Expose 6653 port on each node through NodePort to allow OF Master/Standby connection

[Andrea-Campanella]

This patch exposes a node port for each replica of ONOS that gets deployed on a separate pod when exposeNodePorts is set to true in values.yaml
For example in this case 3 replicas were deployed on 3 nodes and the resulting node ports are:

onos-of-0                  NodePort    10.43.192.158   <none>        6653:31653/TCP                                 19m
onos-of-1                  NodePort    10.43.123.123   <none>        6653:31654/TCP                                 19m
onos-of-2                  NodePort    10.43.132.63    <none>        6653:31655/TCP                                 19m

This allows an aggregation switch to connect to all the instances of ONOS at the same time to achieve the MASTER/STANBY active connections that OF requires.
The switch needs to have the IP of one of the nodes and all the ports configured.
e.g.

OPT_ARGS="-d 2 -c 2 -c 4 -t 10.128.100.70:31653 -t 10.128.100.70:31654 -t 10.128.100.70:31655  -i $DPID"

A small caveat is that with this patch ONOS can't be automatically scaled up in number of instances.

IF your cluster is deployed in Kind there is a further step needed to expose those node ports.
A custom cfg needs to be given to the deployment of your kind cluster

kind: Cluster
apiVersion: kind.sigs.k8s.io/v1alpha3
nodes:
  - role: control-plane
    extraPortMappings:
    - containerPort: 31653
      hostPort: 31653
    - containerPort: 31654
      hostPort: 31654
    - containerPort: 31655
      hostPort: 31655
  - role: worker
  - role: worker

[Andrea-Campanella]

Thanks @kuujo @charlesmcchan I'll coordinate with the SEBA/VOLTHA team not to break tests and then merge this, most likely Monday.

[kuujo]

It's not really even that . doesn't work. It actually does work as intended. This is just a limitation of variable scopes in Go templates. It does make for some pretty messy workarounds. We probably won't get anything out of tracking the issue...

[Andrea-Campanella]

Agreed, but since it took me a little to understand what was going on I still think it's worth tracking, also to show people why I have done something like that and to help if in the future something similar.

[kuujo]

Yep it took me a while to understand how it worked too. It can be pretty frustrating.

[kuujo]

Yes let's be careful. I added the Do Not Merge tag to make sure nobody sees an approved PR and merges it. You can remove it when you're ready.

[Andrea-Campanella]

thanks @kuujo for the Do Not Merge. I'll take it out when we can merge based on upgrade of SEBA/VOLTHA PODs