17:00:46 <LouisF> #startmeeting service_chaining 17:00:47 <openstack> Meeting started Thu May 18 17:00:46 2017 UTC and is due to finish in 60 minutes. The chair is LouisF. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:00:48 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 17:00:50 <openstack> The meeting name has been set to 'service_chaining' 17:00:59 <LouisF> hi all 17:01:06 <pcarver> hi 17:01:12 <LouisF> pcarver: hi 17:01:21 <vks1> hi all 17:01:34 <bcafarel> hello 17:01:48 <doonhammer> hi all 17:02:31 <LouisF> #topic agenda 17:02:35 <LouisF> https://wiki.openstack.org/wiki/Meetings/ServiceFunctionChainingMeeting#Agenda_for_the_Networking-SFC_Meeting_.285.2F18.2F2017.29 17:03:13 <LouisF> i was not at the Summit 17:03:35 <LouisF> can anyone provide highlights? 17:04:11 <bcafarel> no sfc presentations then? 17:04:18 <pcarver> I didn't see a lot SFC related, but I was focussed on Contrail and ODL 17:04:44 <LouisF> bcafarel: no 17:04:45 <pcarver> Some folks from Midonet did a hands one workshop with the midonet-cli for layer 2 service insertion 17:04:59 <pcarver> but it didn't involve networking-sfc at all 17:05:36 <LouisF> ok 17:05:41 <pcarver> There was an talk on SFC as part of the ODL day 17:05:55 <pcarver> but again, not specific to networking-sfc 17:06:35 <pcarver> it did rasie some questions about the relationship between OpenStack and MANO stuff 17:06:50 <LouisF> i think we can do a talk on the sfc pike enhancements for the next summit 17:07:03 <LouisF> pcarver: such as? 17:07:55 <LouisF> is this related to tacker? 17:07:57 <pcarver> I don't know the specfics of things like TOSCA and similar, but there were questions about whether the APIs align 17:08:48 <pcarver> slightly related to tacker, but I'm not sure how well that aligns either 17:09:59 <pcarver> I think it may be a case of no clear standard that everyone agrees is the one standard 17:10:34 <LouisF> the NFV network forwarding path (NFP) aligns very well with port chain 17:11:22 <LouisF> i think the NFP is part of a vnffg tosca definition 17:11:51 <doonhammer> LouisF: do you have a link for NFP? 17:12:47 <LouisF> http://docs.oasis-open.org/tosca/tosca-nfv/v1.0/csd03/tosca-nfv-v1.0-csd03.pdf 17:13:03 <LouisF> its also in the ETSI MANO spec 17:13:05 <vks1> we had shown the demo of chaining in Openstack Summit with GBP 17:13:21 <doonhammer> LouisF: Thanks 17:13:21 <vks1> not this one but earlier ones 17:14:00 <LouisF> vks1: did that use networking-sfc? 17:15:11 <vks1> no we implemented that on CISCO ACI , but was doing chaining like SFC but was bit advanced 17:15:26 <LouisF> Figure 7 of that tosca doc shows NFP - same as port chain 17:15:58 <LouisF> vks1: do you have a link? 17:16:44 <vks1> which link do you need ? I can point you to that 17:16:50 <LouisF> i think igor's work in SF graph will align with nfv/tosca vnffg 17:17:08 <LouisF> vks1: link to the demo 17:17:16 <vks1> oh ok will send you 17:17:22 <LouisF> vks1: thks 17:17:51 <LouisF> #topic cli work 17:18:02 <LouisF> CLI in python-neutronclient https://review.openstack.org/#/c/409759/ 17:18:22 <bcafarel> no recent update? 17:18:44 <LouisF> no - i will contact mohan 17:19:06 <LouisF> #topic API reference 17:19:19 <LouisF> pcarver: what is status? 17:20:12 <pcarver> not much progress, I just haven't been able to spend much time debugging the issues 17:20:22 <LouisF> pcarver: ok thanks 17:20:26 <pcarver> question though 17:20:42 <pcarver> should we move SFC specific exceptions into neutron-lib? 17:21:20 <LouisF> pcarver: how is the handled for other projects? 17:21:25 <pcarver> Are they expected to be used generally or should they remain defined in networking-sfc essentially "privately" 17:22:06 <pcarver> there are some exceptions defined in neutron-lib but I haven't looked at what's not in neutron-lib for other projects 17:22:18 <bcafarel> if the API reference needs to reference them, probably yes, else that would mean a dependency on networking-sfc? 17:22:19 <pcarver> for the BGPVPN API def I didn't encounter the issue 17:22:40 <igordcard> hi, sorry 17:22:47 <bcafarel> but yes the neutron-lib resident experts will probably know :) 17:22:49 <LouisF> igordcard: hi 17:23:06 <pcarver> It's actually the API definition. The API reference is basically complete but Armando wanted the API def as a pre-req to merging the API ref 17:23:42 <igordcard> +1 nfp aligns well with port-chain, sfp does as well 17:24:00 <LouisF> igordcard: agree 17:24:25 <igordcard> but service graphs aligning with vnffg, a bit more difficult... 17:25:07 <LouisF> igordcard: where do you see the differences? 17:25:45 <LouisF> pcarver: can you check with neutron-lib on yr question 17:26:02 <igordcard> LouisF: want to discuss now or after the topic? 17:26:17 <LouisF> pcarver: anything to add? 17:26:54 <pcarver> LouisF: yes, just a matter of time. Got bogged down before the summit and now just barely caught up 17:27:05 <pcarver> need to find time to focus on SFC 17:27:19 <LouisF> pcarver: ok thanks 17:27:31 <LouisF> #topic vnffg alignment 17:27:50 <LouisF> igordcard: lets talk about this now 17:28:01 <igordcard> alright so even though an NFP kinda looks like an SFP, when they're grouped together to form graphs a main difference pops out 17:30:04 <igordcard> you can theoretically insert an SFP based on any condition, and so you can depend on another SFP (well this case is even explicitly written in the sfc arch), while in NFP you have an "nfpRule" which is the condition to get into the path but, from what I understand, it's the classifier used to get into the path 17:30:25 <igordcard> so I'm not sure it was designed taking into account that you could make an NFP depend on another NFP 17:30:56 <igordcard> when you look at the VNFFG, it always starts from a single point and everything is a branch of that point 17:31:29 <igordcard> but if nfpRule can be defined as allowing dependencies, then I see no major difference... 17:31:37 <LouisF> the vnffg spc does not really describe how traffic gets into the vnffg 17:33:04 <LouisF> where do you see nfpRule? 17:33:05 <igordcard> what about multiple instances at each node of the NFP, is that possible? I don't quite remember at the moment 17:33:41 <igordcard> usually, this is my guide: http://www.etsi.org/deliver/etsi_gs/NFV-IFA/001_099/014/02.01.01_60/gs_NFV-IFA014v020101p.pdf 17:34:03 <igordcard> LouisF: 6.4.3.2 17:34:27 <igordcard> yes, NFP supports multiple instances (connection points) at each hop 17:34:55 <LouisF> also http://www.etsi.org/deliver/etsi_gs/NFV-MAN/001_099/001/01.01.01_60/gs_NFV-MAN001v010101p.pdf 17:36:03 <igordcard> LouisF: hmm, seems like "policy" becomes "nfpRule" 17:37:05 <LouisF> igordcard: yes 17:39:32 <LouisF> igordcard: multiple cp corresponds to ppg 17:40:01 <igordcard> LouisF: yes.. and a cp is a pp? 17:40:16 <LouisF> igordcard: yes 17:41:04 <LouisF> igordcard: although is is unclear from the vnffg spec on the intention behind multiple cps per nfp 17:42:13 <LouisF> ok moving on 17:42:22 <igordcard> LouisF: oh! yes... the cardinality isn't right 17:42:46 <igordcard> LouisF: if N is the length of the NFP, then there is only 1 instance per hop 17:43:16 <igordcard> LouisF: unless the "VNF" wraps multiple instances, which might be the case, I'll have to take a look again 17:43:46 <LouisF> igordcard: yes N = number of CPs in a NFP seems correct 17:44:25 <LouisF> ok need to investigate further, but there is pretty good alignment... 17:44:35 <LouisF> #topic SF tap 17:44:48 <LouisF> vks1: what is status? 17:45:21 <vks1> I have submitted patch fr review 17:45:53 <vks1> you are in the review list :) 17:46:00 <vks1> *reviewer 17:46:00 <igordcard> haven't had much time to focus on the bigger patches, but I'm hoping to have a look in about 2 weeks time.. is that alright? 17:46:01 <LouisF> vks1: i see you have posted patches https://review.openstack.org/#/c/449954/ and https://review.openstack.org/#/c/465057/ 17:46:25 <bcafarel> ack, I am a bit behind on big reviews too 17:46:28 <vks1> yeah correct 17:46:33 <LouisF> vks1: thanks - all please review these 17:47:06 <LouisF> #topic OVN work 17:47:19 <LouisF> doonhammer: what is status? 17:48:57 <doonhammer> LouisF: still working patch through OVN/OVS did check my old networking-ovn and networking-sfc code, the ovn code can be merged the sfc code has a few minor issues bcafarel: reached out to help so not too much work 17:49:38 <LouisF> doonhammer: thanks 17:49:53 <bcafarel> doonhammer: I played a bit to set up an OVN / SFC setup, so should be able to try your branches soon 17:50:00 <doonhammer> LouisF: Biggest issue is we have run out of table space in OVN so I need a patch submitted by mickeys_: to be approved before I can submit the patch 17:50:30 <LouisF> doonhammer: i see 17:50:32 <doonhammer> bcafarel: any issues just reach out and I can help 17:50:39 <bcafarel> doonhammer: will do:) 17:51:14 <LouisF> great - thanks 17:51:21 <bcafarel> apart from the table space extension (independant issue), everything is ok on the OVN side? 17:51:25 <doonhammer> LouisF: Hopefully blp: will review mickeys_ patch soon in the meantime I can apply his patch to my fork' 17:51:59 <LouisF> #topic Non-transparent SF 17:52:02 <LouisF> https://review.openstack.org/#/c/458303/ 17:52:07 <doonhammer> LouisF: Just working through issues - I have been testing it against docker and k8s and that works 17:52:18 <LouisF> doonhammer: great 17:52:40 <LouisF> please review 17:52:51 <LouisF> #topic other items 17:53:38 <igordcard> I have to work on a patch before service graphs can be merged.. hoping to finish the first chunk of graphs by next week 17:54:07 <igordcard> the OSC patch might take a bit longer, then I'm moving to NSH 17:54:26 <LouisF> igordcard: thanks 17:54:44 <bcafarel> please check also http://lists.openstack.org/pipermail/openstack-dev/2017-May/116600.html some feedback will be nice 17:55:38 <igordcard> bcafarel: oh, my bad haven't checked the ml often enough lately 17:55:53 <LouisF> bcafarel: will comment on that 17:56:10 <bcafarel> igordcard: np I think you had the summit on your hands at that time 17:56:17 <bcafarel> LouisF: thanks 17:57:42 <LouisF> bcafarel: lets discuss more at next meeting 17:57:50 <LouisF> i have another meeting 17:57:55 <LouisF> thanks all 17:58:17 <LouisF> #endmeeting