13:00:33 <baoli> #startmeeting PCI Passthrough
13:00:34 <openstack> Meeting started Tue Nov 11 13:00:33 2014 UTC and is due to finish in 60 minutes.  The chair is baoli. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:00:35 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:00:37 <openstack> The meeting name has been set to 'pci_passthrough'
13:00:46 <baoli> Hi
13:00:49 <irenab> Hi
13:01:02 <heyongli> hi, everyone
13:01:04 <sadasu> Hi
13:01:17 <irenab> baoli: thanks a lot for posting agenda for this meeting.
13:01:45 <baoli> irenab, hope it doesn't miss anything
13:02:29 <baoli> https://wiki.openstack.org/wiki/Meetings/Passthrough#Agenda_on_Nov._11th.2C_2014
13:03:34 <adrian-hoban> Hi Guys
13:03:49 <irenab> hi
13:04:06 <baoli> adrian-hoban: welcome to join the meeting
13:04:11 <pczesno> hi
13:04:44 <adrian-hoban> baoli: Thanks
13:04:51 <sadasu> Hello adrian-hoban, pczesno
13:04:56 <irenab> shall we start?
13:05:14 <baoli> yep
13:05:22 <sadasu> are we starting from the contents of the ether pad?
13:05:40 <baoli> I put the etherpad there as reference
13:05:49 <baoli> Let's do the bugs first
13:06:13 <sadasu> ok
13:06:22 <irenab> baoli: Ok, then lets discuss how to progress with etherpad items
13:07:34 <heyongli> nova-compute fails to start
13:08:02 <heyongli> patches sent, had comments to address, will release new patch soon
13:08:03 <irenab> heyongli: link to bug/review?
13:08:29 <heyongli> #link https://review.openstack.org/#/c/131321/
13:08:46 <irenab> heyongli: thanks
13:08:49 <baoli> heyongli, looks like one more step to get it approved.
13:09:23 <heyongli> sure, it's likely easy to fix
13:09:28 <heyongli> the comments
13:09:51 <irenab> heyongli: folowing summit concerns, are you testing on  HW as well?
13:10:12 <heyongli> definitely
13:10:19 <heyongli> all hw env
13:10:25 <irenab> heyongli: great, thanks
13:10:30 <baoli> heyongli, combining your patch with https://review.openstack.org/#/c/117042/ will likely fix the pci init issue.
13:10:49 <heyongli> this need rebase
13:10:55 <heyongli> will release soon also
13:11:55 <baoli> I listed two review from Yunhong Jiang at the end https://wiki.openstack.org/wiki/Meetings/Passthrough#Agenda_on_Nov._11th.2C_2014
13:12:13 <heyongli> another bug resize with PCI devices doesn't work
13:12:22 <heyongli> https://bugs.launchpad.net/nova/+bug/1368201
13:12:23 <uvirtbot> Launchpad bug 1368201 in nova "resize with PCI devices doesn't work" [Low,Confirmed]
13:12:45 <heyongli> also got progressed, will come up to a patch in few days
13:13:21 <irenab> baoli: is there any bug without owner?
13:14:39 <baoli> let's have a conversation one at a time. heyongli, what did you find out?
13:15:36 <baoli> irenab: the last one on the list doesn't seem to have an owner, and it's an old bug.
13:15:37 <heyongli> which one?
13:16:06 <irenab> https://bugs.launchpad.net/nova/+bug/1333498
13:16:07 <uvirtbot> Launchpad bug 1333498 in nova "table nova.pci_devices  lost device status every time. && PciDeviceList.get_by_compute_node pass a wrong parameter" [Low,Confirmed]
13:16:15 <heyongli> i assign this bug to me https://bugs.launchpad.net/nova/+bug/1333498
13:16:17 <baoli> heyongli, reszie
13:16:34 <heyongli> resize had several  problem
13:17:24 <heyongli> first one is the prefix _new process code been killed, i re open that in new style, then got pci resiza allocation work,
13:18:08 <baoli> heyongli, yeah, that's a known issue. So you are debugging with Juno code.
13:18:16 <heyongli> second one is when migration, the pci device free on old compute node had problem, i also fond a way to solve the problem
13:18:53 <heyongli> yeah, i always work with new code -:)
13:19:21 <irenab> there also another bug, seems old, withour owner: https://bugs.launchpad.net/nova/+bug/1261256
13:19:22 <uvirtbot> Launchpad bug 1261256 in nova "Cannot assign 2 same type PCI passthrough devices to 2 instances separately" [Undecided,Incomplete]
13:19:52 <heyongli> irenab, that's one is single tigger, not reproduced yet
13:20:22 <heyongli> Requested operation is not valid: PCI device 0000:04:00.0 is in use by domain instance-00000010\n']
13:20:32 <heyongli> and seems configration error
13:21:07 <irenab> heyongli: ok then, not a bug
13:21:22 <heyongli> because libvirt found the pci device used by another domain. sorry, might not a configration error, but before reproduced, nothing can do with it
13:22:00 <heyongli> back to resize
13:22:37 <heyongli> the update from migration need changed to corretly handle the resize,
13:23:09 <heyongli> all those change will solid together to form a patch to fix resize bug
13:23:34 <baoli> heyongli, that sounds good. Can't wait to try it.
13:23:55 <heyongli> this is the live migiration base
13:23:57 <baoli> heyongli, the bug also says after resize, the instance becomes inaccessible.
13:23:59 <heyongli> sorry for this
13:24:19 <heyongli> baoli: no, i can access the instance after resize
13:24:30 <baoli> heyongli, that sounds great
13:24:41 <baoli> heyongli, looking forward to your patch
13:24:50 <heyongli> if new patches released, and you still can not access the instance, we then dicuss it with detail
13:24:50 <irenab> heyngli: please add me to the review as you upload the patch
13:25:01 <heyongli> sure, irenab
13:25:31 <baoli> back to this one then: https://bugs.launchpad.net/nova/+bug/1333498
13:25:32 <heyongli> i got 3 bug on queue, seems all progress well now
13:25:32 <uvirtbot> Launchpad bug 1333498 in nova "table nova.pci_devices  lost device status every time. && PciDeviceList.get_by_compute_node pass a wrong parameter" [Low,Confirmed]
13:26:16 <baoli> so this looks like the compute restart bug.
13:26:18 <heyongli> baoli, this patch need to be rebase and retesting, few days needed
13:26:27 <heyongli> yeah, compute node problem
13:26:44 <heyongli> this because 2 reason
13:26:51 <heyongli> firs one is we use node id
13:27:38 <heyongli> second one is the pci devices need discovery, this is not like memory and vcpu, we can not reset the resource in evevery round of updating resource
13:28:45 <baoli> heyongli, I think that we have several bugs reported on this issue already, and there are several patches out of there dealing with this.
13:29:05 <heyongli> i prefer the patch yunhong released
13:29:26 <heyongli> but we can discuss with mail
13:29:48 <irenab> heyongli: there is no patch linked to latest bug
13:30:12 <heyongli> witch bug?
13:30:29 <heyongli> https://review.openstack.org/#/c/102298/
13:30:46 <baoli> I had yunhong's patches listed on the wiki as well, but I'd like to encourage looking at this one as well:https://review.openstack.org/#/c/117042/
13:30:57 <heyongli> sure
13:31:58 <irenab> heyongli: found it, it seems to be in the merge conflict.
13:32:16 <heyongli> irenab, yeah, no time solt for this now
13:32:35 <heyongli> baoli, Sylvain Bauza's patch is great
13:33:03 <heyongli> i will spend time this week to reivew and make a choice
13:33:18 <baoli> heyongli: sounds great
13:33:40 <irenab> anything else on nova bugs?
13:34:22 <heyongli> baoli, how about the pop problem about whitelist?
13:35:23 <baoli> heyongli, a patch is already available: https://review.openstack.org/#/c/130833/
13:36:00 <heyongli> got, on my review list -:)
13:36:53 <irenab> baoli: there is also a patch (no unit tests yet) for white lists
13:37:24 <baoli> irenab, the first patch listed on the wiki?
13:37:25 <irenab> https://review.openstack.org/#/c/133138/
13:37:32 <irenab> naoili: yes
13:38:02 <irenab> are you ok with this approach?
13:38:08 <heyongli> got but no time slot for this now ...
13:39:19 <heyongli> i hope i can review all the patches in this week
13:39:22 <baoli> irenab: I think that I'm fine with the approach. Although I don't think it's necessary to provide the sections options
13:39:55 <irenab> baoli: so no yaml file, config stays in nova.conf?
13:40:15 <baoli> irenab, that seems to be Ian's choice as well
13:40:48 <irenab> this looks fine for me as well
13:41:12 <baoli> irenab, it's straightforward. yaml is straightforward too, but just another format to look at.
13:41:57 <baoli> I'd encourage everyone to review the patch.
13:42:00 <heyongli> i heared of voice to improve whitlist, this might be a way to do
13:42:35 <irenab> baoli: agree, just need to see that this is aligned with what came out of ml discussion
13:42:37 <baoli> Now that we know how to use local.conf to configure with the existing format, it's not in a rush to get the new format in.
13:43:37 <irenab> baoli: if you have an idea how to simplify Ian suggestion, please comment on his patch
13:44:00 <baoli> irenab, I commented it on patch 1
13:44:17 <irenab> we have 15 mins left, are we ok to move on?
13:44:28 <baoli> irenab, the ML discussion discussed two choices.
13:45:09 <baoli> irenab, so we'll see which one will win the most votes in the end
13:45:22 <irenab> baoli: :-), lets have both
13:45:24 <baoli> irenab, let's move on.
13:46:07 <irenab> baoli: any update on live migration bp?
13:46:49 <baoli> I'm trying to find time working on the specs. I got it started, but stopped with various other things. I'm hoping to get the initial version soon.
13:47:18 <heyongli> same situation to me for the sriov hotplug
13:47:56 <adrian-hoban> baoli: I'm interested to review that once ready. Will it focus on macvtap only at this time?
13:48:14 <baoli> adrian-hoban, yes, macvtap only.
13:48:49 <irenab> as for SR-IOV for InfiniBand, the spec for VIF driver is under review. This is quite small code change to support IB vif_type
13:49:26 <irenab> the rest of the nova support does not require any change
13:49:46 <heyongli> irenab, neutron will check the vender&product to support sriov?
13:50:08 <heyongli> if new device support it neutron need any change?
13:50:57 <irenab> heyongli: not neutron code, just config option for neutron server
13:51:09 <heyongli> ok, got it
13:51:11 <heyongli> thanks
13:51:40 <irenab> How do we want to discuss the etherpad items?
13:52:16 <heyongli> pci stats object model is own by yunhong now, seems in the track
13:52:34 <baoli> I have his patch listed on the wiki
13:53:24 <irenab> back to etherpad, anyone had progress with some item?
13:54:10 <pczesno> on number 9
13:54:28 <pczesno> not much progress yet , but we are working on the spec and libvirt patch
13:54:42 <pczesno> libvirt guys intrsted in this functionality
13:55:13 <pczesno> https://www.redhat.com/archives/libvir-list/2014-October/msg00606.html
13:55:53 <irenab> pczesno: Thanks! Do you have spec under review?
13:56:08 <pczesno> not yet, but should be posted this week
13:56:52 <baoli> pczesno: this looks interesting.
13:56:59 <irenab> We also had talk with Sandhya during the summit regarding collabotation on SR-IOV CI, lets discuss it next week
13:57:02 <heyongli> interesting
13:57:48 <irenab> adrian: The CI effort  maybe be relevant for NFV as well
13:58:50 <sadasu> the CI folks are also starting to attend this meeting
13:59:15 <heyongli> great news
13:59:51 <baoli> I will add CI to the meeting agenda
13:59:57 <sadasu> baoli: thanks
14:00:08 <irenab> baoli: thanks
14:00:21 <heyongli> time is up
14:00:30 <sadasu> don't have an update for this meeting, but will have something starting from next meeting
14:00:42 <heyongli> let's discuss next meeting
14:00:55 <adrian-hoban> irenab: Yes, very much so.
14:00:56 <sadasu> working on changes to my kilo spec after some questions being answered during summit
14:00:57 <baoli> it's time. it's good to see that we have more folks joining the meeting
14:01:13 <irenab> great, I think having agenda was very hepful. Thnaks a lot, baoli
14:01:14 <sadasu> will post it for review today…would appreciate reviews :-)
14:01:35 <irenab> sadsu: add me as reviewer
14:01:36 <baoli> ok, next time
14:01:40 <baoli> #endmeeting