To scale, the MANO will be sending on the topic for cnf deployment and cnf removal. The k8s wrapper needs to be smart with these messages to see when it is a scale event.
adding CNF
topic: infrastucture.k8s.function.deploy
payload example:
service_instance_id: ff0757e7-0d29-417d-8b63-cba80b6ffd9d
vim_uuid: 88888888-2222-3333-4444-888888888888
vnfd:
author: 5GTANGO Test Team
cloudnative_deployment_units:
- connection_points:
- fip: true
id: data
network_id: ff0757e7-0d29-417d-8b63-cba80b6ffd9d_ns_data-2-cnfs
port: 1883
id: cdu01-dfe39daa-0538-4133-8de1-d3cf6b43d14a
image: ubuntu:trusty
connection_points:
- id: data
interface: ipv4
port: 1883
type: serviceendpoint
description: This is a integration test artifact
descriptor_schema: https://raw.githubusercontent.com/sonata-nfv/tng-schema/master/function-descriptor/vnfd-schema.yml
function_specific_managers:
- description: FSM to do some dummy config on the ubuntu image for testing purposes
id: ubuntuk8s
image: tsoenen/ubuntuk8s-fsm
options:
- key: type
value: start
- key: type
value: configure
instance_uuid: dfe39daa-0538-4133-8de1-d3cf6b43d14a
name: cnf1
uuid: 2dcbf6f1-8019-404e-af24-d59b77fb9f97
vendor: eu.5gtango
version: '0.1'
virtual_links:
- connection_points_reference:
- data
- cdu01:data
connectivity_type: E-Tree
id: data-link
When the k8s wrapper receives a message on the .deploy topic, it needs to use the service_instance_id, the vnfd (e.g. image urls) and vim_uuid to check if there is already a pod for this vnfd part of the service instance. If it is, we just need another replica for this pod. If it is not, regular deployment needs to be followed. The response from the wrapper on this message should be the same for both cases.
Removing CNF
topic: infrastructure.k8s.function.remove
example payload:
service_instance_id:
vim_uuid:
vnf_uuid:
When reciving requests to remove a cnf, the K8s wrapper should check which pod is linked to this CNF (based on vim, vnf and service instance id) and check the amount of replicas. If higher than 1, reduce the number by 1. If 1, remove the pod. Responses to the MANO should be similar for both cases. Currently, the functionality for both cases is missing, for now this topic is not correctly handled. A response to a request on this topic looks like this
felipevicens
commented
5 years ago
To scale, the MANO will be sending on the topic for cnf deployment and cnf removal. The k8s wrapper needs to be smart with these messages to see when it is a scale event.
adding CNF
payload example:
When the k8s wrapper receives a message on the .deploy topic, it needs to use the service_instance_id, the vnfd (e.g. image urls) and vim_uuid to check if there is already a pod for this vnfd part of the service instance. If it is, we just need another replica for this pod. If it is not, regular deployment needs to be followed. The response from the wrapper on this message should be the same for both cases.
Removing CNF
example payload:
When reciving requests to remove a cnf, the K8s wrapper should check which pod is linked to this CNF (based on vim, vnf and service instance id) and check the amount of replicas. If higher than 1, reduce the number by 1. If 1, remove the pod. Responses to the MANO should be similar for both cases. Currently, the functionality for both cases is missing, for now this topic is not correctly handled. A response to a request on this topic looks like this