13:02:01 <moshele> #startmeeting sriov
13:02:02 <openstack> Meeting started Tue Aug 30 13:02:01 2016 UTC and is due to finish in 60 minutes.  The chair is moshele. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:02:03 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:02:05 <openstack> The meeting name has been set to 'sriov'
13:02:10 <lbeliveau> o/
13:02:13 <wznoinsk> o/
13:02:13 <moshele> hi everyone
13:02:19 <edand> hi
13:02:25 <janki> o/
13:02:49 <moshele> #topic agenda SR-IOV/PCI Passthrough CI
13:03:34 <moshele> our CI had issues with suspend resume test
13:03:57 <prateek> Hi guys, I have started contributing to nova some time back and uploaded some patches and done some reviews and would love to be a part of this exciting subteam !
13:04:29 <vladikr> o/
13:04:30 <moshele> see https://www.redhat.com/archives/libvir-list/2016-August/msg00406.html
13:05:01 <moshele> prateek: everyone is welcome
13:05:26 <prateek> moshele, thanks
13:05:34 <moshele> it seem that there was security patch which cause problem in qemu <2.6
13:06:20 <moshele> it was revert in Ubuntu https://bugs.launchpad.net/ubuntu/+source/qemu/+bug/1612089
13:06:20 <openstack> Launchpad bug 1612089 in Ubuntu Cloud Archive mitaka "Fix for CVE-2016-5403 causes crash on migration if memory stats are enabled" [Undecided,New] - Assigned to Corey Bryant (corey.bryant)
13:06:36 <moshele> but we still see it in redhat 7.2
13:06:59 <lbeliveau> yeah, so this patch is still not fixed in qemu tree ?
13:07:56 <moshele> lbeliveau: I think the fix is working with qemu 2.6 but Ubuntu and redhat backport it to there distro and that cause problems
13:08:06 <lbeliveau> ok
13:08:33 <moshele> Ubuntu revert it
13:08:52 <moshele> we also have the migration patches in tempest
13:09:10 <moshele> migration test - https://review.openstack.org/#/c/339230/     migration revert test - https://review.openstack.org/#/c/348498/
13:09:12 <vladikr> I guess it's happening only with virtio bus, maybe in the meanwhile you can redefine virtio to ide or anything else?
13:10:12 <moshele> vladikr: currently or CI only run on Ubuntu so this is good for now
13:10:30 <vladikr> ah, ok
13:10:41 <moshele> anything else on CI?
13:10:48 <lbeliveau> wznoinsk: can you please help on those tempest test ?
13:11:03 <wznoinsk> lbeliveau: sure, looking now
13:11:08 <lbeliveau> they haven't been reviewed yet by any tempest cores
13:11:14 <moshele> also we need CI for PF-PT
13:11:30 <moshele> I am not sure if we can do it here in Mellanox
13:12:23 <moshele> wznoinsk: are you planning to test it on your CI ?
13:13:26 <wznoinsk> moshele: not at the start, we're only preparing to take on the SRIOV/PCI here
13:13:54 <wznoinsk> will extend the tempest coverage as we go as well but not add the additional setup
13:14:16 <moshele> wznoinsk: but Mellanox CI already test SR-IOV so it will be redundant
13:15:00 <wznoinsk> moshele: I'm not closing the door here, it's best if we could talk in a couple of weeks re PF-PT... I'll know enough to tell you yes/no for sure
13:15:15 <moshele> ok cool
13:15:36 <moshele> #topic Patches for Subteam Review
13:15:45 <moshele> https://etherpad.openstack.org/p/sriov_meeting_agenda
13:16:03 <moshele> most of the migration patches are merged
13:16:18 <moshele> we have only this one     Fix drop_move_claim() on revert resize - https://review.openstack.org/#/c/349060
13:16:37 <moshele> and maybe this one     pci: in free_device method compare devices according to pci_device id - https://review.openstack.org/#/c/347558/
13:16:44 <lbeliveau> waiting for jaypipes to review it, but I think he's real busy with RT
13:17:03 <moshele> lbeliveau: so maybe dansmith can help
13:17:39 <lbeliveau> I asked him, he looked at the latest patch rev and he was ok with it, but he'd like to see jaypipes review it first
13:17:49 <moshele> without Fix drop_move_claim() on revert resize - https://review.openstack.org/#/c/349060  migration-revert won't work
13:18:11 <lbeliveau> it would be nice if it could land in this release
13:18:26 <moshele> let my try to ping him
13:19:30 <moshele> this is fixed https://bugs.launchpad.net/neutron/+bug/1614092 by edand
13:19:30 <openstack> Launchpad bug 1614092 in OpenStack Compute (nova) "SRIOV - PF / VM that assign to PF does not get vlan tag " [Low,Confirmed]
13:19:41 <moshele> sorry not this one :)
13:20:25 <moshele> I mean this one https://bugs.launchpad.net/neutron/+bug/1614086
13:20:25 <openstack> Launchpad bug 1614086 in neutron "SRIOV- when numvfs=0 SRIOV agent is failed to start " [Low,Fix released] - Assigned to edan david (edand)
13:20:49 <lbeliveau> yeap, I ran into it as well while testing PF passthrough last week
13:21:03 <moshele> vladikr: do you have solution to https://bugs.launchpad.net/neutron/+bug/1614092
13:21:03 <openstack> Launchpad bug 1614092 in OpenStack Compute (nova) "SRIOV - PF / VM that assign to PF does not get vlan tag " [Low,Confirmed]
13:21:30 <moshele> is cloud init an option and we can move it to documentation
13:21:54 <lbeliveau> haven't looked at this one yet myself ...
13:22:17 <vladikr> moshele, not really.. yes, I was thinking about adding it to the metadata service/configdrive for cloud init to pickup
13:22:43 <vladikr> I brought this idea to danpb and dansmith and they didn't really like it
13:23:13 <vladikr> not sure how else to tackle it - maybe they did get my point then
13:23:26 <lbeliveau> what was their arguments ?
13:23:30 <moshele> we can talk about it in the summit
13:24:47 <moshele> we also plan to deprecate all of the SR-IOV agent config https://review.openstack.org/#/c/360447/
13:24:49 <vladikr> lbeliveau, their argument was that this is a vehicle to deliver user defined setting and what we want to do should be separate ..
13:25:06 <lbeliveau> vladikr: I see, thanks
13:25:48 <moshele> the plan is deprecation ocata
13:26:11 <wznoinsk> vladikr: there are 4 types of 'metadata' : http://www.madorn.com/openstack-metadata-types.html#.V7MM5vkrIfI
13:26:45 <wznoinsk> I'm not sure what data (vlan id) the network-data can be carrying tho
13:29:43 <vladikr> perhaps we didn't understand each other. maybe putting it in writing would be easier for everybody
13:29:44 <wznoinsk> try curl http://169.254.169.254/openstack/2015-10-15/network_data.json from any of your openstack VMs
13:30:29 <vladikr> I guess the point is that the vlan would be automatically set , as oppose to user setting it
13:31:21 <wznoinsk> sure, network-data is retrieved from neutron, also nothing user can modify AFAICT
13:31:35 <wznoinsk> only user-data is user-provided, vendor-data, network-data, meta-data not
13:32:48 <vladikr> wznoinsk, ok, thanks, I'll look into that more
13:32:49 <wznoinsk> anyway, just wanted to point out that maybe one of the other 'metadata' sources is the place to look at
13:33:52 <moshele> can we move to the next topic?
13:34:18 <vladikr> wznoinsk, yep, I was thinking about this the same way as we did the device-role-tagging - but this is different you are right
13:34:39 <vladikr> moshele, I think yes
13:35:12 <moshele> #topic  Specs for Review
13:35:31 <moshele> https://etherpad.openstack.org/p/sriov_meeting_agenda  Line 52
13:35:38 <lbeliveau> I dusted off this one https://review.openstack.org/#/c/362287/
13:36:13 <moshele> ok I will take a look
13:36:36 <moshele> if you have SR-IOV spec please add them to the agenda
13:37:18 <moshele> I just added this     Enable to share PCI devices between numa nodes - https://review.openstack.org/#/c/361140/
13:39:15 <moshele> also I forgot that the     SR-IOV: add agent QoS driver to support egress minimum bandwidth - https://review.openstack.org/#/c/347302/ is close for merge
13:39:46 <moshele> so maybe we will have SR-IOV BWG in newton
13:40:21 <moshele> Also we are planning to run the in Mellanox SR-IOV the neutron tempest qos tests
13:41:13 <moshele> ok, anyone want to raise and issue?
13:41:22 <lbeliveau> I'm good
13:41:55 <moshele> is anyone going to the summit in Oct?
13:42:13 <lbeliveau> I'm suppose to have confirmation in 2 weeks
13:42:14 <vladikr> not me.. :/
13:42:39 <wznoinsk> not sure yet, will know in a few weeks
13:44:44 <moshele> ok, if you go let me know so we can meet there
13:44:49 <moshele> anything else?
13:45:17 <wznoinsk> one open from me, if ppl didn't figure it out yet
13:45:57 <wznoinsk> I'm taking the SRIOV/PCI CIs of Intel into my garden, we should be ready with what the two CIs are doing now + more tempest testing before summit, not sure about multinode yet
13:46:20 <lbeliveau> cool
13:46:32 <moshele> good news
13:46:40 <wznoinsk> I'll keep you posted on this meeting, we can figure out what we want to test too
13:47:09 <moshele> ok cool
13:47:13 <wznoinsk> <eom>
13:47:38 <moshele> #endmeeting