13:02:26 <baoli> #startmeeting PCI Passthrough
13:02:27 <openstack> Meeting started Tue Dec  9 13:02:26 2014 UTC and is due to finish in 60 minutes.  The chair is baoli. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:02:28 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:02:30 <openstack> The meeting name has been set to 'pci_passthrough'
13:02:35 <baoli> Hi there
13:02:41 <beagles> hi
13:02:50 <irenab> hi
13:03:44 <pczesno> bi
13:03:45 <pczesno> hi
13:04:23 <itzikb_> hi
13:04:42 <baoli> #topic Bugs
13:04:58 <baoli> we dont' seem to have new bugs
13:05:23 <baoli> itzikb: https://bugs.launchpad.net/neutron/+bug/1397675, this is not a bug anymore?
13:05:26 <uvirtbot> Launchpad bug 1397675 in neutron "Updating admin_state_up  for port with vnic_type doesn't have affect when not using sriov nic agent" [Undecided,Confirmed]
13:05:42 <baoli> can we change its status?
13:06:05 <yongilhe> hi
13:07:41 <itzikb_> baoli: changed to?
13:08:00 <itzikb_> s/changed/change/
13:08:35 <baoli> if it's not a bug, then invalid, I guess
13:09:17 <sadasu> itzik, baoli: unless you want to use this to make a documentation update
13:09:37 <itzikb_> baoli: Not sure I follow. Do you think it's not a bug?
13:09:59 <itzikb_> baoli: I think it's a bug but I see no problem to solve it in Documentation
13:10:28 <itzikb_> baoli: For now at least ..
13:11:45 <itzikb_> I can add it - just not sure where
13:11:58 <baoli> itzikb: ok, it's a doc bug, then. The behavior is consistent with the design.
13:12:19 <baoli> Anything else on bug?
13:12:32 <itzikb_> I want to ask Regarding the resize
13:12:36 <baoli> #topic Reviews
13:12:37 <sadasu> Are there any voluteers to do the doc change? Else, I could take it up
13:13:05 <irenab> Is there any documentation regarding SR-IOV upstream?
13:13:08 <itzikb_> sadasu: I can do it if you'll show me where
13:13:23 <sadasu> baoli: maybe we can give a gap of 1 sec after asking a question before moving on? :-)
13:13:38 <baoli> sadasu, sure.
13:13:52 <sadasu> itzikb_:sure. I was going to look for it myself..but
13:14:03 <sadasu> I can help
13:14:24 <irenab> I am not aware of any documentation except for wiki
13:14:40 <sadasu> thats what I thought too
13:15:00 <sadasu> maybe we should take up documentation as a whole
13:15:08 <itzikb_> sadasu: +1
13:15:22 <yongilhe> does the reviews got any progress?
13:15:47 <baoli> irenab: talking about docs, how is the user doc/manual going on?
13:16:25 <irenab> baoli: main focus was on specs till Monday, so no progress here
13:16:52 <baoli> irenab, do you need help. Sounds like sadasu can lend some help in that front.
13:17:20 <irenab> baoli: help will be hihly appritiated
13:17:41 <irenab> neutron side is fine, since doc is auto generated from config files
13:17:50 <sadasu> do we need to a file a doc spec for this?!!
13:17:54 <irenab> need to document nova side
13:18:16 <irenab> sadasu: hope that node, seems more like bug
13:18:27 <irenab> ^not
13:18:37 <sadasu> irenab: cool
13:18:59 <yongilhe> neutron doc seems very cool.
13:19:18 <baoli> Ok, moving on?
13:19:37 <yongilhe> baoli, any bug, spec review got progress?
13:19:48 <irenab> baoli: ye
13:19:48 <baoli> #topic Reviews
13:19:58 <yongilhe> does shaohe_feng here,
13:20:09 <baoli> https://wiki.openstack.org/wiki/Meetings/Passthrough#Agenda_on_Dec._9nd.2C_2014
13:20:23 <yongilhe> he is cover me for the interface attach/detach blueprints.
13:21:37 <yongilhe> l'd like to introduce shaohe_feng to you
13:21:52 <shaohe_feng> Ｈｉ，　ａｌｌ
13:21:56 <pczesno> hi shaohe_feng
13:22:02 <baoli> we got one patch approved last week. Yongli's patch is pretty close for approaval
13:22:02 <irenab> hi
13:22:08 <baoli> shaohe_feng: HI
13:22:18 <shaohe_feng> Ｉ＇ｍ　ａ　ｆｒｅｓｈ　ｆｏｒ　ｏｐｅｎｓｔａｃｋ
13:22:26 <yongilhe> baoli, my patch is lack of review by core now.
13:22:45 <baoli> shaohe_feng, welcome aboard
13:22:58 <yongilhe> for resize, suspended by CI work for about a week.
13:23:03 <shaohe_feng> ｔｈａｎｋ　ｙｏｕ，　　ｂａｏｌｉ
13:23:55 <sadasu> shaohe_feng: welcome to openstack and pci passthrough
13:23:55 <baoli> yonglihe, Dan reviewed your patch https://review.openstack.org/#/c/131321/. He'd approve it soon, I think, after you fixed the grammar
13:24:17 <baoli> your pci resize patch is still WIP, right?
13:24:25 <yongilhe> baoli, yeah.
13:24:42 <itzikb_> yongilhe: I want to ask you about the resize
13:24:55 <yongilhe> go on, itzikb_
13:25:20 <itzikb_> yongilhe: Will it solve also the cold migration?
13:25:30 <shaohe_feng> sadasu:thank　you
13:25:58 <yongilhe> itzikb_, migration might be another patch if needed. now focus on resize only.
13:26:40 <itzikb_> yongilhe: ok.thanks
13:26:47 <yongilhe> itzikb_, mix them all in one patch might not effeciency and code will mass to reivew i think, plus lots of test cases.
13:27:08 <baoli> yonglihe, resize/cold migration shared the same flow/code, it seems to me.
13:27:51 <yongilhe> baoli, yeah, i think so, but not very sure, top entry might little diffrence, need to check.
13:28:32 <yongilhe> baoli, does the migration got progress?
13:29:05 <yongilhe> bali, i'm missing this windows to consult exper for you problem, sorry for this.
13:29:07 <baoli> yonglihe: I got it working. To upstream, it needs polish and unit tests
13:29:28 <baoli> yonglihe, thanks
13:29:41 <yongilhe> baoli, does the migration will depend on resize?
13:30:19 <baoli> yonglihe, after close look at the flows involved with live versus cold/resize, there doesn't seem to be much in common.
13:30:33 <baoli> So no dependencies
13:30:40 <yongilhe> baoli, seem very cool,
13:31:26 <baoli> The debate is still going on https://review.openstack.org/#/c/133138/. So if anyone has a minute, please weigh in with your opinions
13:32:15 <irenab> I am pretty good with proposed solution to include it in nova.conf
13:32:17 <yongilhe> baoli, sure, i'm going to check detail, the CI on discustion also.
13:33:19 <baoli> Moving on to BPs?
13:33:33 <yongilhe> yeah,
13:33:37 <irenab> yes
13:33:40 <baoli> #topic Blueprints
13:33:56 <yongilhe> shaohe_feng drop a initial interface attach/detach for sriov.
13:34:10 <yongilhe> shaohe_feng , clould you please attach a link here?
13:34:37 <baoli> yonglihe, talking about that, I saw that pczesno's spec also addresses the interface attach
13:34:53 <pczesno> but just from api
13:35:08 <shaohe_feng> https://review.openstack.org/#/c/139910/
13:35:08 <baoli> https://review.openstack.org/#/c/138808/
13:35:11 <pczesno> so i guess it's dependend on shaohe_fengs work
13:35:43 <baoli> pczesno, so your spec only addresses the API side of things, but not the actual implementation?
13:35:54 <baoli> pczesno, how about the nova boot command?
13:36:10 <pczesno> baoli,yes it does that
13:36:57 <yongilhe> pczesno, we can discuss the details about the 2 bps, include me in mail thread.
13:37:17 <pczesno> ok, i can drop the attach part from mine
13:38:03 <yongilhe> pczesno, you are so nice. but we might still need to align to make sure it make sense to comunity.
13:38:12 <baoli> shaohe_feng, please add you BP to this link: https://wiki.openstack.org/wiki/Meetings/Passthrough#Agenda_on_Dec._9nd.2C_2014
13:38:23 <shaohe_feng> baoli:  OK.
13:38:26 <pczesno> yongilhe, sure
13:38:33 <baoli> shaohe_feng, thanks
13:40:33 <yongilhe> any comments on bps? i'd like introduce some CI things.
13:41:21 <yongilhe> could we move on?
13:41:23 <irenab> just asking to review specs, it helps to shape them before cores take a look
13:41:35 <yongilhe> irenab, sure.
13:41:50 <baoli> irenab, I added some comments to your spec. Mainly it needs to address the nova side of changes required.
13:43:01 <irenab> baoli: thanks
13:43:06 <sadasu> irenab: I took a quick look...will review again and post commnets
13:43:19 <irenab> sadasu: great!
13:43:24 <baoli> #topic CI Testing
13:43:45 <baoli> yonglihe, please go ahead
13:43:48 <yongilhe> there some discuss about how to testin the pci/sriov
13:44:09 <yongilhe> first, use third-ci.
13:45:01 <yongilhe> but sdague, suggest fake a libvirt to testing any HW requiments feature.
13:45:14 <yongilhe> there is a bp for that.
13:45:44 <baoli> yongilhe: can you post the link here?
13:46:11 <yongilhe> i'm looking that now, a moment
13:46:26 <yongilhe> #link https://review.openstack.org/#/c/139927/
13:47:21 <yongilhe> on this approch, Daniel Berrange prefer the third CI, i also think so.
13:47:24 <baoli> yongilhe, please add this to the meeting wiki as well
13:47:30 <yongilhe> sure
13:47:34 <irenab> yonglihe: do you think it would be possible to run scenario tests with faked libvirt?
13:48:04 <irenab> sanpshot/migrate/resize/...
13:48:20 <yongilhe> Daniel point out this approche had it's limitation.
13:48:47 <yongilhe> irenab, please post your concern to the bp, then everyone notice it.
13:49:00 <irenab> yongilhe: sure
13:49:02 <baoli> yonglihe: so third CI versus fake libvirt, right? And you perfer third CI?
13:50:39 <yongilhe> baoli, yeah.
13:51:00 <itzikb_> yongilhe: Do you have CI framework that can test SR-IOV scenarios?
13:51:07 <irenab> I think for sriov we must run on real HW.
13:51:12 <yongilhe> the faked libvirt or qemu, will might need create a new project.
13:51:40 <yongilhe> irenab, i include this considerations in my bp alternatives.
13:52:16 <sadasu> Real HW for third party CI and fake libvirt for tempest/scenario tests
13:52:32 <irenab> yonglihe: will follow up on bp review
13:52:42 <yongilhe> before move on, we need dicuss and make things clear. welcom every one to comments that bp, we need some testing method to ensure pci code quality.
13:52:56 <sadasu> I see it as needing both but we could prioritize 3rd party CI over fake libvirt driver
13:53:01 <irenab> sadasu: I think at some point we are supposed to run tempest/scenario on 3rd party CI
13:54:12 <yongilhe> irenab, sadasu, i need rebase the testcases to scenario, to move progress
13:54:17 <baoli> yongilhe, will do.
13:54:21 <sadasu> irenab: I agree...but that would be a gate requirement...but individual contributors who don't have access to hw but want to test privately will have an option
13:54:41 <sadasu> thats why i am calling it lower priority
13:54:48 <irenab> sadasu: makes sense
13:55:05 <yongilhe> sadasu, good point
13:55:28 <yongilhe> please post to testcase reveiw, sadasu
13:55:50 <sadasu> yongilhe: will do. thanks for the reminder
13:56:10 <baoli> From that point of view, the fake libvirt is useful not only for sr-iov, but also everything else.
13:56:15 <yongilhe> irenab, what testcases your CI running?
13:56:42 <yongilhe> baoli, yeah, intend to, but not very easy, mostlikely need fake a qemu not libvirt.
13:57:18 <baoli> yongilhe, sure it's going to be easy.
13:57:20 <yongilhe> baoli, because there is a such thing in nova now, not that good as we expected.
13:57:40 <sadasu> baoli: not sure that is true. but we need to keep our scope to our project for now
13:58:20 <sadasu> baoli: only because I am not completely aware of other features in nova
13:58:40 <irenab> heyongli: currently automatically only API tests
13:58:58 <irenab> heyongli: neutron API
13:59:04 <yongilhe> irenab, good to know, thanks.
13:59:25 <yongilhe> time is up, then it's time to say see you..
14:00:07 <yongilhe> thanks everyone, bye.
14:00:09 <pczesno> bye
14:00:12 <irenab> bye
14:00:23 <itzikb_> bye
14:00:28 <sadasu> yonglihe: what do you mean by "such a thing: in a previous comment " because there is a such thing in nova now, not that good as we expected."
14:00:29 <baoli> until next time ....
14:00:42 <riwinters> bye
14:00:44 <baoli> #endmeeting