13:01:14 #startmeeting powervm_driver_meeting 13:01:15 Meeting started Tue Apr 11 13:01:14 2017 UTC and is due to finish in 60 minutes. The chair is esberglu. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:01:16 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:01:19 The meeting name has been set to 'powervm_driver_meeting' 13:01:36 \o 13:02:13 o/ 13:02:35 #topic In Tree Driver 13:02:35 efried: review this in background 13:02:36 https://review.openstack.org/#/c/448134/ 13:02:39 ack 13:03:11 efried: Need anything from us in-tree? 13:03:38 Other than reviewing once the rebases all go through? 13:03:45 esberglu I don't think so. Assume CI will be a separate topic. 13:03:48 yeah 13:04:52 #topic OOT Driver 13:05:10 thorst: Are there any reviews that we are still trying to push through CI OOT? 13:05:26 I think at least most of them have a CI +1 now 13:05:28 esberglu: tons 13:05:41 we need to do a full on scrub of things 13:06:14 Yeah I think that's the best bet 13:06:36 Gonna be a pain to redeploy until that packaging stuff while stacking is fixed though 13:07:14 well, I think that we have a few things that we should chat through and then once we're relatively solid...we can then go ahead and redeploy OOT 13:07:18 OOT CI... 13:07:30 net, there are about 6 or so change sets we need to scrub through. 13:09:27 Links? I will keep an eye and try to recheck as soon as any fail 13:09:56 well, some will take effort. But here's a first pass (and they're not just blocked by CI, but also reviews) 13:10:06 https://review.openstack.org/448134 13:10:14 https://review.openstack.org/448249 13:10:23 https://review.openstack.org/447762 13:10:29 https://review.openstack.org/453024 13:10:39 https://review.openstack.org/452712 13:10:49 https://review.openstack.org/447763 13:11:10 https://review.openstack.org/416667 13:12:15 I think that's it 13:12:18 from my end 13:12:28 Okay. I'm kicking off rechecks on the ones that haven't had a run yet 13:14:04 Okay is that it for OOT (CI separate topic) 13:14:12 longer term... 13:14:18 The other two that will need reviews are https://review.openstack.org/#/c/452220/ and https://review.openstack.org/#/c/452222/ 13:14:21 I would like to see us start to look at PCI direct attach 13:14:31 the latter is gated behind those ceilometer changes thorst posted 13:14:33 in case anyone has spare cycles. 13:15:01 But the former should be good, though it appears to have failed tempest with different tests than before the recheck... 13:15:41 Are you saying it failed different tests? Or ran different tests? 13:15:59 Failed 13:16:08 Whew okay good 13:16:24 Haha yeah, the latter would definitely be more confusing to deal with 13:17:33 esberglu: Do you know if the CI was down/didn't run on a patch when it first came in and I do a powervm: recheck comment on it, will CI run on that patch? 13:17:44 Yeah it will 13:17:45 Talking https://review.openstack.org/#/c/454506/ in particular 13:17:45 Ok 13:19:29 thorst: You have anything else you wanted to say about PCI direct attach, or just for awareness at this point? 13:20:19 just awareness...start thinking about it 13:20:23 Yeah, that was a bit of a bomb there, thorst 13:20:40 We talking queens? 13:20:45 yeah 13:20:48 long term game 13:20:48 ight 13:21:11 I suspect the code for that right now is very kvm centric 13:21:28 Oh hell yeah. 13:21:35 so I'd like our team to build an understanding of what's there...does it fit...do we have changes needed in the core driver...etc... 13:21:40 #agreed PCI direct attach is a big fish, long term. 13:21:43 core driver --> manager 13:22:00 (been looking for an opportunity to use #agreed) 13:22:15 heh 13:22:20 (oh, it only works for the chair. Rats) 13:22:27 anywho, that's all I had 13:22:39 Okay lets move on 13:22:43 #topic CI 13:22:56 So we already talked about pushing the backlog through 13:23:20 And then at that point I think the plans is to wipe the mgmt node and redeploy from scratch 13:23:34 Cuz this whole saving a sinking ship with a bucket thing isn't working 13:25:14 Other than that we have efrieds local2remote changes 13:25:26 I know you want to have that installed as part of the base image template 13:25:37 Well, that's open for discussion, but I think that's the right place to do it. 13:25:38 But we would still have to reinstall during the prep_devstack part 13:25:40 and we have a long term plan to go 'patchless'? 13:25:58 We shouldn't have to reinstall it at all then. 13:26:12 We don't know which version of pypowervm we want until then 13:26:35 You mean because newton and ocata are still on the table at that point? 13:26:37 Doesn't matter. 13:27:09 If we diddle /etc/environment in the base image, and have the .py present, pre-1.1.2 will simply ignore, and we'll still have to do the old style local2remote patching if newton/ocata. 13:27:15 But 1.1.2 will just work. 13:28:09 To summarize: the new stuff I'm proposing can be part of the base image template; and the old stuff will need to change to be conditioned on pre-pike-only once 1.1.2 is released. 13:28:36 Well we can't pull the version from requirements then 13:28:53 When requirements says 1.1.2, we can. 13:29:10 Cause we should be pulling from the requirements of the branch we're testing. 13:29:15 What if requirements changes between the nightly build and the CI run? 13:29:45 Yeah, we still pull from requirements of whatever branch we're testing. 13:29:57 We don't know which branch we are testing until way after the image build 13:30:03 Maybe I'm missing something here 13:30:06 Doesn't matter if we flipflop back and forth between 1.1.2 and pre-1.1.2. 13:30:22 Yeah but we still have to reinstall at that point is what I'm saying 13:30:35 Only if we're testing pre-pike. 13:30:56 Whole point of this is we don't have to patch pypowervm at 1.1.2. 13:31:29 ...which should be concomitant with pike. And pre-1.1.2 with pre-pike. 13:31:56 My point about the reqs changing between the image build and the run still stands 13:32:11 Edge case, but still 13:32:19 reqs of pypowervm specifically? 13:32:38 I'm still not seeing the problem. 13:32:39 Yeah, what if we bump the reqs up to 1.1.3? We would then have to wipe all of the nodes out 13:32:44 And build from scratch 13:32:54 no, we shouldn't. 13:33:04 Cause 1.1.3 will then get pulled in when we stack. 13:33:18 And 1.1.3 will still just work when the env var & .py are present. 13:33:18 Ohhh duhhhh 13:33:35 Whole reason we're doing things this way. 13:33:41 Yeah my b 13:33:55 Cool 13:34:49 I still owe you a patch / review on that stuff 13:35:16 Yeah, but this is background work, as there are lulls in the flushing-CI-backlog business. 13:35:40 As excited as I am about trying it out and getting it working... 13:35:56 Yep. I would like to get it in for when we do the full CI redeploy 13:36:06 So we don't have to redeploy again right after 13:36:59 For the sake of having it documented, we will ultimately need: o) patches to deploy the .py and /etc/environment change; o) image template rebuild; o) 1.1.2 released; o) g-r and nova project requirements bumps 13:37:47 I think in that order. 13:38:36 the image template rebuild & CI redeploy could happen at any point after the neo-os-ci change is merged. 13:39:36 Yeah I think so 13:39:39 oh, and o) separate neo-os-ci (or powervm-ci?) change to condition manual local2remote patch on pre-pike. I think that needs to come in sync with the reqs bumps. 13:39:49 which will be fiddly. 13:40:00 Cause the manual local2remote patch will simply fail on 1.1.2. 13:40:07 merge conflict there. 13:40:23 I guess we could change that right now to condition on the pypowervm version. 13:40:31 rather than the openstack branch. 13:40:51 Then no fiddly. 13:41:13 Sure I like that 13:41:19 Will probably have to use pbr to do version compare, groan. 13:41:25 Okay, let's move on. 13:41:58 I didn't have anything else base CI 13:42:19 thorst: Did you want to talk OSA CI here? Or separate discussion on that? 13:42:55 lets have a separate discussion on that when we stabilize 13:43:08 Okay 13:43:14 That's it from my end then 13:44:12 #topic Other Discussion 13:44:34 nbante: You have anything you are working through? 13:44:37 #action thorst As able, take a look at https://review.openstack.org/#/c/438729/12..13 (tasks/* bits - the others are rebase) and let's move toward getting this back to nova-powervm for https://bugs.launchpad.net/nova-powervm/+bug/1680947 13:44:38 Launchpad bug 1680947 in nova-powervm "Use PrintingDurationListener, get rid of PowerVMTask" [Wishlist,New] 13:44:56 Low priority. 13:45:39 I tried basic tests with tempest and those are passed successfully. Now working to configure external storage then we can try deploy test cases 13:46:05 it looks like I need to figure out how to get our v7k on a private IP address for nbante 13:46:12 it'll take some wiring :-/ 13:47:04 meanwhile I'll check,what are configuration files need to modify and once all changes done. I'll redploy to get cinder configured 13:48:30 nbante: Cool sounds good 13:48:43 Doesn't look like jay is on 13:48:50 Any final topics before I close? 13:50:50 #endmeeting