13:04:41 <moshele> #startmeeting sriov 13:04:42 <openstack> Meeting started Tue Nov 1 13:04:41 2016 UTC and is due to finish in 60 minutes. The chair is moshele. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:04:43 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:04:46 <openstack> The meeting name has been set to 'sriov' 13:04:47 <moshele> hi 13:05:01 <edand> hi 13:05:04 <wznoinsk> hi 13:05:45 <lennyb> hi 13:05:54 <moshele> let wait for some more pepole 13:07:09 <moshele> #topic Improving SR-IOV/PCI Passthrough CI 13:07:32 <moshele> so lennyb is woking on the SR-IOV multi-node CI 13:08:02 <moshele> we need to merged this tempest patches migration test - https://review.openstack.org/#/c/339230/ and migration revert test - https://review.openstack.org/#/c/348498/ 13:08:30 <moshele> wznoinsk: any update regarding NFV CI ? 13:08:33 <lennyb> yep, they pass manual check, I will comment on gerrit dhortly 13:08:52 <lbeliveau> hi, sorry I'm late 13:08:59 <wznoinsk> yes, I have PCI passthru CI in our lab sandboxing 13:09:26 <wznoinsk> (non-voting): samples - https://review.openstack.org/#/c/387485/4, https://review.openstack.org/#/c/387836/25 13:09:59 <moshele> ok cool 13:10:02 <wznoinsk> "tempest-dsbaremetal-pcipassthrough-xenial" job under http://ci-watch.tintri.com/project?project=nova 13:10:16 <moshele> when do you think it will be ready? 13:10:19 <lbeliveau> this is with multi nodes ? 13:10:41 <moshele> no it just same group managing all of intel CI, right? 13:10:49 <wznoinsk> this is single node, for now I've 1:1 moved from other team in Intel to our lab 13:11:41 <wznoinsk> multinode of PCI passthru CI (if at all) we may do it later, not sure about good testing scanario for this 13:12:18 <wznoinsk> would you have any sugestions? 13:12:24 <lbeliveau> migration is one 13:12:25 <moshele> how about mult-node nfv for NUMA migration/live migration ? 13:13:06 <moshele> yes once lbeliveau patches in tempest will be merged you can ran migration and migration revert 13:13:12 <wznoinsk> moshele, that's done, job: tempest-dsvm-multinode-ovsdpdk-nfv-networking-xenial (no pci passthru here if that's what you're after) 13:13:54 <wznoinsk> moshele, do you have a link again? 13:14:20 <moshele> wznoinsk: what link? 13:14:25 <moshele> for patches? 13:14:28 <wznoinsk> sorry, I think I have it https://review.openstack.org/#/c/348498/ 13:14:44 <lbeliveau> https://review.openstack.org/#/c/339230 13:14:49 <moshele> yes 13:14:53 <lbeliveau> https://review.openstack.org/#/c/348498 13:15:15 <wznoinsk> thanks 13:15:26 <lbeliveau> there was no cores on qa irc yesterday, need to ping them 13:15:36 <moshele> wznoinsk can you check if migration test passes for NUMA? and update us 13:16:14 <wznoinsk> moshele the job "tempest-dsvm-multinode-ovsdpdk-nfv-networking-xenial" has numa_nodes spec in the flavor and it's doing multinode live mgiration 13:17:09 <moshele> live migration is not the same as cold migration. It is different code path 13:17:38 <wznoinsk> http://intel-openstack-ci-logs.ovh/36/387836/25/check/tempest-dsvm-multinode-ovsdpdk-nfv-networking-xenial/3bd89ed/testr_results.html.gz -> tempest.api.compute.admin.test_live_migration (in green) 13:18:05 <moshele> we are looking for scenarios testing not api 13:18:14 <wznoinsk> moshele, yes, cold migration will be used for PCI passhtru multinode as live migration for the same not possible AFAICU 13:18:17 <moshele> not sure how well it is tested with api 13:18:42 <wznoinsk> moshele, sure, would you know if someone's working on such scenario test for PCI? 13:19:49 <wznoinsk> I take it as it's a task for me then ;-) 13:19:51 <moshele> so live migration can work with macvtap port (which is the same flow for pci device) 13:20:20 <moshele> but we know it is broken see live-mig: Add claims and proper resource tracking - https://review.openstack.org/#/c/244489/ 13:20:24 <wznoinsk> wouldn't macvtap be for PCI devices of type 200 only (network) tho? 13:21:04 <moshele> I didn't get the question 13:21:31 <wznoinsk> I thought macvtap is for networkg devices so more of a SRIOV thing, not general PCI device passthru one 13:22:24 <moshele> yes, but for the code point of you the share lots of common paths 13:22:51 <moshele> ok let move the the next topic 13:22:57 <wznoinsk> ok, I'll educate myself on that then, let's move on 13:23:17 <moshele> #topic Documentation 13:23:44 <moshele> lbeliveau: are you planning to work on the Docmement PF Passthrough for nova and neutron ? 13:24:03 <lbeliveau> yes I could 13:24:21 <moshele> ok cool 13:24:28 <moshele> let move on 13:24:41 <lbeliveau> we'll also have to revise the documentation regarding the neutron sriov conf that are being deprecated 13:24:43 <moshele> #topic Patches for Subteam Review 13:25:01 <moshele> see list of patches in https://etherpad.openstack.org/p/sriov_meeting_agenda Line 29 13:25:36 <moshele> I ask jaypipes to review pci: remove pci device from claims and allocations when freeing it- https://review.openstack.org/#/c/369180/ and he added it to his queue 13:25:51 <moshele> we will need to backport this one 13:26:14 <moshele> any other important patches? 13:26:59 <lbeliveau> this one: https://review.openstack.org/#/c/363884/ 13:27:06 <lbeliveau> lots of +1 13:27:29 <moshele> ok let me update the agneda 13:28:01 <moshele> SR-IOV neutron config depreciation/removal 13:28:08 <moshele> SR-IOV: Remove deprecated supported_pci_vendor_devs option https://review.openstack.org/#/c/391911/ 13:28:14 <moshele> Deprecate SR-IOV 'physical_device_mappings' config option - https://review.openstack.org/#/c/360447/ 13:28:37 <moshele> for the last want we need to send deprecation mail, will do that this week 13:29:02 <moshele> right edand? 13:29:09 <edand> yep 13:29:15 <moshele> ok so let move one 13:29:55 <moshele> #topic spec for review 13:30:13 <moshele> so we currently we have 3 spec 13:30:40 <moshele> see https://etherpad.openstack.org/p/sriov_meeting_agenda line 49 13:31:19 <moshele> and we have Add spec to enhance PCI passthrough whitelist to support regex is approved and merged 13:32:09 <moshele> I request jaypipes to review them. we need to merge them until the spec freeze in 17th this month 13:32:22 <moshele> anything else you want to talk about? 13:33:33 <moshele> #endmeeting