17:01:10 <sridhar_ram> #startmeeting tacker
17:01:11 <openstack> Meeting started Tue Apr 12 17:01:10 2016 UTC and is due to finish in 60 minutes.  The chair is sridhar_ram. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:01:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:01:16 <openstack> The meeting name has been set to 'tacker'
17:01:23 <sridhar_ram> #topic Roll Call
17:01:26 <zeih> o/
17:01:29 <dkushwaha> o/
17:01:30 <sridhar_ram> who is here?
17:01:32 <janki_chhatbar> o/
17:01:35 <bobh> o/
17:01:36 <michael_bredel> o/
17:01:36 <vishwanathj__> o/
17:01:42 <sridhar_ram> zeih: howdy, long time ;-)
17:01:57 <natarajk> o/
17:01:59 <sridhar_ram> hello everybody!
17:02:14 <zeih> sridhar_ram: howdy :)
17:02:14 <sridhar_ram> lets start..
17:02:21 <sridhar_ram> #topic Agenda
17:02:36 <sridhar_ram> https://wiki.openstack.org/wiki/Meetings/Tacker#Meeting_Apr_12.2C_2016
17:03:03 <sridhar_ram> beyond what is in the agenda, I'd like to touch base w/ michael_bredel on NFV Catalog
17:03:13 <sridhar_ram> anything else to discuss?
17:03:41 <trozet> o/
17:03:41 <sridhar_ram> #topic Annoucements
17:03:51 <sridhar_ram> trozet: hi there
17:03:56 <trozet> sridhar_ram: hi
17:04:08 <sridhar_ram> Austin Summit is just two weeks away!
17:04:24 <zeih> sridhar_ram: 5min for architectur Diskussion
17:04:40 <sridhar_ram> zeih: sure
17:05:07 <sridhar_ram> #topic Newton Design Summit Planning
17:05:27 <sridhar_ram> #link https://etherpad.openstack.org/p/tacker-newton-summit
17:05:55 <sridhar_ram> please look at the above, tentative topic assignments for our design summit session
17:06:01 <brucet> hello
17:06:06 <sridhar_ram> brucet: hi!
17:06:15 <brucet> In commit hell
17:06:33 <sridhar_ram> the format will be similar to other projects..
17:06:53 <sridhar_ram> we will have folks come up to the front to lead the discussion on a specific topic..
17:07:35 <michael_bredel> is it possible to attend online?
17:07:39 <sridhar_ram> .. it is encouraged for "active" participants to come up to the front row
17:08:18 <sridhar_ram> michael_bredel: I'm not sure if we can pull that off..
17:08:58 <sridhar_ram> michael_bredel: .. I sure can try to get a webex session going
17:09:43 <sridhar_ram> related note, we SHOULD use etherpad and shared edit capable tools - like google docs..
17:09:57 <michael_bredel> That would be great - thanks
17:10:24 <sridhar_ram> any questions on what you see in the etherpad ?
17:10:30 <sridhar_ram> anything glaringly missing ?
17:10:39 <sripriya> sridhar_ram: what would be the duration for BoF?
17:10:46 <sridhar_ram> 40mins
17:11:14 <sripriya> sridhar_ram: ok
17:11:17 <sridhar_ram> That session is technically not specific to Tacker...
17:11:36 <sridhar_ram> .. but a general topic on NFV Orchestration
17:12:07 <sripriya> sridhar_ram: i see ok
17:12:09 <sridhar_ram> it is intentionally cross community and cross project, so I've slotted some of those topic in there
17:12:31 <sridhar_ram> everything looks peachy ?
17:12:55 <sridhar_ram> I'll push this into the summit schedule..
17:13:08 <sridhar_ram> .. we can continue to tweak as we get closer to the summit date
17:13:35 <sripriya> sridhar_ram: i guess you have it captured for all topics needed to discuss :-)
17:13:49 <s3wong> hello. sorry, late
17:14:02 <sridhar_ram> sripriya: yeah, heavy agenda!
17:14:06 <sridhar_ram> s3wong: hi there
17:14:10 <sripriya> sridhar_ram: yup!
17:14:27 <sridhar_ram> Regarding switch Friday's contributor meetup to AM... no luck getting it switched officially.
17:14:56 <sridhar_ram> trading with a project currently slotted in AM looks doubtful..
17:15:24 <sridhar_ram> Worst case will find a place after going there and update the schedule / etherpad
17:15:41 <sridhar_ram> anything else to discuss on Summit ?
17:15:55 <trozet> sridhar_ram: so what is the current contributor time for Friday?
17:16:30 <sridhar_ram> trozet: 3.5hrs
17:16:48 <sridhar_ram> Friday 2:00pm - 5:30pm
17:16:55 <sridhar_ram> April 29th
17:17:04 <trozet> sridhar_ram: thanks
17:17:35 <sridhar_ram> the plan is to switch to Friday 9:00am - 12:30pm
17:17:36 <s3wong> I will be at the airport by 2pm (hopefully) :-)
17:17:46 <sridhar_ram> s3wong: :)
17:18:08 <sridhar_ram> moving on....
17:18:18 <sridhar_ram> #topic NSD spec
17:18:33 <sridhar_ram> dkushwaha: can you give a quick update ?
17:18:47 <dkushwaha> sridhar_ram, I have shared nsd support doc at https://docs.google.com/document/d/1yMIZMYDHx_QbljSlc4Jsju7YV8b_59voLtbJRPCQf70/edit#heading=h.z6ne0og04bp5
17:19:08 <trozet> dkushwaha: i requested access if you could please approve
17:19:14 <michael_bredel> me too
17:19:21 <vishwanathj__> dkushwaha me too
17:19:51 <sridhar_ram> folks - hang on, dkushwaha has pushed an equivalent doc in gerrit... lets use that to iterate
17:20:38 <dkushwaha> yes, please have a look in https://review.openstack.org/#/c/304667/1/specs/newton/nsd-support.rst
17:21:22 <sridhar_ram> dkushwaha: can you also add the sample NSD TOSCA template into this ?
17:22:03 <dkushwaha> sridhar_ram, yes
17:22:16 <sridhar_ram> heads - I'm planning to co-author this spec along with dkushwaha
17:22:22 <trozet> sridhar_ram, dkushwaha: I think this will need to have some basis on the VNFFG spec...I modeled the TOSCA input after part of the NSD
17:23:00 <sridhar_ram> trozet: totally agree, many things are falling in place now .. bit more naturally !
17:23:24 <trozet> sridhar_ram: yeah let just make sure we work together to have a common approach
17:23:35 <sridhar_ram> there are going to be multiple inter-related specs coming by.. NFV Catalog work including
17:23:53 <sridhar_ram> trozet: absolutely...
17:23:55 <trozet> https://review.openstack.org/#/c/292196/3/specs/newton/tacker-vnffg.rst line 158 is the tosca example i came up with...that has some assumptions because NSD isnt currently available (not to de-rail the current topic just FYI)
17:24:49 <sridhar_ram> that's the idea, to get all of us share *early* what we are bringing in..
17:25:23 <sridhar_ram> trozet: hold that thought for a sec, lets circle back to that..
17:25:37 <trozet> sure
17:26:13 <sridhar_ram> the scope for NSD is, IMHO, is just to take the next logical step.. to instantiate a sequence of VNFs..
17:26:38 <zeih> sridhar_ram: we are facing more and more NFVO specs and implementations in newton. So for me raises the question how to have a defined interface between NFVO and VNFM. :( currently it will all in one servercomponent. Any thoughts on this? Is it possible to Separate NFVO and VNFM? (ETSI like)
17:27:05 <zeih> Ups..
17:27:05 <zeih> Sry
17:27:16 <sridhar_ram> zeih: that's a fair question...
17:28:03 <zeih> Thats my question for Arch discussion
17:28:10 <sridhar_ram> If you remember our midcycle discussion, I'm in favor of decomposing tacker service into multiple smaller / micro services...
17:28:37 <zeih> sridhar_ram: yes,
17:28:53 <sridhar_ram> we sure need to discuss the architecture evolution of Tacker..
17:29:05 <sridhar_ram> .. it is a tad related to the technical debt/refactoring ...
17:29:15 <sridhar_ram> zeih: are you going to be in the summit ?
17:29:22 <zeih> sridhar_ram: isnt it easier to to this earlier than later?
17:29:50 <zeih> sridhar_ram: unfortunately not in Austin
17:29:59 <sridhar_ram> zeih: absolutely, however the trade off is we need to pause bringing in new features..
17:30:19 <dkushwaha> sridhar_ram, I have just added sample nsd tosca template in spec, please have a look
17:30:26 <sridhar_ram> dkushwaha: thanks!
17:30:34 <zeih> sridhar_ram: I will follow in etherpads
17:31:07 <sridhar_ram> either case, I'll try to slot in an arch evolution topic into the summit discussion...
17:31:32 <zeih> sridhar_ram: cool
17:31:46 <sridhar_ram> .. perhaps we can find a way to put a target steady state arch.. and chew portions of it using different specs in flight..
17:31:56 <sridhar_ram> may be I'm moonlighting :)
17:32:28 <sripriya> zeih: fyi, with multisite feature we took a first stab at defining that interface between NFVO and VNFM, ofcourse this needs to evolve
17:32:37 * sridhar_ram realizes specific agenda doesn't slot well, so...
17:33:04 <zeih> sridhar_ram: yes
17:33:12 <sridhar_ram> #topic NFV Catalog + NSD + VNFFGD
17:34:02 <sridhar_ram> michael_bredel: can you provide a quick update from your side on NFV Catalog evolution ?
17:34:33 <sridhar_ram> michael_bredel: are you / your team comfortable to start writing a tacker-spec for it ?
17:34:58 <michael_bredel> Sure. Sonata is already working on a catalog implementation for its own system.
17:35:30 <michael_bredel> We will have a project-internal face-2-face meeting and hackathon next week.
17:36:02 <michael_bredel> Hope is to have something running after that (includes also an integration with a - simple - sdk ...)
17:36:54 <michael_bredel> Our NSD is based on ETSI so far - but we already had some interesting discussions on that which you (hopefully) can benefit from
17:37:22 <michael_bredel> We are ready to start writing the tacker-spec after the meeting
17:37:42 <sridhar_ram> michael_bredel: great...
17:38:05 <sridhar_ram> michael_bredel: your thoughts on NSD would absolutely help here as well
17:38:22 <sridhar_ram> looking forward you comments on the tacker nsd spec
17:38:44 <michael_bredel> Sure - I will do that.
17:38:50 <sridhar_ram> michael_bredel: thanks!
17:39:07 <sridhar_ram> lets switch to VNFFG...
17:39:22 <sridhar_ram> trozet: do you want to share the course correction you are up for ?
17:39:36 <trozet> sridhar_ram: sure
17:39:45 <dkushwaha> michael_bredel, I had stuck some other things in last weeks, so could not discussed with you regarding nsd.
17:39:59 <trozet> so the original POC for using tacker to do SFC, was creating chains and classifiers directly with OpenDaylight
17:40:21 <trozet> that effort kind evolved spec wise into calling into networking-sfc to do chain/classifier creation
17:40:34 <trozet> with an ODL driver living in networking-odl and called by networking-sfc
17:40:53 <trozet> the tacker spec was modified to be vnffg (rather than SFC) but still focused on chain and classifier creation
17:41:11 <trozet> I realized its probably much more useful to make it into a real vnffg, so I have re-written the spec to abandon the chain/classifier direct interaction
17:41:17 <trozet> and instead use TOSCA templates
17:41:31 <sridhar_ram> woohoo!
17:41:34 <trozet> so a user defines a VNFFGD, and then creates a VNFFG - which may in turn create many chains and classifiers
17:41:37 <s3wong> no APIs
17:41:44 <trozet> by invoking a networking-sfc driver
17:41:58 <trozet> would appreciate reviews on the spec: https://review.openstack.org/#/c/292196/
17:42:24 <trozet> I outline some assumptions about the TOSCA implementation for this effort in the doc
17:43:42 <trozet> this also tries to marry the idea of ETSI VNFFG to IETF SFC/NSH
17:44:02 <trozet> so I have written it with that in mind, and have some NSH related changes in there for VNFM
17:44:10 <trozet> for supporting some extra parameters
17:45:00 <sridhar_ram> trozet: two different worlds - ETSI and IETF colliding thru Tacker :)
17:45:07 <s3wong> trozet: NSH won't be there (in OVS) until later
17:45:11 <sridhar_ram> s3wong: my understanding is, with this new approach, there won't be any APIs to create chains and classifiers...instead there will be new APIs to onboard VNFFGD template and to CRUD vnffg based on those template. trozet is this understanding correct ?
17:45:29 <trozet> sridhar_ram: thats correct
17:45:43 <sridhar_ram> trozet: thanks!
17:45:49 <trozet> s3wong: sure but that is outside of our scope
17:45:55 <sridhar_ram> s3wong: does this make sense ? do you see any issues ?
17:45:55 <trozet> s3wong: we already do it in OPNFV, with patches for OVS
17:45:58 <s3wong> trozet: since we need to use ODL ML2 driver also, we need to know what OVS version would be used by them at Newton timeframe, even if OVS is going to have NSH support (allegedly soon)
17:46:34 <sridhar_ram> LouisF: ^^
17:46:37 <s3wong> sridhar_ram: I don't have any problem... I had discussion with trozet  last time he raised this point on #tacker channel also
17:46:50 <trozet> s3wong: yeah let me clarify that...
17:46:57 <trozet> this spec doesnt mean you have to use NSH
17:47:10 <trozet> it should work also with networking-sfc's mpls implementation
17:47:20 <sridhar_ram> trozet: +1, agree
17:47:21 <trozet> it just accounts for future NSH parameters encap type
17:47:36 <trozet> so that when its here, tacker VNFFG will be able to use it
17:47:37 <s3wong> trozet: I wonder if we would touch NSH at Tacker level, really...
17:47:57 <trozet> s3wong: its VNFM attributes of a VDU
17:48:03 <trozet> s3wong: like service type
17:48:13 <s3wong> trozet: is encap type something that is specified at VNFFG at NFV profile?
17:48:15 <trozet> s3wong: we need to know if the VNF supports NSH
17:48:23 <s3wong> trozet: I see
17:48:37 <s3wong> trozet: sounds very low level...
17:48:52 <trozet> s3wong: yeah for the virtual links that connects VNFs in the VNFFG, we need to know what encap type they are using to communicate (VXLAN+GPE) etc
17:48:53 <sridhar_ram> s3wong: we can extend NFV profile for these needs..
17:50:00 <trozet> I feel like Tacker is the place to do this since it knows/manages the VNFs, it shoudl know its properties
17:50:13 <s3wong> trozet: I see... would be nice if this can be auto-negotiated by the VNFs themselves... but not getting ahead of ourselves here...
17:50:14 <trozet> rather than networking-sfc or something querying the VNF for the info
17:50:45 <s3wong> trozet: networking-sfc API does include specifying the encap type and parameters
17:51:28 <trozet> s3wong: yeah but im saying i dont think its networking-sfc's job to figure out what VNFs support that
17:51:38 <trozet> im saying its the VNFM job to hold the info about the VNFs
17:51:45 <s3wong> trozet: but yeah, I agree... someone needs to specify that somewhere before Tacker can pass that info to networking-sfc
17:51:53 <trozet> so i think its the appropriate place to list VNF specific attributes
17:52:11 <sridhar_ram> vnf metadata could be the ideal place for such things
17:52:21 <s3wong> sridhar_ram: sure
17:52:27 <sridhar_ram> trozet: thanks for sharing the changes.. I'm quite ecstatic about this course correction.. this take closer to where we want to go and avoids introducing some APIs with the idea that it would be deprecated !
17:52:44 <s3wong> trozet: I generally agree with this --- thinking about value Tacker provides here
17:52:44 <zeih> sridhar_ram: +1
17:52:57 <sridhar_ram> now, back to putting NSD + VNFFGD .. they need to parse / gel well together..
17:53:11 <trozet> so I dkushwaha I would be interested in your feedback with respect to the NSD
17:53:20 <trozet> and michael_bredel
17:53:22 <s3wong> trozet: if each VNF supports multiple encap types, perhaps Tacker VNFFG can select which one is common across the entire VNFFG
17:53:36 <sridhar_ram> NSD could just be a "envelope" with pointers to VNFDs and VNFFGDs
17:53:38 <trozet> s3wong: yeah
17:54:01 <trozet> s3wong: I have a bunch of features for VNFFG, but going to hold off on those
17:54:01 <s3wong> trozet: cool
17:54:16 <sridhar_ram> trozet: yes, lets pace ourselves...!
17:54:24 <s3wong> trozet: cool, we can discuss them during the summit
17:54:29 <trozet> s3wong: one really cool feature I briefly mention, is if a VNFFGD has many paths, we can actually optimize the chains that are created by consolidating common paths into single chains
17:54:52 <trozet> s3wong: so the orchestrator can make some intelligent decisions there
17:55:02 <s3wong> trozet: sounds cool --- but security policy would have to kick in to ensure we actually CAN do that :-)
17:55:03 <sridhar_ram> trozet: you are now going into real graph theory ... !
17:55:09 <dkushwaha> trozet, yes sure, I wiil look into that
17:55:17 <sridhar_ram> I'm game.. that's my elective in my undergrad
17:55:33 <trozet> sridhar_ram: I think what's lacking now is reviews on this, would like to get it through so I can code it
17:55:39 <trozet> sridhar_ram: good will need your help then :)
17:56:16 <sridhar_ram> anyone else beyond me and s3wong interested in reviewing VNFFG ?
17:56:36 <zeih> o/
17:56:37 <s3wong> trozet: yeah, let's come up with a list of cool stuff we can discuss during the summit --- but of course we need to get the basic stuff done first
17:56:43 <vishwanathj__> I would like to get involved to understand, but am no expert though
17:56:48 <sridhar_ram> it would need some investment in time to understand and ask question.. but SFC is really hot and cool feature
17:56:56 <janki_chhatbar> o/
17:57:09 <natarajk> o/
17:57:19 <dkushwaha> sridhar_ram, I would like to review VNFFG
17:57:31 <sridhar_ram> trozet: there you go.. you got many!
17:57:48 <trozet> cool :)
17:57:49 <s3wong> I am holding all these volunteers accountable for actually reviewing the spec :-)
17:57:58 <sripriya> s3wong: lol
17:58:01 <sridhar_ram> seriously, just start by asking questions if you don't understand...
17:58:38 <sridhar_ram> .. those, sometime, seemingly simple questions expose some design flaws
17:58:49 <sridhar_ram> an "outsider" is the ideal candidate for that
17:59:09 <sridhar_ram> s3wong: totally.. I need +1s for all these volunteers to land trozet's spec & code
17:59:17 <sridhar_ram> times up folks...
17:59:32 <sridhar_ram> that kinda felt a bit like open-discussion ;-)
17:59:41 <s3wong> good discussion, good meeting
17:59:51 <sridhar_ram> talk to you all next week..
17:59:53 <michael_bredel> yep - very interesting!
17:59:58 <sridhar_ram> bye
18:00:00 <s3wong> bye
18:00:01 <sripriya> thanks bye
18:00:02 <michael_bredel> bye
18:00:02 <dkushwaha> bye
18:00:04 <zeih> Next week I will be at plane
18:00:07 <zeih> Bye
18:00:12 <sridhar_ram> zeih: safe travels
18:00:15 <sridhar_ram> #endmeeting