04:30:58 <gongysh> #startmeeting tacker 04:30:59 <openstack> Meeting started Wed Jan 3 04:30:58 2018 UTC and is due to finish in 60 minutes. The chair is gongysh. Information about MeetBot at http://wiki.debian.org/MeetBot. 04:31:00 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 04:31:02 <openstack> The meeting name has been set to 'tacker' 04:33:53 <gongysh> YanXingAn, hi 04:34:26 <YanXingAn> hi 04:34:58 <gongysh> phuoc, hi 04:35:09 <phuoc> hi Mr.gongysh 04:35:20 <gongysh> #topic bp 04:35:24 <gongysh> lets start 04:35:32 <gongysh> with container vnf 04:35:41 <gongysh> phuoc, http://paste.openstack.org/show/636966/ 04:36:03 <longkb_> hi gongysh 04:36:18 <gongysh> please look at my c-vnf descriptor 04:36:55 <phuoc> I will update it 04:37:20 <gongysh> this way will keep the vnfd almost the same 04:37:28 <phuoc> original, I intend to support multiple containers in 1 VDU 04:37:45 <phuoc> but I will update with 1 container in 1 VDU 04:38:05 <phuoc> and provide 1 example about deploy c-VNF with real case 04:39:30 <gongysh> how do you mapping vnf resources to k8s elements? 04:39:51 <gongysh> vnf is a k8s service, vdu is a pod? 04:40:20 <phuoc> VNF is made of many VDU 04:40:38 <gongysh> phuoc, right. 04:41:23 <gongysh> and how do you map to k8s elements? 04:41:37 <phuoc> each of VDU, I define it by a Service, Deployment (deployment have many pods) and a Horizon Pod Auto-scaler (optional) 04:44:14 <gongysh> vnf can have many vdus. k8s service can have many pods, k8s pod can have many containers. 04:44:17 <phuoc> And VDU is exposed by Service (it looks like load balancer to Pods) 04:46:53 <gongysh> phuoc, do you think your spec is clear about the mapping? 04:47:10 <gongysh> if that is, I can read it again. 04:47:41 <phuoc> gongysh, I will update my spec to make it clear 04:48:00 <gongysh> phuoc, please make sure the mapping is clear and then do coding, otherwise, you probably waste your time. 04:48:01 <phuoc> because there are many ways to define c-VNF in Kubernetes 04:48:06 <phuoc> my spec is one of them 04:48:37 <phuoc> gongysh, thanks 04:48:52 <gongysh> phuoc, in openshift, the pod can have many containers. 04:49:37 <phuoc> gongysh, it is the same to kubernetes, that is why I try to define VNFC 04:50:07 <phuoc> 1 VDU have many VNFCs (VNFC is defined by a container) 04:51:57 <YanXingAn> phuoc: A VNFC have only one VM? 04:52:54 <gongysh> phuoc, https://review.openstack.org/#/c/529116/2/samples/tosca-templates/vnfd/tosca-vnfd-containerized.yaml, why use a new configurable_properties instead of vnfc? 04:54:52 <phuoc> YanXingAn, I tried to map VNFC definition to k8s object 04:55:11 <phuoc> I think it is better to map VNFC to a container in Pod 04:55:37 <phuoc> gongsyh, I followed to this document from TOSCA http://docs.oasis-open.org/tosca/tosca-nfv/v1.0/csd04/tosca-nfv-v1.0-csd04.html#_Toc482896065 04:56:27 <phuoc> it is only one document, they talk about VNFC 04:58:39 <gongysh> it seems there is no independent node type to define the vnfc 05:00:06 <gongysh> how about changing https://review.openstack.org/#/c/529116/2/samples/tosca-templates/vnfd/tosca-vnfd-containerized.yaml 05:00:19 <gongysh> configurable_properties: 05:00:19 <gongysh> redis-master: 05:00:20 <gongysh> properties: 05:00:20 <gongysh> num_cpus: 0.5 05:00:22 <gongysh> mem_size: 512 MB 05:00:22 <gongysh> into: 05:00:32 <gongysh> nfvc: 05:02:20 <gongysh> phuoc , http://paste.openstack.org/show/637006/ 05:03:16 <phuoc> actually, there are no definition in TOSCA about VNFC 05:03:31 <phuoc> so, I think about your advice http://paste.openstack.org/show/636966 05:04:04 <phuoc> we will have simple case, 1 VDU is defined a container 05:04:09 <gongysh> 636966 is one vdu has one container. 05:04:27 <gongysh> ok, lets do it first. 05:04:44 <phuoc> Kompose project, they also define one container for one pod 05:05:13 <phuoc> so, we can't need VNFC 05:06:30 <gongysh> I mean since your current patch is trying to do multiple containers, we can do it 05:07:05 <gongysh> 637006 is justing change 'configurable_properties' into 'vnfcs'. 05:07:52 <gongysh> 637006 is just changing 'configurable_properties' into 'vnfcs'. 05:08:03 <phuoc> gongysh, I got it 05:08:05 <phuoc> thank you 05:08:54 <gongysh> phuoc, so what is your next direction? 05:09:30 <phuoc> after this spec? 05:09:44 <gongysh> In this spec. 05:10:02 <gongysh> http://paste.openstack.org/show/636966 or http://paste.openstack.org/show/637006/ ? 05:10:34 <phuoc> maybe I will go with multiple containers 05:10:55 <phuoc> but I have some issues with volumes in Kubernetes 05:11:19 <gongysh> phuoc, ok with multiple containers. 05:11:43 <phuoc> https://wiki.onap.org/pages/viewpage.action?pageId=16007890 05:12:09 <phuoc> this is one proposal project in ONAP to do the same thing 05:12:27 <phuoc> I will update Tacker if they have better solution 05:13:04 <gongysh> phuoc, you can let volume problem to next phase. 05:13:08 <gongysh> phuoc thanks 05:13:22 <gongysh> #info https://wiki.onap.org/pages/viewpage.action?pageId=16007890 05:13:32 <gongysh> next bp 05:13:35 <gongysh> longkb_, hi 05:13:51 <xuan0802> hi 05:14:00 <gongysh> xuan0802, hi 05:14:08 <gongysh> I have seen the bp and its codes. 05:14:11 <longkb_> Hi, I and xuan0802 are working on bp: https://blueprints.launchpad.net/tacker/+spec/policy-based-vnf-cluster 05:14:20 <gongysh> have you guys tested that? 05:14:45 <gongysh> maybe you can record a demo before next meeting? 05:14:46 <xuan0802> all functions as spec have been tested 05:15:04 <xuan0802> however, we will add functional testcases soon 05:15:16 <gongysh> xuan0802, that's fine. thanks 05:15:38 <gongysh> xuan0802, if you can record a demo, it will be cool. 05:15:39 <longkb_> From your point of view, do we need to update our spec? 05:16:34 <longkb_> Ok, I will make a video demo and show to you in tacker channel 05:16:35 <gongysh> I cannot get a picture of how it will work if I just review the spec. 05:17:30 <gongysh> longkb_, thanks 05:18:02 <gongysh> #topic open discussion 05:18:23 <gongysh> do you guys have stuff to talk? 05:18:41 <longkb_> How about multi-site scenario in Tacker? 05:18:59 <longkb_> Do we have a plan to integrate with Tricircle? 05:19:12 <gongysh> longkb_, I see your spec relates to this. 05:20:31 <gongysh> longkb_, to work with tricircle, we need add some properties to our VLD elements. 05:21:17 <longkb_> yep, I see 05:21:19 <gongysh> of course, if we don't create VL, I mean use the existed neutron networks, it should work with tricircle now. 05:23:21 <gongysh> longkb_, if you think there are some we need to do in order to work with tricircle, you are welcome to file a bp and impement it. 05:23:31 <gongysh> impement -> implement 05:24:30 <longkb_> Thanks gongysh, I will think about it. Hence, I will make a bp that relates to Tricircle :) 05:24:45 <gongysh> anything else? 05:25:09 <gongysh> ok. lets end this meeting 05:25:15 <gongysh> #endmeeting