13:00:44 <moshele> #startmeeting sriov
13:00:44 <openstack> Meeting started Tue May  3 13:00:44 2016 UTC and is due to finish in 60 minutes.  The chair is moshele. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:00:45 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:00:48 <openstack> The meeting name has been set to 'sriov'
13:00:51 <moshele> hi all
13:01:00 <lennyb> moshele, hi
13:01:08 <efried_> Howdy.
13:01:50 <moshele> let wait for more people to join
13:01:56 <sfinucan> o/
13:02:17 <lbeliveau> hi
13:02:56 <moshele> ok let start
13:03:04 <moshele> here is that agenda https://etherpad.openstack.org/p/sriov_meeting_agenda
13:03:49 <lbeliveau> I did some cleanups last couple of days, let me add them in the etherpad
13:03:52 <moshele> currently I am still working cleaning up the pci code with jaypipes and also fixing the resize
13:05:59 <moshele> ok thanks lbeliveau
13:06:20 <sfinucan> If we're doing updates, I've started working on docs for general high-performance (NFV?) features, including SR-IOV and PCI passthrough
13:06:49 <lbeliveau> sfinucan: do you have a draft/review ready ?
13:06:59 <moshele> sfinucan: ok do you have patches or is it a wiki?
13:07:21 <sfinucan> lbeliveau: not even close yet, but I'll add you to the review when I do
13:07:29 <sfinucan> moshele: patches (nova-docs)
13:07:47 <moshele> sfinucan: can you add them  to the etherpad ?
13:08:01 <sfinucan> Will do, once I publish them (tomorrow'ish)
13:08:11 <moshele> cool
13:08:13 <lbeliveau> perfect, can I help ?  I know this area pretty well ?
13:08:53 <moshele> and I can help with the SR-IOV PCI passthrough  part
13:08:53 <sfinucan> lbeliveau: Yes, please. I want to cover the PCI stuff, but I've started with the stuff I know best - NUMA and pinning
13:09:20 <sfinucan> lbeliveau: Actually, I think you signed up to help at the summit :)
13:09:21 <lbeliveau> ok, I can have a stab at pci if you want
13:09:25 <sfinucan> Sure thing
13:09:29 <lbeliveau> yeah I did
13:09:34 <lbeliveau> ok I'll get started
13:10:21 <lbeliveau> so that you know, don't want to step on your toes :)
13:11:02 <sfinucan> lbeliveau: Sure thing, heh. Would like you insight into where they should go too. I'm making additional changes to the extra-specs docs at the moment, but I'm thinking a dedicated NFV section would be useful
13:11:30 <moshele> I think so too
13:11:31 <sfinucan> #link http://docs.openstack.org/admin-guide/compute-flavors.html
13:11:44 <lbeliveau> sfinican: I'll have a look after this meeting, will ping you
13:12:17 <sfinucan> Sweet
13:12:21 <sfinucan> Sorry for sidetracking, moshele. Back to the agenda :)
13:12:49 <moshele> it fine I will add a Doc section to see we are making progress
13:13:59 <moshele> so I hope to finish the resize fix by the end of this week
13:14:19 <lbeliveau> that would be sweet
13:14:20 <moshele> and then I will need lbeliveau help with the colde migration
13:14:26 <lbeliveau> sure thing
13:14:31 <moshele> ok cool
13:15:07 <lbeliveau> any news on CI ?
13:15:45 <lbeliveau> I sent an email to smooney to get insight of his plan to support multi-node testing, no answer yet
13:15:46 <moshele> currently we we adding more test to the Mellanox SR-IOV CI
13:16:17 <moshele> but I don't no what is the status with the intel NFV CI
13:16:24 <lbeliveau> moshele, those tests are not public right ?  meaning not in public git ?
13:16:43 <sfinucan> moshele: I can provide one, if it would help?
13:16:44 <moshele> public in git they are in tempest
13:17:11 <moshele> sfinucan: that would be great
13:17:40 <lbeliveau> ok, how does it work with tempest in terms of reviewing, nova cores also review there ?
13:17:40 <sfinucan> moshele: Sure
13:18:08 <efried_> Who is responsible for owning the CI?  Since it requires some rather expensive hardware.
13:18:08 <lennyb> sfinucan: are those tests somewhere on github?
13:18:14 <sfinucan> So we've been working on migrating everything to the upstream model (nodepool, zuul, etc.)
13:18:15 <moshele> sfinucan: also can you check if you CI can test pci-passthrough resize
13:18:34 <sfinucan> That's now complete, and our CI should be switching over this week
13:18:58 <moshele> is this just for the NUMA testing?
13:19:46 <sfinucan> NUMA, hugepages and some pinning
13:19:50 <lbeliveau> sfinucan: what is the name of this CI ?  It is run in the gate ?
13:20:04 <sfinucan> We run the standard tempest tests with different flavors first
13:20:20 <sfinucan> (hugepages enabled, NUMA topology specified and possibly some pinning)
13:20:34 <sfinucan> Then we run our custom tempest tests, which can be found here
13:20:42 <sfinucan> #link https://github.com/openstack/intel-nfv-ci-tests/
13:21:14 <sfinucan> lennyb: ^^ is that what you're looking for?
13:21:17 <lbeliveau> is there a plan to test multi-node ?
13:21:52 <sfinucan> lbeliveau: Is that addressed at me?
13:21:56 * lennyb checking link
13:22:04 <lbeliveau> sfinican: yes
13:22:12 <moshele> for mellanox CI is something that we need to investigate
13:22:56 <sfinucan> In that case, I don't know tbh. Despite suggestions at the summit, wznoinsk (rather than smooney) is the best person to ask
13:22:58 <lbeliveau> for cold migration we need at least two nodes (or maybe a chroot env)
13:23:38 <sfinucan> I can ask him to join next week, if that would help?
13:24:01 <moshele> sfinucan: yes
13:24:17 <sfinucan> moshele: pci-passthrough is probably something that the Intel PCI CI should/does test
13:24:45 <sfinucan> I don't have much information on that at the moment, other than the fact that we (Intel Shannon) should be taking ownership of it shortly
13:24:50 <lbeliveau> sfinican: yes, we absolutely need that, cores are resistant merge some of these patchs since it's not CI tested at the moment
13:25:12 <sfinucan> Again wznoinsk is the guy to ask RE: specifics
13:25:38 <wznoinsk> hi all
13:25:51 <moshele> wznoinsk: hi
13:25:57 <lbeliveau> wznoinsk: hi
13:26:03 <efried_> Will tests automatically filter out if required hardware is not present?  Or based on a setting in tempest.conf?
13:26:51 <wznoinsk> moshele we use SRIOV VF in our NFV CI at the moment, but the proper testing of SRIOV (like resize) would be done in the Intel SRIOV CI, https://wiki.openstack.org/wiki/ThirdPartySystems/Intel-SRIOV-CI
13:27:41 <sfinucan> efried_: Again, is that directed at me? :)
13:28:35 <moshele> wznoinsk: Mellanox CI is testing SR-IOV ports and it working fine, I just want to undestand what is the status of the pci-passthrough  testing
13:28:45 <wznoinsk> I see things coming from different angles, could we throw it into an etherpad... a wishlist about SRIOV testing?
13:29:23 <wznoinsk> moshele I'd have to talk to the guys from the SRIOV CI above on that, they're in PRC hence e-mail is probably best to catchup
13:30:12 <moshele> wznoinsk: ok also I would like to talk to after the meeting if that is possible
13:30:23 <wznoinsk> moshele sure
13:30:31 <moshele> let move on
13:31:36 <moshele> we have some bugs around the PF passthrough which Nikola work on, do you know if anyone continue the work
13:32:49 <lbeliveau> I guess we can share the work, do we have a list of those ?  Some are most likely PCI related and other numa
13:33:21 <moshele> lbeliveau: I put is on the ehterpad under PF passthrough
13:33:24 <sfinucan> moshele: NUMA and pinning work would be preferred, but I can look into slotting this into my Newton backlog if necessary
13:33:28 <lbeliveau> I think we should put down a list and add it to the etherpad for people to pick up this work
13:34:41 <sfinucan> lbeliveau: I think mriedem or johnthetubaguy has one. I'll put it in the etherpad if I find one
13:35:36 <lbeliveau> I can take over some of them (not all)
13:35:50 <moshele> I guess we will split the work :)
13:36:26 <lbeliveau> sure :)
13:36:26 <sfinucan> ^ yes, please :)
13:36:26 <moshele> let move to the NUMA and pinning
13:37:01 <moshele> so I am not expert on that so sfinucan you can update
13:37:30 <sfinucan> lbeliveau: johnthetubaguy doesn't know about any specific doc, but the notes from the summit would be a good start
13:37:32 <sfinucan> #link https://etherpad.openstack.org/p/newton-nova-performance-vms
13:37:35 <sfinucan> moshele: Sure
13:37:49 <lbeliveau> thanks
13:38:05 <lbeliveau> I'll start a list and put it in the etherpad
13:38:06 <johnthetubaguy> I want to get this all into the feature classification matrix
13:38:19 <johnthetubaguy> so we are all clear about the current docs, testing and implementation state
13:38:29 <sfinucan> So around NUMA and pinning, there a few ongoing things
13:38:29 <johnthetubaguy> but thats very much a work in progress at this point
13:39:15 <johnthetubaguy> honestly, as a sub group, if you can agree a list of "issues" where we need help, then narrow that down to a list of newton targets, that would be awesome
13:39:23 <johnthetubaguy> then send a summary to the ML
13:39:32 <johnthetubaguy> at least, thats would I would recommend you focus on
13:40:08 <sfinucan> For pinning, there are two bugs with pinning that need to be resolved and backported (they're listed in the agenda)
13:40:51 <sfinucan> We also need more work on documenting this feature for people (ongoing,sfinucan) and some additional functional tests (tbd)
13:40:54 <moshele> johnthetubaguy: so the agenda etherpad covers all the current issues https://etherpad.openstack.org/p/sriov_meeting_agenda
13:41:38 <sfinucan> For NUMA, the only big issues I know of are live migration. I _think_ ndipanov resolved most of these, but tbh I don't know what got merged in the end vs. what didn't
13:42:23 <moshele> sfinucan: do you have the bug at least?
13:42:36 <sfinucan> moshele: One moment
13:44:32 <sfinucan> So there are two closely related bugs
13:44:36 <sfinucan> #link https://bugs.launchpad.net/nova/+bug/1417667
13:44:37 <openstack> Launchpad bug 1417667 in OpenStack Compute (nova) "migration/evacuation/rebuild/resize of instance with dedicated cpus needs to recalculate cpus on destination" [Medium,In progress] - Assigned to Nikola Đipanov (ndipanov)
13:44:48 <sfinucan> #link https://bugs.launchpad.net/nova/+bug/1289064
13:44:49 <openstack> Launchpad bug 1289064 in OpenStack Compute (nova) "live migration of instance should claim resources on target compute node" [Medium,In progress] - Assigned to Nikola Đipanov (ndipanov)
13:45:38 <sfinucan> But tbh, I don't know the status of these now so it might be best to leave this discussion until next week
13:46:11 <sfinucan> ...when I can come back with a full report and we can decide who to take over this stuff (if anyone)
13:46:47 <moshele> sfinucan: ok, by the way the meeting is biweekly do you want to do it every week?
13:48:40 <moshele> anyone else? do you want to change it to weekly meeting?
13:49:04 <lbeliveau> every week is good with me as well
13:49:23 <sfinucan> moshele: Yes, every week would be good
13:49:27 <lbeliveau> there are no many tasks that need to be tracked that every week would make sense
13:49:50 <moshele> ok I will update the irc-meeting
13:49:54 <sfinucan> ...at least while getting these priorities in order
13:50:05 <moshele> agree
13:50:30 <lbeliveau> when PCI (and numa) gets more stable I guess it will become natural to do this every 2nd week
13:51:26 <sfinucan> agreed
13:52:10 <moshele> So I will reorganize the https://etherpad.openstack.org/p/sriov_meeting_agenda and send an update on the ML
13:52:29 <moshele> is that cool
13:53:18 <lbeliveau> perfect
13:53:43 <moshele> anything else?
13:54:03 <sfinucan> sounds great
13:54:17 <sfinucan> Bar getting eyes on those pinning bugs, nothing else from me
13:54:42 <lbeliveau> good with me
13:55:02 <moshele> ok cool, that it, thanks everyone for joining the meeting
13:55:08 <moshele> see you next week :)
13:55:12 <sfinucan> ttfn
13:55:12 <moshele> #endmeeting