13:01:42 <baoli> #startmeeting PCI passthrough
13:01:44 <openstack> Meeting started Tue Jan 20 13:01:42 2015 UTC and is due to finish in 60 minutes.  The chair is baoli. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:45 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:45 <nuritv> Hi
13:01:47 <openstack> The meeting name has been set to 'pci_passthrough'
13:02:33 <beagles> \o
13:04:03 <baoli> #topic Bugs
13:04:58 <baoli> https://bugs.launchpad.net/nova/+bug/1406486
13:05:16 <moshele> yes I fixed it
13:05:34 <baoli> moshele: suspend/resume was working before. Did you track down what has been changed to cause it not to work any more?
13:06:29 <baoli> moshele: I saw the change, and it's good I think. Just wondering what makes the get_admin_context() not to work any more.
13:06:47 <moshele> baoli: no, I didn't look at the hisrory
13:06:58 <moshele> history
13:07:48 <baoli> moshele: ok.
13:07:54 <moshele> as I see now it just return context with None value and admin flag is True
13:08:35 <baoli> so get_admin_context() doesn't return a valid context.
13:09:10 <moshele> let me look at the code again
13:11:01 <moshele> def get_admin_context(read_deleted="no"):
13:11:01 <moshele> return RequestContext(user_id=None,
13:11:01 <moshele> project_id=None,
13:11:01 <moshele> is_admin=True,
13:11:01 <moshele> read_deleted=read_deleted,
13:11:02 <moshele> overwrite=False)
13:12:49 <moshele> this is the context that returns
13:12:49 <baoli> moshele: ok, I'll give it a try some time today.
13:13:02 <moshele> ok
13:13:17 <baoli> Anything else to bring up on bugs?
13:13:58 <moshele> baoli: I need your second review for this  https://review.openstack.org/#/c/145035/
13:14:08 <moshele> Add range support to all address in white list
13:14:39 <baoli> moshele: will do
13:14:44 <moshele> thanks
13:15:02 <yongli_he> sorry, late
13:15:33 <baoli> yongli_he: hi
13:15:42 <yongli_he> hi, baoli
13:15:48 <baoli> yongli_he: we are talking about bugs
13:15:53 <shaohe_feng> hi all
13:16:01 <shaohe_feng> sorry for late
13:16:20 <baoli> shaohe_feng: hi
13:16:24 <yongli_he> i focus on ci these 2 weeks, not show up here, sorry
13:16:32 <shaohe_feng> just  come back hom.
13:17:07 <baoli> yongli_he: thanks for the CI work, good stuff.
13:18:25 <yongli_he> baoli, we does need CI to move PCI forward, and it does help, catch patches broken the pci
13:19:09 <moshele> baoli: we plan  to add CI in mellanox for nova as well, because it is blocking all the sriov features
13:19:17 <baoli> yongli_he: the CI is important for sr-iov/pci related blueprints as well.
13:20:05 <baoli> moshele: great.
13:20:13 <shaohe_feng> yes, CI can make openstack more robust
13:20:55 <nuritv> yongli_he: are you running API tests or other scenarios?
13:21:29 <baoli> let me post the meeting link here: https://wiki.openstack.org/wiki/Meetings/Passthrough
13:25:15 <baoli> beagles: on your BP, do we know how to determine a NIC supports single/multiple physical functions per physical port
13:26:14 <beagles> baoli, not to my knowledge.
13:27:16 <baoli> The cisco vNIC supports that. Not sure if intel/mlnx supports multiple physical functions per port.
13:28:06 * beagles nods
13:29:08 <baoli> yongli_he, moshele: any idea?
13:29:57 <moshele> I will check it here
13:31:46 <moshele> baoli: just for my understanding this bp will allow to allocate to vf from 2 diffrent port/nic so bonding on the vm can be done ?
13:33:38 <baoli> moshele, it seems to me that it allocates across physical functions (not ports)
13:34:06 <baoli> beagles: is that right?
13:34:09 <beagles> it is based on PCI information really... whatever nova has at hand at scheduling time
13:35:23 <beagles> so it'll spread consumption across all root PCI devices instead of consuming all from one and then going onto the next
13:36:08 <beagles> it is a *soft* improvement that could enable bonding, etc. but fundamentally it strives to provide better distribution of resource consumption
13:36:47 <beagles> the former is of interest to us, but the latter is simply hard to argue with... making it a more reasonable change for general acceptance.
13:38:08 <moshele> bealges: can we talk offline regarding you bp  we are also interested in it
13:38:10 <moshele> ?
13:38:15 <beagles> moshele, sure
13:41:37 <baoli> anything else to bring up?
13:43:23 <baoli> ok, let's call it a day, then.
13:43:36 <yongilhe_> my network broken, anything calling me?
13:44:29 <baoli> yonglihe: we had a question about whether or not intel NIC support multiple PFs per port
13:44:44 <moshele> youngihe_: what kind of CI testing are you doing api or scenarios testing?
13:45:17 <yongilhe_> baoli, i don't know either, let me ask someone later
13:45:39 <yongilhe_> moshele, we do scenarios testing, all testcases on git hub
13:46:01 <moshele> ok thanks
13:46:20 <baoli> yongihe_: sure, do let us know
13:50:03 <baoli> Ok, see you guys next week.
13:50:09 <yongilhe_> see you
13:50:12 <baoli> #endmeeting