16:00:35 <sgordon_> #startmeeting nfv
16:00:35 <openstack> Meeting started Thu Sep  4 16:00:35 2014 UTC and is due to finish in 60 minutes.  The chair is sgordon_. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:36 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:38 <bauzas> ohai - aloha - morning - bonjour - guten tag - \o
16:00:38 <openstack> The meeting name has been set to 'nfv'
16:00:48 <sgordon_> !topic roll call
16:00:49 <openstack> sgordon_: Error: "topic" is not a valid command.
16:00:54 <sgordon_> #topic roll call
16:00:57 <sgordon_> whoops :)
16:01:02 <bauzas> ohai - aloha - morning - bonjour - guten tag - \o
16:01:04 <sgordon_> who is here for the nfv meeting?
16:01:13 <cloudon> hi
16:01:31 <sgordon_> expecting to be short/quick today as i expect a lot of people are dealing with feature freeze related work
16:01:44 <lukego> Howdy
16:01:50 <sgordon_> hi luke
16:02:00 <sgordon_> #link https://etherpad.openstack.org/p/nfv-meeting-agenda
16:02:16 <sgordon_> #topic review action items
16:02:26 <sgordon_> so russell scored two action items last week
16:02:48 <sgordon_> one on clarification of FeatureProposalFreeze wiki, which i am unclear on - possibly still outstanding
16:02:56 <sgordon_> the other was related to the NUMA I/O scheduling
16:03:08 <sgordon_> now as ndipanov appears to be requesting an FFE for the base NUMA work
16:03:18 <sgordon_> i am not sure where this one stands, maybe jchapman can update
16:03:28 <danpb> FYI on NUMA  FFE
16:03:30 <jchapman> sgordon, he
16:03:32 <jchapman> hi
16:03:40 <danpb> it has sufficient core sponsors but the questions are around testing
16:03:59 <danpb> i think it would be helpful if anyone were  able to step forward and publically say they are going to test this before release
16:04:23 <sgordon_> #info Base NUMA functionality will require an FFE, has sponsors but needs testing commitment
16:04:25 <danpb> (say it in response to the email requesting NUMA FFE )
16:04:44 <sgordon_> right
16:04:56 <danpb> the pushback we're seeing against NUMA is all related to level of testing
16:05:05 <danpb> so anything to mitigate that is useful
16:05:15 <sgordon_> #info need a lucky volunteer to test NUMA functionality before release
16:05:28 <adrian-hoban> danpb: We will test the NUMA functionality
16:05:42 <danpb> oh and they need to say this on the mail within the next 24 hrs
16:05:52 <sgordon_> #link http://lists.openstack.org/pipermail/openstack-dev/2014-September/044878.html
16:06:16 <sgordon_> and specifically:
16:06:18 <sgordon_> #link http://lists.openstack.org/pipermail/openstack-dev/2014-September/044906.html
16:06:50 <sgordon_> adrian-hoban, indeed - i think the key is to have a documented commitment on list as to how this will happen
16:07:41 <adrian-hoban> danpb: sgordon: We will reply to the mail saying what we intend to test
16:07:49 <sgordon_> anyone else with a particular interest in this feature that would be interested would be welcome as well
16:07:51 <danpb> adrian-hoban: thanks
16:07:59 <sgordon_> e.g. providers etc.
16:08:15 <sgordon_> we kind of skipped right into this
16:08:17 <sgordon_> but
16:08:20 <sgordon_> #topic juno status
16:08:34 <sgordon_> obviously we have hit feature freeze, i have seen the NUMA FFE request we just discussed
16:08:41 <sgordon_> looks like one was also just filed for SR-IOV
16:08:56 <sgordon_> #link http://lists.openstack.org/pipermail/openstack-dev/2014-September/044923.html
16:09:00 <danpb> yes, SR-IOV FFE got a positive reception in nova meeting too
16:09:11 <sgordon_> #info SR-IOV FFE proposed and received positive reception in nova meeting
16:09:35 <nidpanov> here now too
16:09:37 <sgordon_> #info two cores so far signed up for SR-IOV FFE
16:10:03 <adrian-hoban> That's good news
16:10:25 <sgordon_> so, those are the two ones i can see already related to work this group has been tracking
16:10:42 <sgordon_> does anyone from this group have plans to submit additional FFE requests?
16:10:58 <nidpanov> danpb, sorry to go back - missed your comment about numa testing - so the issues were mostly due to me mentioning it in the email, and HP already kind of signed up to test it as well iiuc
16:11:24 <sgordon_> nidpanov, i think it is probably necessary to confirm with Paul that was the intent of his email
16:11:32 <jchapman> we are not lookign for FFE  for I/O (PCIe) Based NUMA Schedulin  :(
16:11:37 <sgordon_> it seemed somewhat unclear to me
16:11:48 <nidpanov> sgordon_, hence my response
16:11:55 <sgordon_> ack
16:12:06 <sgordon_> jchapman, understoof
16:12:10 <nidpanov> jchapman, is that because of the NUMA delay or?
16:12:25 <jchapman> nidpanov, that and no core dev reviews
16:12:36 <nidpanov> jchapman, :(
16:12:45 <nidpanov> I had my hands full
16:12:45 <sgordon_> #info I/OC (PCIe) Based Numa Scheduling FFE not being requested due to issues with base NUMA functionality and lack of interested core sponsors
16:13:01 <jchapman> nidpanov, ye i guessed you guys were busy
16:13:09 <danpb> realistically the only stuff that's going to get approved FFE is stuff that already had a +A and was in the gate at time of freeze
16:13:22 <nidpanov> if that
16:13:22 <cloudon> lukego, is your accel data plane stuff going to make it?
16:13:22 <sgordon_> yeah understood
16:13:30 <danpb> so yeah, i'm afraid I/O PCI numa scheduling would  loose on that count even if proposed
16:13:39 <adrian-hoban> The I/O Numa scheduling will be re-targeted at Kilo now
16:13:46 <sgordon_> yes
16:13:50 <jchapman> yep
16:14:01 <lukego> cloudon: ‘fraid not. Neutron core asked me to develop it more out of tree first, and then Nova wouldn’t take the VIF until Neutron code is accepted.
16:14:04 <sgordon_> hopefully the desire to come up with a way to preload designs for summit will go somewhere useful...
16:14:46 <lukego> cloudon: at least within this NFV group those of us interested in userspace virtio-net seem to be on the same page and on track to share code going forward, which I think is the most important thing
16:14:51 <cloudon> lukego: shame - are you hopeful for Kilo?
16:15:00 <bauzas> sgordon_: yeah, at least we need to see what will be the new format for the Design Summit
16:15:51 <sgordon_> soooo
16:15:57 <sgordon_> one other thing i wanted to quickly mention
16:16:03 <sgordon_> #topic other discussion
16:16:05 <sgordon_> #link http://lists.openstack.org/pipermail/openstack-dev/2014-September/044614.html
16:16:07 <lukego> cloudon: I am hopeful but it does depend on the success of initiatives in Neutron and Nova to accommodate more drivers/features
16:16:23 <cloudon> lukego: understand - good luck
16:16:31 <sgordon_> itai put together a very well thought out email on some of our current issues and attempting to frame discussion about where we go from here
16:16:45 <lukego> cloudon: on the list we also discuss the possibility of sharing a mech driver instead of having a separate snabb vs. ovdk vs. etc ones, which I suspect would resolve the Neutron core concerns (which are roughly “too many drivers”)
16:16:47 <sgordon_> i havent had a chance to fully consider and respond yet but i would encourage others to also review
16:17:39 <cloudon> sgordon: +1 - on first read looked very sensible
16:18:06 <adrian-hoban> +1 on that note. Nicely written Itai
16:19:02 <sgordon_> does anyone have anything else they would like to discuss this week? i would probably prefer to close down asap and let those dealing with FFE items get back to it
16:19:14 <adrian-hoban> Itai: One thing to consider is the focus on the wiki... I think we need to broaden horizons in getting the info you suggest out to the community.
16:19:20 <sgordon_> and discuss some of the medium to long term issues/goals starting next week
16:19:32 <sgordon_> incl. kilo planning
16:19:58 <sgordon_> even the wiki needs to be broken up and organized a little imo
16:20:35 <cloudon> sgordon: quick question on Paris summit - see you +adrian-hoban are giving a telco strategies track talk on this group - congrats on getting selected & what's your plan/do you need further stuff from this group?
16:20:54 <sgordon_> cloudon, good question - we're not really sure yet :)
16:21:22 <sgordon_> cloudon, the idea we had discussed was to try and present some use cases, what's there today, and more importantly what's not
16:21:29 <cloudon> sgordon: lucky you have two months then :)
16:21:39 <sgordon_> but we will be trying to flesh out further
16:21:58 <sgordon_> obviously wrt juno we made some progress but not as much as we would like
16:22:21 <sgordon_> which is obviously not ideal but also provides us with some good examples of solid proposals that exist but ultimately did not get implemented
16:23:54 <cloudon> sgordon: cool - if you want any help on use cases etc. am happy to assist
16:24:50 <adrian-hoban> cloudon: Thanks! Also, alank co-presenting too.
16:26:05 <bauzas> sgordon_: do you think we should try to promote NFV work to developers too ?
16:26:19 <sgordon_> bauzas, of course - but that is a challenging thing
16:26:24 <bauzas> sgordon_: I'm also concerned about the coding and reviewing bandwidth
16:26:37 <sgordon_> bauzas, design summit proposals need to be framed around specific goals
16:26:53 <sgordon_> bauzas, yes - in the context of nova i expect that to be a hot topic in general
16:26:58 <sgordon_> bauzas, not an nfv specific issue
16:27:12 <bauzas> sgordon_: lemme check if it's in the nova priorities for the summit
16:27:21 <sgordon_> (contributors interested in nfv need to contribute to this discussion to help the project move forward though or we're all in the log jam togetjer)
16:27:36 <bauzas> sgordon_: there is an etherpad and I don't remember having seen the NFV stuff to be debated
16:28:09 <bauzas> there
16:28:22 <bauzas> #link https://etherpad.openstack.org/p/kilo-nova-summit-topics Proposals for Nova Kilo Summit
16:28:48 <bauzas> sgordon_: can't see NFV in there ^
16:28:59 <sgordon_> bauzas, this is the thing though
16:29:09 <sgordon_> bauzas, proposing "let's talk about nfv" in there will get nacked
16:29:18 <sgordon_> bauzas, needs to be about specific issues or proposals
16:29:19 <bauzas> I don't know for Neutron but I assume there will be the same etherpad
16:29:25 <sgordon_> so for instance wrt ci
16:29:28 <bauzas> sgordon_: indeed
16:29:33 <sgordon_> newer libvirt versions is probably relevant for nfv
16:29:40 <bauzas> sgordon_: but can't see a related topic to that either
16:29:43 <sgordon_> (esp since some of the nfv features are now getting push back on testing)
16:29:49 <sgordon_> related to what?
16:29:58 <bauzas> related to the nfv usecases
16:30:02 <sgordon_> scheduler split in there is another one
16:30:38 <sgordon_> bauzas, so the backstory here is there were several such proposals for atlanta
16:30:56 <sgordon_> all were rejected with clear guidance that sessions need to have a very specific goals/proposals
16:31:02 <bauzas> got it
16:31:10 <adrian-hoban> resource tracker seems to be covered
16:31:31 <sgordon_> getting time allocated generically to NFV in the track will be very difficult, similar for HPC or other user story type things
16:31:37 <bauzas> I just want to make sure that the next Juno developments will be covered and agreed there :)
16:31:38 <sgordon_> adrian-hoban, right
16:31:54 <sgordon_> key is to focus on what we specifically want to achieve for kilo
16:32:01 <sgordon_> and ensure where necessary that is covered in there
16:32:11 <sgordon_> the scheduler / RT / ERT discussion is a great example
16:32:12 <danpb> also note that if you propose a topic for a session
16:32:16 <bauzas> +1 - and I just add: make sure it's covered in the design talks
16:32:20 <adrian-hoban> Is there a deadline on getting items into the lists?
16:32:21 <sgordon_> as a lot of features are running into issues on that front
16:32:35 <danpb> it will be expected that the topic has already had a spec proposed or some other detailed discussions or write up
16:32:56 <danpb> they're very unlikely to approve sessions for topics where this is the first time anyone has heard of the topic
16:33:01 <bauzas> sgordon_: the scheduler/RT thing will probably be highly debated, don't know if we will have time to cover extra things in there
16:33:14 <bauzas> danpb: +1
16:33:37 <adrian-hoban> danpb: Ok, but is there an order in which things need to happen? E.g. add to the list, release bp?
16:33:50 <danpb> the goal is that the sessions will focus on resolving actual problems that need f2f time
16:33:59 <danpb> rather than just presenting new ideas
16:34:50 <danpb> also i hope there will be more unscheduled time for ad-hoc discussions which will be an avenue to bring up any other items which are relevant
16:35:20 <adrian-hoban> danpb: Is this a living doc for the next few weeks? Perhaps up until the bp submission window closes?
16:35:27 <sgordon_> adrian-hoban, let me see if i can find mikal's email
16:36:48 <sgordon_> #link http://lists.openstack.org/pipermail/openstack-dev/2014-August/044196.html
16:36:53 <sgordon_> no date afaict
16:37:17 <sgordon_> let me bump the thread now and ask that question...
16:37:34 <adrian-hoban> sgordon: Thx
16:38:52 <sgordon_> im going to close up the meeting for now
16:38:56 <sgordon_> thanks all for the discussion
16:39:11 <sgordon_> let's see what michael says wrt that list and when it is going to be finalized
16:39:14 <bauzas> sgordon_: thanks
16:39:27 <sgordon_> i expect it will be a while yet while the dust settles on j-3 and the first RCs but we will see
16:39:30 <sgordon_> #endmeeting