13:02:03 <ndipanov> #startmeeting sriov 13:02:04 <openstack> Meeting started Tue Nov 24 13:02:03 2015 UTC and is due to finish in 60 minutes. The chair is ndipanov. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:02:05 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:02:08 <openstack> The meeting name has been set to 'sriov' 13:02:36 <ndipanov> so I had 2 things in mind for today 13:03:01 <ndipanov> look at the specs that are outstanding seeing that the freeze is close 13:03:15 <ndipanov> look at the patches currently up for review 13:03:25 <lbelivea> sounds good 13:03:36 <ndipanov> (and I have to say that due to some internal stuff I was doing last 2 weeks I was not reviewing as much as I should have) 13:03:41 <ndipanov> but that should improve this week 13:04:04 <ndipanov> and #3 if we have time - re-visit the neutron<-> nova interactions 13:04:36 <vladikr> hi everyone 13:04:36 <ndipanov> ok so specs that need love 13:05:01 <ndipanov> hi vladikr - saw your patches - queued them for this afternoon with coffee and cake 13:05:08 <lbelivea> do we have a list of the specs that we are interested in ? 13:05:19 <ndipanov> lbelivea, we should do 13:05:21 * ndipanov digs 13:06:01 <ndipanov> lbelivea, https://etherpad.openstack.org/p/mitaka-nova-priorities-tracking 13:06:07 <vladikr> ndipanov, :) thanks, it's a draft, going to add tests today and still having problems with my sriov card, so still didn't tests the whole thing 13:06:34 <ndipanov> and then scroll down to SR-IOV subteam 13:06:39 <lbelivea> got it 13:06:46 <moshele> hi 13:06:52 <ndipanov> #info priorities tracking etherpad has stuff that we should be looking at 13:06:58 <ndipanov> hi moshele 13:07:06 <vladikr> moshele, Hi 13:07:34 <shaohe_feng> ndipanov: can you have a look at this spec? https://review.openstack.org/#/c/139910/ 13:08:08 <ndipanov> you seem to have a -1 from baoli 13:08:51 <shaohe_feng> baoli: hi 13:08:53 <baoli> shaohe_feng: I saw your comments, I think that You just need to clarify in the text. 13:09:39 <shaohe_feng> baoli: OK. I need need to update the spec? 13:09:48 <ndipanov> shaohe_feng, yeah 13:10:21 <ndipanov> so after that - I am +1 on it 13:10:31 <ndipanov> shaohe_feng, do you already have some code ready? 13:11:34 <ndipanov> once you upload a fixed version - I can just go and ping john (or you can do it too) and he'll likely approve it 13:11:36 <baoli> shaohe_feng: will +1. 13:11:42 <ndipanov> as this was discussed at the summit 13:11:46 <ndipanov> johnthetubaguy, ^ 13:11:52 <shaohe_feng> ndipanov: just prototype long time ago . If the spec approve, I will start the code. 13:12:02 <ndipanov> ok makes sense 13:12:16 <shaohe_feng> ndipanov: thank you. 13:12:20 <shaohe_feng> ndipanov: let me do ti now. 13:13:38 <ndipanov> ok so there are 2 more that might be interesting to review 13:13:39 <ndipanov> https://review.openstack.org/#/c/195662/ 13:13:50 <ndipanov> this one is not related only to pci 13:13:56 <ndipanov> but can be quite useful 13:14:02 <johnthetubaguy> ndipanov: shaohe_feng: yeah, lets get the -1 addressed, then it should pop higher up my list 13:14:15 <ndipanov> thanks johnthetubaguy ! 13:14:49 <lbelivea> ndipanov: Gonna review it later today 13:14:53 <ndipanov> baoli, would you mind taking a look too 13:15:43 <baoli> ndipanov, sure 13:15:55 <baoli> ndipanov, will take a look 13:15:59 <johnthetubaguy> ndipanov: once you folks are all +1 on these specs, do ping me and I will try jump on those quickly, they all sound like useful things 13:16:21 <ndipanov> johnthetubaguy, awesome - will probably ping you later today about 2 of them 13:16:34 <ndipanov> so one more that seems to have reached some consensus is https://review.openstack.org/#/c/199488/ 13:16:49 <ndipanov> so we can add that one to the list of "ready for wider review too" 13:16:51 <ndipanov> cool 13:16:56 <ndipanov> so is there something we missed 13:16:57 <ndipanov> ? 13:17:16 <lbelivea> I have 3 patches in the pipeline that are ready for review 13:17:37 <lbelivea> https://review.openstack.org/#/c/216049/ 13:17:38 <lbelivea> https://review.openstack.org/#/c/242555/2 13:17:45 <lbelivea> https://review.openstack.org/#/c/242573/9 13:17:54 <ndipanov> ok that's the next part if there are no more specs questions 13:18:15 <vladikr> ndipanov, there is this one as well : https://review.openstack.org/#/c/239875/ 13:18:50 <ndipanov> vladikr, ah right - completely forgot about that one as I got sucked into a different thing recently 13:18:59 <lbelivea> vladikr: Going to take some time today to review it as well 13:19:01 <moshele> you are depended on this neutron patch https://review.openstack.org/#/c/246923/ 13:19:08 <vladikr> ndipanov, I've missed moshes comments on it though 13:19:10 <ndipanov> seems like there is some questions on it by moshele 13:19:17 <vladikr> moshele, yeah, will update it 13:19:57 <ndipanov> will review that one today as well and add it to the therpad (with a line over it since it's pending update) 13:20:49 <vladikr> thanks 13:21:13 <ndipanov> ok so now onto patches 13:21:46 <ndipanov> first - I'd like to talk about this one: https://review.openstack.org/#/c/227160/ 13:22:00 <moshele> libvirt fix for pf with no VF https://www.redhat.com/archives/libvir-list/2015-November/msg00934.html 13:22:00 <moshele> https://www.redhat.com/archives/libvir-list/2015-November/msg00960.html 13:22:14 <ndipanov> aaahhhh ok 13:22:37 <moshele> they will also add max VF in the xml 13:22:52 <lbelivea> mosele: nice ! 13:23:02 <ndipanov> that is pretty cool 13:23:18 <ndipanov> however I feel we want to work around it until they do int he libvirt driver 13:23:47 <ndipanov> though that may not be so easy 13:24:30 <baoli> ndipanov, maybe the same method (is_physical_function) be called from inside the libvirt driver to determine if a device is a PF? 13:24:41 <ndipanov> baoli, was about to say 13:24:55 <ndipanov> that means rootwrap call for every pci device 13:24:58 <baoli> ndipanov, we are thinking the same, then. 13:25:07 <ndipanov> so not really ideal 13:25:09 <ndipanov> :/ 13:25:59 <moshele> you want me to update the patch with this change 13:26:01 <moshele> ? 13:26:18 <ndipanov> moshele, thinking... 13:26:28 <ndipanov> let me comment on the patch after the meeting 13:26:52 <baoli> ndipanov: the device FS can be read by anybody, right? Do we need rootwrap for it? 13:26:58 <moshele> sure I will update the comments the but that it will be fixed in libvirt 13:27:54 <ndipanov> baoli, right we can get away without it 13:27:55 <moshele> s/but/bug 13:28:04 <ndipanov> moshele, so yeah go that route and add a comment 13:28:17 <moshele> ok cool 13:29:36 <ndipanov> anything else? 13:30:06 <moshele> I have this simple patch https://review.openstack.org/#/c/245108/ 13:31:05 <lbelivea> ndipanov: All the patches I have mentionned above, they have been in the pipeline for quite a while, some of them were reviewed by core 13:31:41 <ndipanov> moshele, +2 13:31:46 <ndipanov> lbelivea, ok let me lok 13:32:26 <ndipanov> lbelivea, so this one is a bit tricky I can't review it now but will look asap https://review.openstack.org/#/c/242573/ 13:32:35 * alex_xu lurks 13:33:08 <baoli> lbelivea: https://review.openstack.org/#/c/242573/9. Didn't get a chance to test the patch. But my concern about it is to save the pci requrest id in the binding profile. 13:34:12 <lbelivea> baoli: What is your concern ? I need to know more so that I can address your concerns. 13:34:24 <shaohe_feng> ndipanov: baoli: https://review.openstack.org/#/c/139910 13:35:21 <ndipanov> shaohe_feng, ok cool let me look 13:35:31 <shaohe_feng> ndipanov: thank you. 13:36:29 <baoli> lbelivea: I put it in comments, https://review.openstack.org/#/c/242573/6/nova/network/neutronv2/api.py. 13:36:41 <ndipanov> shaohe_feng, +1 13:36:58 <shaohe_feng> thank you. 13:37:00 <ndipanov> let's get baoli to like it too and then I'm sure john will be happy to +2 it 13:37:08 <ndipanov> and then onto teh codez 13:37:13 <shaohe_feng> great. 13:37:20 <shaohe_feng> baoli: ^ 13:38:00 <baoli> shaohe_feng: looks good. 13:39:01 <shaohe_feng> baoli: thank you. 13:39:09 <ndipanov> so lbelivea I will look at your patches later today too 13:39:12 <baoli> shaohe_feng: you are welcome 13:39:20 <lbelivea> ndipanov: thanks 13:39:20 <ndipanov> and then we can take it from there 13:39:33 <ndipanov> unless you have some concrete questions 13:40:48 <ndipanov> ok so ... some progress has been made :) 13:41:41 <ndipanov> so last thing I was going to mention the QoS thing and scheduling around it 13:42:06 <ndipanov> I've spoken to some people about it 13:42:21 <ndipanov> not necessarily representative of the Neutron community :) 13:42:25 <ndipanov> but "in the know" 13:42:37 <ndipanov> and also spent some time looking at it myself 13:42:50 <ndipanov> there are several ways to accomplish it 13:42:58 <ndipanov> none of them particularly easy :) 13:44:10 <ndipanov> so the bottom line is that someone will have to pick it up and try to figure it out. 13:44:20 <ndipanov> It's unlikely to be me this cycle though 13:45:08 <moshele> yes that what ajo said as well also the scheduler is going to be refactor in this cycle 13:45:49 <ndipanov> yeah well I am not a believer in "that's a good idea, but you have to wait until we do this massive refactoring here" 13:46:08 <ndipanov> also the scheduler folks have been "refactoring" it for several cycles now 13:46:46 <ndipanov> imho - if we have someone interested enough to pony up developer time to solve it 13:47:06 <ndipanov> they shouldn't be blocked on scheduler refactoring - though they should work with the team doing it of course 13:47:59 <ndipanov> but I don't see any serious specs being proposed around it sooo 13:49:20 <ndipanov> anywho 13:49:41 <ndipanov> if that's all we had - onto reviewing stuff (for me at least) and let's call it a meeting 13:50:34 <ndipanov> ah one more thing lbelivea 13:50:35 <ndipanov> https://review.openstack.org/#/c/239123/2 13:50:47 <ndipanov> why do we need a spec for this exactly? 13:51:00 <ndipanov> this is a bugfix! 13:51:51 <lbelivea> ndipanov: I don't think so, our initial intent was to bring all the fixes for cold migration under the same umbrella 13:52:05 <lbelivea> ndipanov: I think we can just focus on bugs/tickets 13:52:18 <ndipanov> lbelivea, so you are saying we don't need it right? 13:52:54 <ndipanov> I am interested in having that fixed too - and it is way more likely to get done if we can not block on the nova process here 13:52:57 <lbelivea> ndipanov: yeap, I can put it in "abandoned" 13:53:02 <ndipanov> yes please 13:53:09 <ndipanov> and open relevant bugs 13:54:08 <lbelivea> ndipanov: agreed, we have customers that uses our patches for supporting cold migration with PF and VF passthrough, trying to help the community and push those fixes :) 13:54:16 <moshele> we already have a bug for resize https://bugs.launchpad.net/nova/+bug/1368201 13:54:16 <openstack> Launchpad bug 1368201 in OpenStack Compute (nova) "resize with PCI devices doesn't work" [Low,In progress] - Assigned to Yongli He (yongli-he) 13:54:32 <ndipanov> moshele, excellent 13:54:58 <ndipanov> so I fixed some of that borkedness around NUMA/CPU pinning last cycle 13:55:12 <ndipanov> but pci devices use completely different code paths sadly 13:55:19 <lbelivea> moshele: Yes this is similar, but my patches focus on migration, this bug is a bit different where a different number of PCI devices needs to be taken care of during the resize 13:55:24 <ndipanov> the fixes also don't require db changes 13:55:34 <ndipanov> I think 13:55:38 <ndipanov> so that makes it easier 13:56:59 <ndipanov> anyway - we're hitting the top of the hour almost so let's get 5 minutes back for coffee 13:57:33 <ndipanov> lbelivea, I'll follow up with you on IRC about the migration fixes and patche reviews 13:57:44 <lbelivea> ndipanov: perfect 13:57:47 <ndipanov> good meeting everyone - progress is unstoppable ! :) 13:57:54 <ndipanov> #endmeeting sriov