| *** dangtrinhnt_ has quit IRC | 00:36 | |
| *** dangtrinhnt has joined #openstack-fenix | 00:36 | |
| *** csatari_ has joined #openstack-fenix | 00:55 | |
| *** csatari has quit IRC | 01:02 | |
| *** csatari_ is now known as csatari | 01:02 | |
| hyunsikyang | Good morning:) OK. I am sorry to late reply. | 01:12 |
|---|---|---|
| tojuvone | hyunsikyang, morning :) np, it happens | 02:32 |
| hyunsikyang | tojuvone Actually, My question is that does fenix need the information about VNF or Host or not? | 08:07 |
| tojuvone | in OpenStack, Fenix maps project_id to VNF | 08:09 |
| tojuvone | As Fenix sends project_id specfific notifications with callbacks, that is all it needs | 08:09 |
| tojuvone | Fenix knows if VNF / project_id is capable of Fenix interaction as thre exists poriject_ip specific subscription in AODH if it is so | 08:11 |
| tojuvone | project_id | 08:11 |
| tojuvone | However, this might change in future | 08:11 |
| tojuvone | Instead of VNF registering subscription to AODH, it might do so directly to Fenix | 08:12 |
| tojuvone | Then VNF will directly register VNF / project_id specific API in VNFM for any interaction | 08:12 |
| tojuvone | Anyhow this is not the current thing and depends a bit of ETSI NFV standardization discussions | 08:13 |
| hyunsikyang | YES. In the case of VNF, VNFM fenix plugin send it to fenix directly and it make sense. | 08:13 |
| tojuvone | but then you asked about hosts | 08:13 |
| hyunsikyang | both. | 08:14 |
| tojuvone | Fenix need to know every host and isntances on top | 08:14 |
| hyunsikyang | Yes so how? | 08:14 |
| tojuvone | In openStack this information comes from Nova to Fenix | 08:14 |
| tojuvone | and hosts under mainteancne/upgrade... | 08:14 |
| hyunsikyang | ah. Ok | 08:14 |
| tojuvone | are either given by infrastructure admin making the maintenance session | 08:14 |
| tojuvone | or | 08:14 |
| hyunsikyang | In the case of Host, fenix get the current info from nova. | 08:14 |
| hyunsikyang | or asking it. | 08:15 |
| tojuvone | he can leave it "blank", so Fenix discovers all the hosts | 08:15 |
| tojuvone | like controllers and computes nova knows | 08:15 |
| tojuvone | Yes, that is the use case where Fenix workflow is meant for Nova specif control | 08:16 |
| tojuvone | Workload and action plugins would need to be different in different cloud | 08:16 |
| tojuvone | or made to support other kinds of clouds | 08:16 |
| tojuvone | to have the same like from Kubernetes | 08:17 |
| hyunsikyang | Yes but, in the case of VNF? Whatif i do not send VNF info directly to the fenix, How Fenix start maintenance with specific VNF? | 08:17 |
| tojuvone | VNF either support Fenix or not | 08:18 |
| tojuvone | if VNF is not subscribed.. | 08:18 |
| tojuvone | there can only be some predefined actions on that VNF instances | 08:19 |
| tojuvone | or if seen as totally Telco cloud | 08:19 |
| tojuvone | Fenix could simple failt the maintenance | 08:19 |
| tojuvone | and somebody would need to fix his VNF with the support :D | 08:19 |
| tojuvone | That is surely up to operator | 08:20 |
| hyunsikyang | yes so we just assume that Fenix know the VNF info? | 08:20 |
| hyunsikyang | I just consider procedure of forwarding VNF info which have maintenance option... | 08:20 |
| tojuvone | 1. Fenix knows in is subscribed to AODH | 08:20 |
| tojuvone | 2. Fenix has default behaviour if not subscribed | 08:21 |
| tojuvone | 3. Fenix will fail, if not subscribed | 08:21 |
| tojuvone | For now the maintenance option is just subscribing VNF to AODH | 08:22 |
| tojuvone | Otherwise the subscription might come to Fenix later with something I have preliminary described here: | 08:23 |
| tojuvone | https://drive.google.com/file/d/1cByI-MPo_a4-wrXrCFU_ssQmi5f7rjbo/view?usp=sharing | 08:23 |
| tojuvone | That is some input for ETSI NFV discussion in PTG: https://etherpad.openstack.org/p/DEN2019-fenix-ETSI-NFV-PTG | 08:24 |
| tojuvone | and there is a lot of details VNF will set as constraints. | 08:24 |
| hyunsikyang | Ok i will check it too:) | 08:24 |
| tojuvone | that would make the current interaction a bit less | 08:25 |
| tojuvone | but Fenix workflow tons of more complex | 08:25 |
| hyunsikyang | Actually, tacker doesn't follow the definition of ETSI such as IFA or SOL.. | 08:25 |
| hyunsikyang | It needs. | 08:25 |
| tojuvone | yes, let's see how it goes | 08:27 |
| tojuvone | perhaps not impossible to keep both ways, but might it be feasible if looking generic definitions on VNF / EM / VNFM | 08:28 |
| *** tojuvone has quit IRC | 08:30 | |
| *** tojuvone has joined #openstack-fenix | 08:31 | |
| *** tojuvone has quit IRC | 09:48 | |
| *** tojuvone has joined #openstack-fenix | 09:49 | |
| *** JangwonLee_ has quit IRC | 09:56 | |
| *** JangwonLee_ has joined #openstack-fenix | 09:56 | |
| *** tojuvone has quit IRC | 10:57 | |
| *** tojuvone has joined #openstack-fenix | 10:57 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!