13:04:46 <baoli> #startmeeting PCI Passthrough
13:04:47 <openstack> Meeting started Tue Mar  4 13:04:46 2014 UTC and is due to finish in 60 minutes.  The chair is baoli. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:04:48 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:04:50 <openstack> The meeting name has been set to 'pci_passthrough'
13:04:56 <baoli> Hi
13:05:01 <sadasu> what generic pci md?
13:05:02 <irenab> hi
13:05:23 <baoli> sorry for being late. My neighbor happens to be at my doorstep
13:05:25 <irenab> sadasu: the one you call SriovMechanismDriverBase
13:05:31 <sadasu> ok :-)
13:05:48 <sadasu> same here...but I have time now, and I was going to volunteer
13:05:54 <irenab> baoli: thank you for putting info on wiki
13:06:01 <baoli> irenab, np
13:06:02 <sadasu> do u already have BP for that...if so sorry I missed it
13:06:31 <baoli> Are you guys talking about the MD Base for SRIOV?
13:06:38 <sadasu> baoli: yes
13:06:44 <irenab> I have this one ml2-sriov-nic-switch
13:06:52 <irenab> and thought to do it as part of it
13:07:22 <sadasu> ok..
13:07:23 <irenab> we have bp opened by baoli, it can land there if need some neutral pace :-)
13:08:00 <sadasu> is ml2-sriov-nic-switch originally intended for your mech driver?
13:08:14 <irenab> https://blueprints.launchpad.net/neutron/+spec/pci-passthrough-sriov
13:08:15 <baoli> irenab, I think that it deserves a separate BP. Yours and Sandy's are focusing on the vendor MD.
13:08:42 <irenab> baoli: agree, it can be implemented on the one I posted before
13:09:00 <baoli> irenab, that's great.
13:09:19 <irenab> sadasu: yes, it is intended for next gen of our solution. The original Mlnx_MD hopefully will be merged in comming days
13:09:45 <sadasu> irenab, ok got it
13:10:18 <sadasu> baoli's existing BP seems to be a better place
13:10:23 <irenab> sadasu: I think I have in mind what is needed, we can chat if you like
13:10:45 <irenab> so agreed, doing it on baoli's bp
13:11:01 <sadasu> yes, I had sent email to list after last meeting, do we continue there or another IRC?
13:11:51 <irenab> sadasu: seems that we are the only one who answer there, so we can chat on neutron channel incase there will be more interest and maybe send update on the list
13:12:09 <sadasu> irenab, sounds good
13:12:33 <heyongli> hi baoli,  how about yours nova side work?
13:12:43 <baoli> heyongli, Hi
13:13:24 <baoli> Yongli, Did you see my response to your comments?
13:13:30 <heyongli> yeah
13:14:44 <heyongli> i hope you try another patch set as base, that's work,  maybe the interface to sriov part is little diffrent
13:15:16 <baoli> Do you mean your patch?
13:15:31 <heyongli> or, you can just drop my last patch and use you interface implement
13:16:36 <irenab> baoli, heyongli: I wanted to ask general question
13:16:44 <heyongli> please
13:16:53 <baoli> irenab, go ahead
13:17:29 <irenab> baoli, seems you are pushing all in one commit changes and heyongly pushes many small commits
13:17:57 <irenab> what is the best handled by nova cores?
13:18:19 <irenab> any reviews done?
13:18:22 <heyongli> small is good, for easy review
13:18:40 <heyongli> what do you mean/
13:19:03 <irenab> currently we need this code to be accepted, wa tis the best way to push it
13:19:50 <heyongli> smaller patch make it easy for review to catch up our idea i think
13:19:57 <irenab> alos need to be sure work is done, not duplicated and no conflicts
13:20:14 <heyongli> yeah, i also worry this
13:20:38 <baoli> Irenab, please see the response I put in the bug review: https://review.openstack.org/#/c/67500/
13:20:51 <heyongli> now baoli given all in one solution had some duplicated function
13:21:24 <heyongli> https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/pci-extra-info,n,z
13:21:37 <heyongli> yunhong and my patch set is here
13:22:37 <irenab> Ok. maybe just need to update on wiki what patches are required
13:22:46 <heyongli> baoli, i think  you can depnd on our patch except the "pci request mark interface", your patch is work , that's cool
13:23:04 <baoli> yongli, I'll take a look at your patch again once I have a chance. The duplication is minimal. And as I have indicated in my response, the two routines are used in different places
13:23:33 <irenab> I should get equipment in comming days and want to try all the patches together, just want to be sure what to apply
13:23:37 <heyongli> i know , if you want merge to nova ,then the match spec is dup
13:24:47 <sadasu> you mean https://review.openstack.org/#/c/57859/?
13:24:52 <sadasu> dup of what?
13:25:23 <irenab> baoli: is there anything that prevent further work on your patch?
13:25:24 <heyongli> not 57859
13:26:24 <baoli> Yongli, I gave comments to  your patch a while back and didn't see response.
13:26:33 <heyongli> i mean baoli's  nova side patch set for basic sriov suport, https://review.openstack.org/#/c/67500/
13:27:35 <heyongli> baoli, i'm sorry, i check it but which patch?
13:27:44 <baoli> irenab, can you clarify your question?
13:27:59 <baoli> yongli, the same patch
13:28:17 <baoli> Let me repeat what I did in my patch:
13:28:22 <heyongli> to my patch set?
13:28:29 <baoli> heyongli, yes
13:28:41 <baoli> So give me one minute
13:28:51 <irenab> baoli: wanted to ask if there is some disagreements of nova guys on your patch, I see it in your response comment
13:29:15 <irenab> "But it also depends on the community consensus on how it should be implemented eventually"
13:29:28 <baoli> Irenab, as you know, we didn't actually reach agreement.
13:29:57 <heyongli> 
13:29:58 <heyongli> Baodong (Robert) Li		Feb 7
13:29:58 <heyongli> Please specify how this is done. In this wiki:https://wiki.openstack.org/wiki/PCI_passthrough_SRIOV_support_Icehouse, it indicates that the existing PCI alias will be extended. For neutron SRIOV, we need to tag networking sriov ports with the attribute "net-group" (or whatever name we can agree on), and stats report for these ports is based on that.
13:29:58 <heyongli> This patch doesn't seem to reflect the wiki and support the neutron sriov requirement.
13:30:23 <irenab> baoli: so the question is, how to proceed here
13:30:40 <heyongli> your concern of this  is solved by https://review.openstack.org/#/c/74633/
13:31:38 <baoli> So this patch is based on the wiki https://wiki.openstack.org/wiki/Nova-neutron-sriov that we have discussed and agreed, and a few things that I have asked to Yongli and Yunhong, plus some fixes to the existing code
13:31:53 <heyongli> yunhong write this, i foget to reply you done
13:32:26 <baoli> And I think that you and sadasu would need something soon. That's why I put all them together
13:33:02 <irenab> baoli: agree and appreciate, but this is to POC
13:33:26 <irenab> what about production version and plan for Juno
13:33:29 <irenab> ?
13:34:11 <baoli> Irenab, as you have seen from the meeting wiki, I put the recap of discussion over there
13:34:33 <baoli> So what's for juno is something for discussion
13:34:53 <baoli> As you know, the nova side of things are not approved for Icehouse
13:35:02 <irenab> baoli: It is very helpfull. What wil be the next step once we review and add/comment?
13:35:35 <irenab> I think we need to make it some sort of proposal with options, share with community. What do you think?
13:36:17 <baoli> Irenab, I'd recommend that you use the patch for testing because it works
13:36:49 <baoli> As for the next step, we said that we are going to plan for a session in the summit
13:36:59 <irenab> baoli: thanks, will  do so
13:37:21 <irenab> I just thought we already doing it :-)
13:38:04 <baoli> Irenab, any comments about the Recap I put in the wiki?
13:38:35 <irenab> need to read it more carefully, will send comments later
13:39:02 <baoli> Irenab, thanks
13:39:10 <irenab> I saw some mails on neutron flavors for services
13:39:46 <irenab> I think we probably need to initiate something similar for tenant/admin API for NIC related options
13:40:06 <baoli> Irenab, what's the subject of the email?
13:40:29 <irenab> baoli: Flavor Framework
13:41:09 <irenab> it sis related to neutron services, but just see it as example of something that starts the discussion on ML in advance
13:41:33 <baoli> cool, I'll take a close look at it today.
13:42:42 <irenab> I think we need to reach some model we agree and propose it
13:42:55 <heyongli> baoli , i just read your recap, i might wrong, but PCI Group is just a extra info tag as we discussed.
13:43:31 <irenab> baoli: by the way, need to update meeting time to  Tue
13:44:10 <heyongli> and i don't remember a pci stats group concept,  do i missing something?
13:44:35 <baoli> Yongli, I think that the difference is not just one against many
13:44:42 <irenab> baoli: any updates from beagles on documentation?
13:45:19 <baoli> yongli, pci stats group is there, right? how do you come up with an stats entry?
13:45:37 <heyongli> are you mean the pool?
13:45:38 <baoli> irenab, I didn't see anything from beagle, I thought he'd show up today
13:45:48 <heyongli> define by the pci_flavor_attr
13:46:04 <baoli> yongli, yes.
13:46:33 <irenab> seems people continue to come and go...
13:47:06 <heyongli> do you mind i update the nova side description a litte bit?,
13:47:18 <heyongli> or i can put to dev mail first
13:47:44 <heyongli> i try to get a slide to show design choice and why
13:47:57 <irenab> hayongli: It can be great
13:48:00 <baoli> Yongli, you can come up with a paragraph and put a link in to the wiki
13:48:14 <heyongli> maybe usefull to get some core sponsor.
13:48:40 <heyongli> baoli: if i do that , we had 2 desciption about same thing
13:49:09 <baoli> yongli, if it's the same, then no need. But it could be useful to see what' exactly in your mind
13:49:41 <heyongli> baoli, i think it should be same as we discussed so long time
13:49:53 <irenab> guys, we need one place to put all info and communicate to ML. I guess on Summit we need to present and get core's sponsorship
13:49:56 <heyongli> i try to put it to mail first
13:50:10 <baoli> yongli, cool.
13:50:29 <baoli> Irenab, agreed
13:51:11 <heyongli> just put it to the recap?
13:51:56 <baoli> Yongli, I didnt' do full recap on the PCI Flavor since the wiki has more detailed info. So you may update that section to give a summarized good description on the key things in PCI flavor
13:52:53 <irenab> Do we want to propose some action items for next week?
13:53:03 <heyongli> i worry what should be put to the recap section, let's just do it. and every one can review it and we can agree to a identical description.
13:53:25 <irenab> please send email once wiki updated
13:53:38 <baoli> Yongli, I had a link in the PCI flavor section to Ian's wiki
13:53:44 <heyongli> i will just send mail at fist
13:54:28 <baoli> yongli, ok, let's start with that
13:54:50 <irenab> I think we need cover Top Down picture from user(tenant/admin) perspective before drill down to technical side
13:55:23 <baoli> irenab, sure. I also think that we should collect the use cases
13:55:46 <irenab> baoli: agree
13:56:05 <irenab> Shall we make it agenda for next meeting? After we sync on ongoing patches
13:56:09 <heyongli> in my bp there had some, but i wonder is that we want?
13:56:34 <baoli> irenab, sure.
13:58:29 <baoli> I can start a agenda section in the wiki, and you guys can put whatever you want to talk about in that section before next meeting. How does that sound?
13:58:40 <heyongli> cool
13:58:59 <irenab> great
13:59:10 <baoli> Do you know what's the deadline for proposing summit session?
13:59:50 <baoli> I'll check it out
14:00:04 <irenab> not sure, but I think ~week before
14:00:05 <heyongli> no idea, i try to setup one, but not find where to do that, seems no invite code sent out
14:00:26 <irenab> I think its not opne yet for design summit
14:00:26 <baoli> Thank you guys. time is up for this meeting
14:00:33 <irenab> thanks
14:00:37 <heyongli> bye
14:00:42 <baoli> #endmeeting