13:00:49 <moshele> #startmeeting sriov
13:00:50 <openstack> Meeting started Tue Jun 21 13:00:49 2016 UTC and is due to finish in 60 minutes.  The chair is moshele. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:00:51 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:00:53 <openstack> The meeting name has been set to 'sriov'
13:01:01 <moshele> hi
13:01:06 <moshele> let get started
13:01:20 <moshele> #topic Improving SR-IOV/PCI Passthrough CI
13:01:50 <moshele> lennyb: can you update on the SR-IOV CI for multlnode
13:02:27 <lennyb> we decided to leave the Docker container and implement multinode CI on 2 physical servers
13:03:14 <saneax> o/
13:03:23 <lennyb> currently I have a main job that starts controller and compute, and I am working on reporting logs with restrictions of infra ci and
13:03:49 <lennyb> logging
13:04:01 <moshele> ok cool
13:04:25 <moshele> sfinucan: do you have any update on the intel CI?
13:05:08 <sfinucan> moshele: Last I checked, it's up, stable and reporting
13:05:40 <sfinucan> Waiting on "a week of green" before we look to get voting rights
13:06:08 <moshele> sfinucan: I meant if you are planning to do multi node CI for NUMA
13:06:12 <lbeliveau> sfinucan: what has changed ?
13:06:36 <sfinucan> sfinucan: ah, yes
13:06:40 <moshele> the Mellanox CI multi node will test migration with direct port
13:06:51 <sfinucan> moshele: ah, yes (whoops)
13:07:15 <moshele> ok cool
13:07:22 <moshele> anything more on CI?
13:07:23 <sfinucan> lbeliveau: moshele: We were converting from custom scripts etc. to a CI based on upstream
13:07:40 <sfinucan> That's done now. Multi-node CI next, I guess
13:07:46 <sfinucan> ...once it's stable :)
13:07:59 <moshele> sfinucan: cool, thanks for the update
13:08:06 <lennyb> sfinican: upstream CI is quit stable
13:08:22 <lbeliveau> sfinican: do you know who's responsible for multi-node CI ?
13:08:24 <sfinucan> lennyb: I mean the Intel NFV CI
13:09:09 <sfinucan> lbeliveau: That would be wznoinsk
13:09:28 <moshele> can move to the next topic
13:09:32 <lbeliveau> sfinucan: ok thanks
13:09:54 <moshele> #topic bug fixes
13:10:36 <moshele> the list of patches under L 20 https://etherpad.openstack.org/p/sriov_meeting_agenda
13:11:00 <moshele> lbeliveau and I tested the migration patches
13:11:27 <moshele> I think for now they are good enough
13:11:57 <lbeliveau> I'll do more test for the race condition, I want to understand that a bit more
13:12:04 <moshele> so I will try to ping some of the nova cores, but sfinucan you are welcome to review them as well
13:12:26 <sfinucan> moshele: will do
13:12:58 <moshele> lbeliveau: agree, but I guess we can handle the race condition in different patch
13:13:05 <sfinucan> moshele: In addition, I've a few bugfixes coming around CPU pinning/NUMA topology. Can I stick them on that etherpad?
13:13:07 <lbeliveau> absolutely
13:13:17 <moshele> sfinucan: sure
13:13:29 <lbeliveau> sfinucan: yes, I'll look at them shortly
13:14:17 <moshele> anything else on bugs?
13:14:35 <lbeliveau> I'm good
13:14:49 <sfinucan> ditto
13:14:53 <moshele> #topic Specs for review
13:15:19 <saneax> we have one more sriov related spec for IRONIC
13:15:32 <saneax> https://review.openstack.org/#/c/331564/
13:15:49 <saneax> this is to push the grub parameters via IPA
13:15:55 <sfinucan> there's this one for PCI passthrough without NUMA too
13:15:56 <sfinucan> https://review.openstack.org/#/c/298179/
13:16:15 <moshele> saneax: this for support SR-IOV with  TriplO right
13:16:21 <saneax> yes
13:16:23 <sfinucan> (tl;dr: Don't do PCI passthrough/SR-IOV if you don't have a NUMA topology on the host)
13:17:10 <saneax> ok sfinucan
13:18:12 <karthiks> to support the spec related to ironic , we have raised a RFE with nova
13:18:14 <karthiks> https://bugzilla.redhat.com//show_bug.cgi?id=1348566
13:18:14 <openstack> bugzilla.redhat.com bug 1348566 in openstack-nova "[RFE] Provide a mechanism to configure boot params" [Unspecified,New] - Assigned to eglynn
13:18:50 <saneax> sfinucan, do you have more doc as why?
13:19:14 <moshele> saneax: I added the ironic spec
13:19:22 <moshele> saneax: to etherpad
13:19:27 <saneax> I amnot sure what breaks if NUMA topology is not present and we allow passthrough/SR-IOV
13:19:33 <saneax> thanks moshele
13:19:41 <saneax> request for review
13:20:58 <moshele> sfinucan: you are the NUMA expert can you review https://review.openstack.org/#/c/298179/
13:21:41 <sfinucan> saneax: No more than what's in the commit messag
13:21:55 <saneax> ok sfinucan, thanks
13:22:08 <sfinucan> though I think there might be something on the mailing listr
13:22:51 <sfinucan> saneax: Yup, search for 'PCI Based NUMA Scheduling can't really achieve pci numa binding in some cases'
13:23:11 <saneax> thanks sfinucan
13:23:40 <sfinucan> moshele: Don't know about "expert", but I'll check it out :)
13:24:01 <sfinucan> *I plan to check it
13:24:28 <moshele> sfinucan: cool
13:24:32 <saneax> sfinucan, got the mail thanks - bug https://bugs.launchpad.net/nova/+bug/1551504
13:24:32 <openstack> Launchpad bug 1551504 in OpenStack Compute (nova) "I/O (PCIe) Based NUMA Scheduling can't really achieve pci numa binding in some cases." [Medium,In progress] - Assigned to Jinquan Ni (ni-jinquan)
13:25:24 <moshele> moving to my spec User-controlled SR-IOV ports allocation - https://review.openstack.org/#/c/182242/
13:25:55 <moshele> sfinucan: so I need to address the NUMA part
13:26:51 <sfinucan> moshele: Yes, you do
13:27:01 <sfinucan> I've no idea how to address this myself, I'm afraid :(
13:27:35 <moshele> sfinucan: I first need to have good understanding of the problem
13:28:03 <moshele> sfinucan: so maybe we can talk after the meeting
13:28:08 <sfinucan> moshele: Sure thing
13:28:28 <moshele> sfinucan: cool
13:28:55 <moshele> we also have this spec  new spec: tripleo-sriov - https://review.openstack.org/#/c/313872
13:29:18 <moshele> I haven't review the new updates, I hope to do it this week
13:29:34 <moshele> anything more about SR-IOV specs?
13:30:08 <sfinucan> moshele: Not off the top of my head
13:30:17 <lbeliveau> moshele: good too
13:30:33 <moshele> #endmeeting