprints.launchpad.net/openstack-ansible/+spec/multi-platform-host 16:40:25 the rally role is actually ready already, but before we register a new repo for it we need to agree to register independent repositories for roles 16:41:04 ie we need the agreement in principle to do it which is covered in https://review.openstack.org/213779 16:42:02 I'd like us to get that review ready so that I can push it out to the ML for feedback next week. 16:43:23 #action hughsaunders and odyssey4me to prepare a spec for https://blueprints.launchpad.net/openstack-ansible/+spec/split-aio-gates 16:43:59 does anyone want to work with me to prepare a spec for https://blueprints.launchpad.net/openstack-ansible/+spec/multi-platform-host ? 16:44:39 * hughsaunders looks around for redhat people 16:45:17 hughsaunders so the idea I have is simply to instrument the plays to support alternative platforms 16:45:27 it's easy enough to do - I have tested it out and have a method 16:45:34 Yeah, there's a standard pattern for that now 16:45:34 at least from a package standpoint 16:45:45 Per-distro vars/playbooks 16:45:56 aside: https://review.openstack.org/215093 has merged on hpcloud - master is fair game (just rebase) 16:46:19 palendae no need for per distro playbooks - just vars will do 16:46:23 Ah, cool 16:46:33 it won't make the code path onerous, so I think it's actually pretty easy 16:46:33 I must be thinking of puppet then 16:46:38 Yeah 16:47:08 also https://blueprints.launchpad.net/openstack-ansible/+spec/add-support-for-systemd will help 16:47:23 I think that cloudnull has mostly got that worked out in a lab already 16:47:45 Yep 16:47:57 And I would imagine there will be talk of moving ot 16.04 16:48:02 This needs to be done before that 16:48:28 Or, rather, would make doing that move much, much easier 16:48:34 yep 16:48:36 And supporting distros already on systemd 16:49:42 #topic Reviews 16:49:58 as always, please keep an eye out on https://review.openstack.org/#/q/starredby:%22Jesse+Pretorius%22+project:stackforge/os-ansible-deployment,n,z 16:50:00 #link https://review.openstack.org/#/q/starredby:%22Jesse+Pretorius%22+project:stackforge/os-ansible-deployment,n,z 16:50:20 #topic Open discussion 16:50:30 anyone want to raise anything? 16:52:33 alright, that's it then - thank you all for your time! 16:52:37 #endmeeting