14:02:59 <edmondsw> #startmeeting PowerVM Driver Meeting
14:03:00 <openstack> Meeting started Tue Aug 14 14:02:59 2018 UTC and is due to finish in 60 minutes.  The chair is edmondsw. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:03:01 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:03:04 <openstack> The meeting name has been set to 'powervm_driver_meeting'
14:03:08 <edmondsw> #link agenda: https://etherpad.openstack.org/p/powervm_driver_meeting_agenda
14:04:29 <edmondsw> efried mujahidali mdrabe gman-tx ^
14:04:40 <efried> ō/
14:04:57 <edmondsw> #topic In-Tree Driver
14:05:44 <edmondsw> nothing new here for Rocky
14:05:47 <efried> .
14:06:01 <efried> I'm going to propose update_provider_tree probably this week, hopefully today.
14:06:08 <edmondsw> ah, true
14:06:16 <efried> oh, but not for rocky.
14:06:23 <edmondsw> you are going to do that both IT and OOT, right?
14:06:25 <efried> yes
14:06:39 <edmondsw> right, it would just be Stein
14:06:42 <efried> yes
14:07:03 <efried> I don't see us doing anything more for rocky at this point.
14:07:16 <efried> unless we find actual bugs
14:07:16 <edmondsw> in the coming weeks we'll need to start talking about what else we do in Stein, but I'm not sure we're ready to dig into that today
14:07:29 <edmondsw> yep, I think we're done with Rocky
14:07:51 <edmondsw> #topic Out-of-Tree Driver
14:08:11 <edmondsw> as efried mentioned, will be update_provider_tree changes coming soon here
14:08:28 <gman-tx> what do we mean by soon?
14:08:36 <efried> this week / today
14:08:39 <edmondsw> see above... "hopefully today"
14:08:53 <gman-tx> got it
14:09:16 <efried> This will be a backward-compatible update_provider_tree, btw.
14:09:33 <efried> Which will then be extended later for dev passthrough work (coming up in a later topic in this meeting)
14:10:46 <edmondsw> yep, there is nova code that has been doing x for us, and we're just reproducing x in our code as the first step, and then we'll start tweaking it
14:11:03 <efried> correct
14:11:39 <edmondsw> mdrabe I'm still trying to get a multinode devstack working for you in my spare minutes here and there
14:11:45 <edmondsw> if this is blocking you and you want to try to work on it yourself, let me know and I can get you access to the systems.
14:12:11 <edmondsw> otherwise, hopefully I can get that done this week
14:12:28 <edmondsw> learning a lot about devstack
14:12:35 <edmondsw> painful...
14:12:50 <edmondsw> (especially networking)
14:13:22 <edmondsw> on a brighter note, I got the readthedocs builds working for nova-powervm and networking-powervm
14:13:48 <edmondsw> and we branched stable/rocky and tagged rc1 for all the *-powervm projects and
14:14:26 <edmondsw> have updated stable/rocky's .gitreview and tox.ini to point to rocky
14:14:48 <edmondsw> anything else for OOT discussion?
14:15:04 <efried> I've been keeping an eye on the req project, still hasn't forked.
14:15:15 <edmondsw> yeah, it'll probably be a couple weeks
14:15:28 <edmondsw> tx for looking out
14:15:59 <edmondsw> gman-tx all the storage work that I'm aware of your team doing has been merged at this point... anything else coming?
14:16:25 <gman-tx> that should be it
14:16:30 <edmondsw> cool
14:16:45 <edmondsw> #topic Device Passthrough
14:16:52 <edmondsw> efried over to you
14:17:46 <efried> okay, patch series is up through the driver work.
14:18:07 <efried> #link device passthrough nova-powervm series https://review.openstack.org/#/q/status:open+project:openstack/nova-powervm+branch:master+topic:bp/device-passthrough
14:18:52 <efried> The WIP needs test, but it passes our CI, which means a) the update_provider_tree bits for the compute RP are backward compatible; and b) we (probably) properly no-op the child providers when no inventory.yaml is specified.
14:21:17 <efried> The spec is getting attention and has some good (post-merge) commentary on it
14:21:36 <efried> pursuant to which I believe I'm actually going to post a fup patch to edit accordingly, and it will impact the design.
14:21:58 <edmondsw> nova or nova-powervm spec?
14:22:08 <efried> #link merged nova-powervm device passthrough spec https://review.openstack.org/#/c/579359/10
14:22:33 <efried> Meanwhile, kosamara has published a corresponding spec in nova that is libvirt-focused.
14:22:39 <edmondsw> hey, look at that... non-powervm folks commenting on our spec, I like it
14:23:09 <efried> #link proposed nova spec for PCI-via-placement https://review.openstack.org/#/c/591037/
14:23:22 <efried> I have, ahem, reviewed it with some suggestions.
14:24:23 <efried> TL;DR, what I'd like to see is for that spec to a) define more clearly the split of responsibilities between the virt driver and the rest of nova; and b) include both powervm and libvirt, so we can be more sure that split is being done appropriately.
14:24:53 <efried> There is a possible future where we land generic device passthrough in the powervm driver *in tree* in Stein.
14:25:33 <edmondsw> that'd be nice
14:25:51 <gman-tx> yea it would
14:25:55 <edmondsw> I think we're about to the point where it makes sense to be doing co-development IT and OOT
14:26:04 <edmondsw> and this would be the perfect topic on which to do that
14:26:05 <efried> Don't hatchet your counts before they chicken, but it's a possibility.
14:26:20 <edmondsw> yep, understood
14:26:59 <efried> So
14:27:06 <efried> The plan for this week is
14:28:31 <efried> - Split https://review.openstack.org/589668 into a) generic backward-compatible compute-node-only update_provider_tree outside (but as a prereq) of the dev passthrough series; and b) add device passthrough to (a) within the series
14:28:40 <efried> - Mirror ^ (a) in tree
14:28:48 <efried> - Propose a spec update
14:28:59 <efried> - Keep shepherding the nova spec
14:29:03 <efried> EOM
14:29:42 <edmondsw> sounds good
14:29:51 <edmondsw> #topic PowerVM CI
14:29:57 <edmondsw> mujahidali you're up
14:30:09 <mujahidali> CI looks good after the outage and redeployment.
14:31:27 <mujahidali> We are able to stack successfully for multinode setup on stage env(thanks to :esberglu) but there are 3 tempest failures.
14:32:03 <edmondsw> are you looking at those? Let me know if you need help
14:32:19 <mujahidali> yeah, I need help.
14:32:55 <mujahidali> and fo the vscsi stable ocata also, there are 3 tempest failures.
14:34:06 <edmondsw> ok, ping me details after the meeting
14:34:12 <mujahidali> sure.
14:34:43 <edmondsw> mujahidali have you started working to add rocky to the CI?
14:34:57 <mujahidali> not yet.
14:35:26 <esberglu> mujahidali: Sounds like the zuul merger issues are resolved. Most of those are 1 time setup steps for new servers
14:35:26 <edmondsw> shouldn't be too difficult... but I'd like it done sooner rather than later
14:35:31 <edmondsw> maybe something you can work on tomorrow?
14:35:39 <edmondsw> (I mean the rocky CI)
14:36:07 <esberglu> However, the steps should be automated in the playbooks so that we don't hit the same issues if we ever use new merger severs
14:36:11 <esberglu> *servers
14:36:32 <esberglu> Or, at a minimum, the steps should be documented
14:36:47 <mujahidali> I wanted to give it a try today, but I already deployed the CI with multinode change, so most probably will do it on thursday(tomorrow is holiday).
14:37:35 <edmondsw> ok
14:37:52 <esberglu> mujahidali: You shouldn't need to do a full redeploy to test the stable/rocky zuul changes
14:38:03 <mujahidali> esberglu: thanks for the help for resolving the zuul_merger issues.
14:38:20 <esberglu> You can just edit /etc/zuul/layout/layout.yaml
14:38:25 <esberglu> And then restart the zuul service
14:38:27 <edmondsw> esberglu it looked like you enabled stable/queens CI runs without trying on staging first? The commit included updates to both staging and production yaml in one commit
14:39:03 <edmondsw> or maybe you tried manually?
14:39:11 <edmondsw> what do you suggest?
14:39:49 <esberglu> edmondsw: I'm sure I went through manual runs for the new local.conf files
14:40:43 <esberglu> For the zuul layout.yaml I probably didn't even test them tbh
14:41:24 <esberglu> I would test the new local.confs. Then make sure I'm on the zuul layout review
14:41:24 <edmondsw> pretty straightforward
14:41:31 <edmondsw> wfm
14:42:11 <edmondsw> ok, anything else for CI?
14:42:42 <esberglu> mujahidali: Do you have time to take this patch over?
14:42:42 <mujahidali> esberglu: apart from changing the /etc/zuul/layout/layout.yaml I think I need to redeploy the cloud_nodes and zuul_mergers.
14:42:45 <esberglu> https://review.openstack.org/#/c/565239/
14:44:01 <edmondsw> mujahidali if zuul mergers are working now, why redeploy?
14:44:06 <esberglu> mujahidali: To start running changes on stable/rocky, you just have to make the change to the zuul layout.yaml file and restart the zuul service
14:44:17 <esberglu> No redeploy is necessary
14:44:28 <mujahidali> okay
14:44:47 <mujahidali> :esberglu will take it https://review.openstack.org/#/c/565239/, but will need a little bit help as well.
14:44:52 <edmondsw> mujahidali but first need to make the powervm-ci changes to add local.conf for rocky
14:45:45 <esberglu> The new local.conf files *should* just work. Unless something has gone into rocky that is breaking since they branched
14:46:05 <esberglu> Just copy the master local.conf files
14:46:10 <edmondsw> esberglu yep, should be simple
14:46:27 <mujahidali> okay
14:46:52 <edmondsw> and I'd like that done quickly so that if we need to make any changes on rocky, i.e. a bug fix, it will be tested by CI
14:47:05 <edmondsw> so that should take precedence over multinode, vscsi, etc.
14:47:46 <mujahidali> yes
14:48:14 <mujahidali> I think it should be completed by end of this week.
14:48:27 <edmondsw> let's shoot for Thursday if we can
14:49:02 <edmondsw> if you can get things ready for review Thursday, I will try to review quickly
14:49:22 <edmondsw> esberglu I'm hoping you could also review quickly
14:50:10 <edmondsw> alright, moving on
14:50:15 <edmondsw> #topic Open Discussion
14:50:20 <edmondsw> anything else to discuss?
14:50:49 <mujahidali> I will be OOO from 22nd to 24th.
14:50:49 <esberglu> edmondsw: Yeah I can review quickly, just ping me when they are ready
14:50:59 <edmondsw> tx
14:51:09 <efried> PTG and summit?
14:51:22 <efried> Negotiations for the former have commenced. Anything to report?
14:51:44 <edmondsw> I expect efried gman-tx and myself to be attending the PTG but haven't received final word
14:52:20 <edmondsw> summit discussions aren't as far along
14:52:33 <edmondsw> probably similar
14:52:43 <efried> Update on dev passthrough topic: sean-k-mooney is proposing an alternative schema
14:52:43 <efried> #link etherpad to brainstorm device passthrough schema https://etherpad.openstack.org/p/generic-device-schema
14:53:34 <edmondsw> will look
14:55:34 <efried> no hurry, imagine it'll go on for a bit, just thought I'd bring it up.
14:56:01 <edmondsw> tx
14:56:11 <edmondsw> sounds like that's it for today
14:56:14 <edmondsw> thanks everyone
14:56:17 <edmondsw> #endmeeting