13:00:39 <baoli> #startmeeting PCI Passthrough 13:00:40 <openstack> Meeting started Tue Oct 14 13:00:39 2014 UTC and is due to finish in 60 minutes. The chair is baoli. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:41 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:43 <openstack> The meeting name has been set to 'pci_passthrough' 13:00:49 <baoli> Hi there 13:02:18 <irenab> hi 13:03:09 <baoli> irenab: regarding the review, https://review.openstack.org/#/c/126357/. 13:03:30 <baoli> any thing special with 00:00:00:00:00:01? 13:03:47 <irenab> baoli: I checked it for Mellanox NIC, it works 13:04:36 <baoli> irenab, it's great to hear about that. 13:04:45 <irenab> baoli: I jut do not like the mac address to stay without change 13:04:54 <irenab> once released 13:07:04 <irenab> baoli: can you please check if setting 00:,,,:01 works for Intel nic? 13:07:59 <baoli> irenab, it's not a problem with intel either. 13:08:43 <irenab> baoli: does it look reasonable to set this 'dummy' mac once VF is released? 13:09:50 <baoli> That's what I'm thinking about. 00::01 has no difference from any other valid mac addresses. 13:10:46 <irenab> baoli: I doubt if there would be such a prefix option defined for neutron as 00:00:00:00 13:11:42 <irenab> but any way since the VF is put to down, it is not a real problem to leave the previous mac 13:13:10 <baoli> That's what I was thinking aobut. What I can do though is to put comments like "Mac addr can't be flushed, so keep the previous mac" 13:14:17 <irenab> baoli: so you prefer not setting it to 'dummy' mac. I am Ok with it. 13:14:32 <baoli> I've also tested that It doesn't affect later assignment. 13:14:48 <irenab> baoli: yes, me too :-) 13:15:34 <baoli> irenab, when you say 'dummay' mac, are there some mac that are commonly designated as 'dummy' 13:16:24 <irenab> baoli: I do not know, I meant 00:...:01 or something like this, but maybe worth to check 13:16:53 <irenab> meanwhile, I think the comment is good enough 13:17:53 <baoli> irenab, cool. 13:18:07 <irenab> I opened 2 bugs for SR-IOV documentation 13:18:18 <irenab> https://bugs.launchpad.net/openstack-manuals/+bug/1381017 13:18:19 <uvirtbot> Launchpad bug 1381017 in openstack-manuals "PCI passthrough_whitelist configuration option documentation" [Undecided,New] 13:18:28 <irenab> https://bugs.launchpad.net/openstack-manuals/+bug/1381060 13:18:30 <uvirtbot> Launchpad bug 1381060 in openstack-manuals "Add SR-IOV networking solution documentation" [Undecided,New] 13:19:39 <baoli> that's cool. 13:20:17 <baoli> irenab, no documentation for PCI passthrough either. 13:21:10 <irenab> baoli: I start to get questions regardig SR-IOV networking, feels like something that really need to be documented 13:21:27 <baoli> irenab, I agree 13:21:48 <irenab> at least there is a wiki page 13:22:41 <baoli> We have a doc person in our team. Let me see if I can get some help. 13:23:33 <baoli> the wiki page is good. 13:23:43 <irenab> baoli: This can be great. 13:24:20 <irenab> I have a question regarding macvtap 13:24:25 <baoli> sure 13:24:37 <irenab> As far as I understand, currently it does not allow live migration, right? 13:24:51 <baoli> By they way, I made a few minor changes in your wiki, 13:25:08 <irenab> baoli: thanks, will take a look 13:25:32 <baoli> irenab, nova doesn't support it yet 13:25:48 <baoli> but libvirt does 13:26:18 <irenab> baoli: are you talking about libvirt network? 13:26:36 <baoli> irenab, yes 13:26:38 <irenab> is this your plan for kilo? 13:27:34 <baoli> We discussed three options earlier to support this. And I'm leaning towards it. 13:27:35 <irenab> with previous mellanox solution, we did some interface renaming, which also allowed live migration support, but I think the libvirt networks is cleaner solution 13:28:12 <irenab> baoli: cool, keep me in loop with your investigation 13:28:20 <baoli> I'm about to write the nova spec for it, and I'm planning to describe all the three options we have discussed 13:28:33 <baoli> absolutely 13:29:10 <baoli> the interface reanming was one of the options 13:29:34 <irenab> I still need to see how to set default network vnic type, will update once have basic understanding how to deal with this 13:30:15 <baoli> irenab, cool 13:31:02 <baoli> So I'll update the patch, and would you +1 it afterwards? 13:31:30 <irenab> sure 13:31:41 <baoli> irenab, great and thanks! 13:32:05 <baoli> irenab, regarding the devstack change your colleague is making 13:32:06 <irenab> no problem 13:32:23 <irenab> baoli: what about it? 13:32:43 <baoli> The devstack change to be able to configure the whitelist and alias have all been merged. 13:32:59 <baoli> So the patch doesn't have to do that any more. 13:33:13 <irenab> baoli: great news, I'll update him 13:33:35 <irenab> I will suggest him to talk with you on irc if he has questions, is it ok? 13:33:42 <baoli> I also made a couple of other comments, please check it out and see if they make sense. 13:33:48 <baoli> irenab, sure 13:34:03 <irenab> baoli: on the mcvtap patch or devstak? 13:34:31 <baoli> ireanb, the devstack one 13:35:09 <irenab> can you please share the gerrit link? 13:35:41 <baoli> https://review.openstack.org/#/c/126580/ 13:36:00 <irenab> thanks 13:36:30 <irenab> I do not have more to update or to discuss 13:37:03 <rpothier> irenab: I have a question 13:37:09 <irenab> sure 13:37:17 <rpothier> I am adding heat support, do you support security groups for sr-iov? 13:37:27 <irenab> no 13:37:52 <irenab> there are plans to support it in the future 13:37:59 <rpothier> so security groups can be restricted in the heat template? 13:38:14 <rpothier> ok, in that case, I won't restrict it 13:38:29 <irenab> rpothier: great, thanks 13:39:20 <baoli> Anything else? 13:39:36 <irenab> not for me 13:40:00 <baoli> Ok, thanks and next time 13:40:06 <baoli> #endmeeting