18:15:51 <adetalhouet> #startmeeting Armoury Weekly 18:15:51 <odl_meetbot> Meeting started Mon Oct 26 18:15:51 2015 UTC. The chair is adetalhouet. Information about MeetBot at http://ci.openstack.org/meetbot.html. 18:15:51 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:15:51 <odl_meetbot> The meeting name has been set to 'armoury_weekly' 18:15:56 <adetalhouet> #topic Roll call 18:16:06 <adetalhouet> please #info pound in 18:16:15 <subh> #info subh 18:16:18 <ariel_noy> #info ariel_noy 18:17:28 <adetalhouet> let's wait a bit to see if others will join 18:17:42 <mmarsale> #info mmarsale 18:18:23 <alagalah> #info alagalah (here but not here) :) 18:18:52 <adetalhouet> ok let's move on 18:18:57 <adetalhouet> #topic Agenda bashing 18:19:10 <adetalhouet> #link agenda https://wiki.opendaylight.org/view/Armoury#Agenda_for_next_meeting_.2810.2F26.29 18:19:37 <adetalhouet> #info On going tasks 18:19:52 <adetalhouet> #info Be release plan (documentation) 18:20:02 <adetalhouet> #info Armoury design 18:20:08 <adetalhouet> #info SFC use case with Tacker 18:20:20 <adetalhouet> do you have anything you'd like to discuss today? 18:20:44 <ariel_noy> want to move to asynch design 18:20:55 <adetalhouet> Ok, let's get there now 18:21:00 <adetalhouet> #topic Armoury desgin 18:21:22 <adetalhouet> #link mail thread related to latest design: https://lists.opendaylight.org/pipermail/armoury-dev/2015-October/000066.html 18:21:49 <adetalhouet> ariel_noy: where you able to commit a patch with the proposed design? 18:22:04 <ariel_noy> Not yet. 18:22:25 <ariel_noy> Wanted first agreement and also can we commit the sequance diagram 18:22:51 <adetalhouet> ariel_noy: sure, I was actually referring to the sequence diagram 18:23:06 <ariel_noy> Ok where that should be 18:23:25 <ariel_noy> The orig txt file not apear in the git 18:23:49 <adetalhouet> ariel_noy: right, it hasn't been merge yet: https://git.opendaylight.org/gerrit/#/c/27263/ 18:24:37 <ariel_noy> Ok can we merge that and I will commit the changes 18:25:01 <adetalhouet> Sure 18:25:17 <odp-gerritbot> A change was merged to armoury: Do NOT merge yet: NF_instantiation web-sequence-diagram https://git.opendaylight.org/gerrit/27263 18:26:05 <adetalhouet> for folks you doesn't have the diagram under their eyes, here is the txt https://gist.github.com/adetalhouet/2c3fa4f1bdb9e02f3051 18:26:11 <adetalhouet> to paste into https://www.websequencediagrams.com/# 18:26:12 <ariel_noy> Thanks I will commit the hange after that 18:27:28 <adetalhouet> so about the diagram, there were some feedback on the mail thread 18:27:48 <adetalhouet> but I believe there were not Armoury related, where they? 18:27:58 <adetalhouet> where/were 18:28:11 <ariel_noy> The armoury related we fixed. 18:28:45 <adetalhouet> ok so the plan is to use http2, is that it? 18:28:52 <ariel_noy> Later stage we will be able to move to notification based API and not polling 18:30:02 <adetalhouet> Yes ok, this is what I understood from the thread 18:30:13 <adetalhouet> Is everyone fine with this? 18:31:02 <adetalhouet> I can hear crickets.. so I guess the answer is yes 18:31:16 <adetalhouet> ariel_noy: is there anything you want to add here? 18:31:29 <ariel_noy> no I am ok 18:31:32 <adetalhouet> before we move on to the next topic 18:31:49 <adetalhouet> #topic On going tasks 18:32:13 <adetalhouet> #link troll board: https://trello.com/b/zsERKgeV/odl-armoury 18:32:43 <adetalhouet> so far we have the model for our 3 components in: driver/catalog/workload manager 18:32:59 <adetalhouet> I've seen subh started work on the workload manager RPC 18:33:23 <adetalhouet> subh: do you plan on filling up those method? 18:33:54 <subh> I'll need help for working on those method 18:34:04 <ariel_noy> I will help. 18:34:24 <subh> sure.. please share the details 18:34:24 <ariel_noy> First need to change the API to the asynch design 18:34:59 <subh> for the asnc design do we have dependency on yang pub-sub ? 18:35:47 <ariel_noy> Let's not wait for them as they are not promissing in this release 18:35:59 <subh> okey 18:36:30 <ariel_noy> We better work on notification on the model and notification to the client will be later 18:37:03 <ariel_noy> Internal app in the ODL should be able to get notifications 18:37:50 <adetalhouet> I created a trello card for this task so you can sync up: https://trello.com/c/JJXovInM/6-implementation-of-workload-manager-rpcs 18:38:11 <subh> okey 18:38:17 <adetalhouet> also created this task https://trello.com/c/poXL86JL/7-redefining-worloadmanager-api-to-be-async 18:38:30 <adetalhouet> to rework the worloadmanager api 18:39:43 <subh> sure I'll sync up with ariel_noy 18:40:06 <adetalhouet> Regarding drivers now 18:40:25 <subh> https://git.opendaylight.org/gerrit/#/c/28265/ 18:40:31 <adetalhouet> we have a base model to create driver, it would be nice to start working on a tacker driver 18:40:44 <adetalhouet> #link https://git.opendaylight.org/gerrit/#/c/28265/ 18:40:49 <adetalhouet> subh: txs 18:41:14 <adetalhouet> There is some discussion around this for SFC purposes 18:41:31 <adetalhouet> An email was sent by alagalah at this regard 18:41:50 <adetalhouet> #link https://lists.opendaylight.org/pipermail/armoury-dev/2015-October/000068.html 18:41:53 <ariel_noy> What they send is without armoury. So I did not understand why they send us 18:42:49 <adetalhouet> alagalah: are you around and able to add more context here? 18:42:54 <adetalhouet> I didn't really get it either 18:43:22 <adetalhouet> #link Gdoc presenting the integration: https://docs.google.com/document/d/1ibuf1j49G97xUAr2zYC9mzQTW6kqZ7cAw0g-FPAZM20/edit# 18:43:53 <alagalah> adetalhouet, ? 18:44:02 <alagalah> What are you missing? 18:44:23 <adetalhouet> Does this present to way Tacker gets integrated into SFC? 18:45:09 <adetalhouet> If so, I don't really see how this involves Armoury - I mean yes we want to have our own Tacker driver, so SFC can use it. But this is not really what I can see here 18:45:18 <ariel_noy> Not sure how Tacker interact with armoury on the flow you send 18:45:33 <alagalah> adetalhouet, Ok so here is the issue 18:45:46 <alagalah> adetalhouet, In general SFC should be able to request and destroy instances of SFs 18:45:50 <alagalah> to the VNFM 18:45:55 <alagalah> There is no way of doing that today 18:46:08 <adetalhouet> ok but this is kind of our workload manager 18:46:28 <alagalah> adetalhouet, Well then great, we will have a common API for the implementation of the sfc-vnfm methods 18:46:49 <alagalah> There is nothing stopping armoury making a sfc-vnfm-armoury that implements the methods..... 18:46:58 <alagalah> ...and if armoury is using a tacker driver, SFC shouldn't care 18:47:14 <alagalah> Or if Armoury is using a CoreOS/K8 driver, SFC shouldn't care 18:47:43 <adetalhouet> Yes, I agree that SFC doesn't need to be aware of the underlying driver 18:48:00 <adetalhouet> So this common API will live in SFC, right? 18:48:20 <alagalah> Right, but in order for the schedulers and RSP maintainer to do CRUD (where D is the most important) it needs a way to signal 18:48:31 <adetalhouet> Because so far, we create a workload manager API in Armoury defining RPC to interact with VMs/images 18:48:40 <alagalah> Its a common set of methods that the Schedulers and RSP manager will use to interact with a VNFM.... 18:48:51 <alagalah> Ok thats great, this doesn't preclude that. 18:49:03 <adetalhouet> Ok 18:49:17 <adetalhouet> I'm asking because we don't want to overlap what is being done somewhere else 18:49:27 <alagalah> In fact it enables it... I mean, why not just map the sfc-vnfm to the armoury API in the sfc-vnfm-armoury impl of those methods and call it a day ? 18:50:03 <alagalah> In fact, I see one of the major benefits of armoury is multiple driver management 18:50:26 <adetalhouet> Yes sure that would be a good integration of Armoury APIs 18:50:39 <alagalah> ie sfc-vnfm calls one CreateSF (AbstractType...) method, and instead of doing VNFM arbitrage, that could be a big value add of Armoury 18:51:03 <alagalah> Cos a big use case is multi-datacentre and mutliDCOS 18:52:07 <alagalah> At the moment I don't see any pre-existing places to hook that allows for SFC to say "I need an SF of this type, and I'd prefer it [here]" 18:52:38 <alagalah> So this is a proposal to change that 18:52:53 <alagalah> Anyway, feel free to comment on the doc, its there for us all to get together on this. 18:53:07 <adetalhouet> yes so using Armoury you could define the [here] 18:53:09 <ariel_noy> Ok we agree on the armoury target not sure how armoury can use Tcker 18:53:20 <alagalah> As a committer on Armoury, I have as much skin in the game as making this succeed, and I see this as a clean interface 18:53:27 <alagalah> ariel_noy, ???? 18:53:42 <alagalah> ariel_noy, I thought a Tacker driver was on the RP ? 18:54:11 <ariel_noy> The flow of Tacker you send is a complete flow. How we get armoury involve 18:54:17 <alagalah> I mean, SFC is meant to interoperate with VNFMs.... 18:54:47 <alagalah> ariel_noy, So as I said on email, this was for OPNFV.... so rather than welding these things together.... 18:54:58 <alagalah> I had to show how it works with pre-existing work 18:55:18 <ariel_noy> Ok so we need to work the flow with armoury. 18:55:40 <adetalhouet> alagalah: what you're saying here is that Armoury can be one of the various VNFMs that can be used for SFC 18:55:54 <ariel_noy> Since it was not clear in that which function is Tacker taking and which armoury work loadmnager taking 18:56:21 <alagalah> adetalhouet, Right... 18:56:54 <alagalah> As I understand it, cataloging is a thing a VNFM would talk to... if ODL SFC decides it wants to also be a VNFM .,. (????) then it would need that .... 18:56:59 <adetalhouet> alagalah: although we clearly stipulate at the proposal that Armoury was not yet another orchestrator 18:57:10 <alagalah> adetalhouet, Ok and it doesn't need to be 18:57:33 <alagalah> Nothing here shuts out Armoury.... I'm not killing any API.... I'm opening up new ones 18:57:48 <alagalah> gotta jump to another meeting.... 18:57:53 <alagalah> we can take it up on email if you like ? 18:57:54 <adetalhouet> alagalah: yes I think I'm understanding what you're saying here 18:57:58 <alagalah> I don't see this as a BAD thing 18:58:07 <adetalhouet> alagalah: right 18:58:26 <adetalhouet> ok let's bring questions over ML if there are some things to clarify 18:58:29 <ariel_noy> I am not clear. Can you send a flow diagram with armoury 18:58:55 <adetalhouet> ariel_noy: I also agree we need to define a sequence diagram for our tracker driver and how it will work in Armoury 18:59:57 <ariel_noy> Ok need to drop off too 19:00:00 <adetalhouet> ok we're almost at the top of the hour 19:00:08 <adetalhouet> just wanted to talk about this topic 19:00:14 <adetalhouet> #topic Be release plan documenation 19:00:28 <adetalhouet> ariel_noy: ok sure, thanks for attending :) 19:00:37 <adetalhouet> subh: started the documentation outline 19:00:39 <adetalhouet> with anipbu 19:00:54 <adetalhouet> #link doc https://git.opendaylight.org/gerrit/#/c/28541/4 19:00:58 <subh> I haven't made much progress in this week 19:01:11 <adetalhouet> @all, please review the patch to help subh in his task 19:01:33 <adetalhouet> subh: it is really fine 19:01:34 <adetalhouet> :) 19:02:01 <subh> I'll try to take some content from wiki and add it to docs 19:02:16 <adetalhouet> Does anyone has anything to ask/say> 19:02:55 <adetalhouet> Ok then, thank you for attending the meeting, and let's bring questions in mailing list if there are any 19:02:59 <adetalhouet> #topic cookies 19:03:05 <adetalhouet> #endmeeting