14:00:06 <edmondsw> #startmeeting PowerVM Driver Meeting
14:00:07 <openstack> Meeting started Tue May  1 14:00:06 2018 UTC and is due to finish in 60 minutes.  The chair is edmondsw. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:08 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:10 <openstack> The meeting name has been set to 'powervm_driver_meeting'
14:00:13 <edmondsw> efried esberglu mujahidali ^
14:00:17 <esberglu> \o
14:00:27 <openstackgerrit> Merged openstack/ceilometer-powervm master: Trivial: Update pypi url to new url  https://review.openstack.org/565438
14:00:30 <efried> ō/
14:00:37 <edmondsw> #link https://etherpad.openstack.org/p/powervm_driver_meeting_agenda
14:00:43 <efried> mujahidali: Greetings, nice to meetcha.
14:00:44 <edmondsw> #topic In-Tree Driver
14:00:54 <edmondsw> #link https://etherpad.openstack.org/p/powervm-in-tree-todos
14:01:07 <edmondsw> esberglu walk us through the latest?
14:01:44 <mujahidali> efried: thanks :)
14:01:45 <esberglu> We need scenario testing to move forward on snapshot and localdisk. I'll go into more detail during CI section
14:02:06 <esberglu> vSCSI is blocked until we have some sort of volume testing
14:02:17 <esberglu> And cold mig/resize is waiting on multinode CI
14:02:29 <esberglu> That's pretty much it
14:02:39 <edmondsw> esberglu I dropped some comments on localdisk
14:02:46 <esberglu> ack
14:02:50 <edmondsw> of course CI is the priority atm
14:03:19 <edmondsw> looks like z got back into a runway quickly after they resolved their CI issues... hopefully things go as well for us
14:04:36 <edmondsw> #topic Out-of-Tree Driver
14:04:47 <edmondsw> #link https://etherpad.openstack.org/p/powervm-oot-todos
14:05:25 <edmondsw> I don't know of any progress made on OOT since the last meeting
14:06:04 <edmondsw> I think everyone's been focused elsewhere
14:06:11 <edmondsw> so, moving along...
14:06:23 <edmondsw> #topic Device Passthrough
14:06:28 <edmondsw> efried anything new here?
14:06:55 <efried> Looks like tetsuro is taking over the nrp-in-alloc-cands series, which is gooood.  He proposed the microversion patch last night.
14:07:08 <edmondsw> sweet
14:07:22 <efried> His bp/spec for including all resources in provider summaries got approved this morning, which is also gooood.  That was going to hold up the series.
14:07:46 <efried> I'm getting real close to having granular ready.  Just a couple more tests to write and a bug to work out.
14:07:56 <efried> Course, we're going to be racing each other.
14:08:06 <efried> Whoever loses is going to have some serious rebasing hell.
14:08:17 <efried> But at least progress is being made.
14:08:20 <efried> end.
14:08:25 <edmondsw> tx
14:08:37 <edmondsw> #topic PowerVM CI
14:08:42 <edmondsw> #link https://etherpad.openstack.org/p/powervm_ci_todos
14:08:49 <edmondsw> welcome mujahidali!
14:09:06 <esberglu> 3 major things going on right now for CI
14:09:16 <efried> Seems ill since last night
14:09:46 <esberglu> efried: Yep. We can start there. The queens cloud is seeing those same EOF errors we were seeing on the staging
14:10:18 <esberglu> However, on staging we changed the NIC type from some weird protocol to virtio and everything was fine
14:10:32 <efried> oh, right, this is what we were thinking was the glance wsgi business
14:10:42 <esberglu> Production is already using virtio
14:10:42 <efried> 'cept that should be in queens
14:10:57 <esberglu> efried: The glance wsgi thing was a real issue
14:11:06 <mujahidali> edmondsw: thanks :)
14:11:13 <esberglu> I'm running glance not with wsgi on production
14:11:51 <esberglu> So I will be doing more recon on the EOF errors today
14:12:09 <edmondsw> I assume that's top priority
14:12:20 <esberglu> yes
14:12:29 <edmondsw> and then next would be the scenario tests so we can resume IT efforts
14:12:32 <esberglu> 2) Scenario CI
14:12:54 <esberglu> I've got working scenario tests for OOT, running the suite with the same changes IT right now
14:13:19 <esberglu> I will be running those tests as part of the base CI job
14:13:33 <edmondsw> how can we tell they're working if the CI as a whole is not working?
14:13:41 <esberglu> edmondsw: Staging env.
14:13:44 <edmondsw> ah
14:14:15 <esberglu> Still some things to clean up there, but I think that should be ready by the end of the week
14:14:22 <edmondsw> got numbers on what that added to the OOT CI times?
14:14:38 <esberglu> edmondsw: I've only ran twice all the way through
14:14:57 <esberglu> 1 was 100s longer, the other actually ran faster than the base time I was comparing against
14:15:03 <edmondsw> wow
14:15:24 <edmondsw> that would be really nice if we find it's negligible
14:15:31 <efried> ++
14:15:51 <esberglu> edmondsw: Yep. Obviously I want to get a lot more runs going through to confirm first, but seems that it may be
14:16:03 <edmondsw> sure
14:16:18 <esberglu> 3) Multinode CI
14:16:42 <esberglu> I know what I want to do here and am 90% sure it's gonna work, just need to finish scenario before I test
14:17:22 <esberglu> I learned that the subnodes are NOT always on the same underlying compute host
14:18:17 <esberglu> So we're going to have to define an AZ for each neo host. Then we can force the subnodes to be on the same neo host in nodepool
14:18:56 <edmondsw> i.e. read the AZ of the parent, then specify that AZ for the subnode?
14:20:11 <edmondsw> my concern with AZs is that if we split each host into a different AZ, don't we have to specify the AZ on any deploy?
14:20:22 <esberglu> edmondsw: Not exactly
14:20:25 <edmondsw> and we don't care which AZ is used for the parent, so how do we pick one?
14:20:39 <esberglu> edmondsw: So we will have a bunch of providers in the nodepool conf, 1 for each AZ
14:20:57 <esberglu> And then we will say spawn 1 node and 1 subnode
14:21:14 <esberglu> Since the provider is a specific AZ, both node and subnode will be in it
14:21:37 <esberglu> The IP of the subnode is then saved in some nodepool files on the subnode
14:21:42 <edmondsw> so we just tell nodepool we want 1 node and 1 subnode, and for the node it picks a provider, we don't have to?
14:21:44 <esberglu> *on the main node
14:21:52 <esberglu> edmondsw: Yep
14:21:56 <edmondsw> ok nice
14:22:16 <esberglu> edmondsw: Only downside is that the config file is gonna be kinda gross, but oh well
14:22:22 <edmondsw> so how does the subnode AZ get specified?
14:22:39 <edmondsw> subnode will always come from the same provider?
14:22:47 <esberglu> edmondsw: Yes, you just have to config it right
14:22:50 <edmondsw> k
14:22:58 <esberglu> That's all I have
14:23:00 <edmondsw> work your magic :)
14:23:10 <edmondsw> let's talk a bit about vSCSI CI
14:23:17 <esberglu> Sure
14:23:26 <edmondsw> I assume that's next after the 3 we just covered
14:23:48 <esberglu> I thought we determined we didn't have the HW for it?
14:24:10 <edmondsw> I think that depends on which solution we're talking about
14:24:25 <edmondsw> in order to test vSCSI on every CI run, yeah, I think we would need more hardware
14:24:48 <edmondsw> but mriedem suggested that it would be ok to just have a job that could be run on-demand
14:24:59 <edmondsw> and I think we could probably do that with the hardware we have
14:25:44 <edmondsw> I don't think we've ever done anything like that before, but I assume there are other examples out there we could look at
14:26:41 <esberglu> edmondsw: I can start thinking about it
14:26:41 <edmondsw> esberglu thoughts?
14:26:44 <edmondsw> tx
14:26:54 <esberglu> edmondsw: Idk how we will run on demand once jenkins goes away
14:27:10 <esberglu> Maybe zuul v3 has a way to do so, but I haven't seen anything like that ever
14:28:08 <edmondsw> esberglu I'd probably start by asking mriedem if he knows of an example you could look at
14:28:12 <esberglu> We can sort through the details once I know more
14:28:35 <edmondsw> and then go from there, talk to him or infra about the zuul v3 options
14:28:58 <edmondsw> I wonder if that's one of the reasons zuul v3 isn't ready for 3rd party CI
14:29:45 <edmondsw> mujahidali coming up to speed?
14:30:02 <edmondsw> have a quick update on your status?
14:30:54 <esberglu> He's got access to all of the repos, etc. now. I've started adding him to CI reviews and will start giving him some tasks
14:32:11 <edmondsw> alright tx
14:32:17 <edmondsw> #topic Open Discussion
14:32:29 <edmondsw> anything else?
14:32:34 <mujahidali> esberglu: thanks for the update. edmondsw: yeah, I am looking into the code and reading the wiki
14:32:53 <edmondsw> ok good
14:33:37 <edmondsw> if there's nothing else, we can get some time back. Thanks!
14:33:41 <edmondsw> #endmeeting