Skip to content

Deploying µONOS micro-services with HELM

One of the goals of the µONOS project is to provide simple deployment options that integrate with modern technologies. Deployment configurations can be found in the onos-helm-charts repository. Each onos service has a directory containing its chart. As an example the onos-config chart is in onos-helm-charts/onos-config.

Deploying on Kubernetes with Helm

Helm is a package manager for Kubernetes that allows projects to provide a collection of templates for all the resources needed to deploy on k8s. ONOS Config provides a Helm chart for deploying a cluster for development and testing. In the future, this chart will be extended for production use.

Individual charts or overarching (umbrella) chart

The individual components of µONOS may be deployed one at a time, or altogether through an overarching (unbrella) Helm chart, or some combination of both.

In all cases the prerequisites must be satisfied:

  • Creation of a namespace
  • deployment of Atomix controller in the namespace.

The individual components in the umbrella chart are:

  • onos-topo:
  • onos-config:
  • onos-cli:
  • onos-gui:
  • onos-ric:
  • onos-ric-ho:
  • onos-ric-mlb:
  • ran-simulator:
  • nem-monitoring:

The choice of which of these is deployed can be chosen at deployment time with an option like:

--set import.<component>.enabled=true

In this way all, none or some of the components can be deployed together.

Resources

The Helm chart provides resources for deploying the config service and accessing it over the network, both inside and outside the k8s cluster:

  • Deployment - Provides a template for ONOS Config pods
  • ConfigMap - Provides test configurations for the application
  • Service - Exposes ONOS Config to other applications on the network
  • Secret - Provides TLS certificates for end-to-end encryption
  • Ingress - Optionally provides support for external load balancing

Deployment options

Local Deployment Setup with KinD

To deploy the Helm chart locally: 

brew install kubernetes-helm
  • Once Kind has been installed, start it with 
kind create cluster
  • Once Kind has started, set your environment to the Kubernetes cluster:
export KUBECONFIG="$(kind get kubeconfig-path --name="kind")"

Bare metal deployment

ONOS can also be deployed on a bare metal cluster provisioned with Rancher or equivalent.

Kubectl and Helm are can be run from your local PC to control the remote cluster.

Prerequisites

For any deployment scenario a number of steps must be performed first.

The steps below assume the KUBECONFIG environment variable to point kubectl to your cluster.

Add the "CORD" Helm chart repo

The Prometheus and Grafana installations are derived from the CORD Helm charts. Run:

helm repo add cord https://charts.opencord.org

Check out the Helm charts

The helm charts need to be present on your PC. Run:

git clone https://github.com/onosproject/onos-helm-charts && cd onos-helm-charts

Configure the micro-onos namespace

The various onos services can be deployed to any namespace.

A "namespace" partitions the cluster in to independent islands. For consistency between documentation we are going to use the micro-onos here.

To create the micro-onos namespace run:

kubectl create namespace micro-onos

Deploy Atomix Controller

The various onos services leverage Atomix as the distributed store for HA, scale and redundancy. The first thing that needs to be deployed in any onos deployment is the Atomix go controller. To deploy the Atomix controller do:

helm install atomix-controller atomix-controller -n micro-onos --set scope=Namespace

If you watch the pods you should now see:

$ kubectl -n micro-onos get pods
NAME                                 READY   STATUS    RESTARTS   AGE
atomix-controller-6bb9555f48-fk8kb   1/1     Running   0          12s

Deploy the SD-RAN set of onos services

A complete set of onos services can be deployed with just the sd-ran chart.

Run the build of dependent charts:

helm dep build sd-ran

Finally, run the install:

helm -n micro-onos install sd-ran sd-ran . \
    --set onos-ric.store.controller=atomix-controller.micro-onos.svc.cluster.local:5679 \
    --set onos-ric.store.raft.backend.image=atomix/local-replica:latest

this will deploy onos-ric, onos-ric-ho, onos-ric-mlb, ran-simulator, onos-topo, onos-cli and onos-gui, but not onos-config (as it's not currently needed for SD-RAN).

It should give an output like:

NAME: sd-ran
LAST DEPLOYED: Tue Apr  7 15:56:45 2020
NAMESPACE: micro-onos
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
Thank you for installing sd-ran Helm chart.

Your release is named sd-ran in namespace micro-onos.
See https://docs.onosproject.org/developers/deploy_with_helm/

To learn more about the release, try:
  $ helm -n micro-onos status sd-ran
  $ helm -n micro-onos get all sd-ran
  $ watch kubectl -n micro-onos get pods

You can attach to:
* Onos CLI pod with
$ kubectl -n micro-onos exec -it $(kubectl -n micro-onos get pods -l type=cli -o name) -- /bin/sh
* SD-RAN GUI at http://<server_IP>:31180
* Prometheus at http://<server_IP>:31301/targets
* Grafana at http://<server_IP>:31300 (admin/strongpassword)

If you are using KinD as a Kubernetes server, you will have to use a "port-forward" to access the GUI, Grafana and Prometheus e.g.
$ kubectl -n micro-onos port-forward $(kubectl -n micro-onos get pods -l type=gui -o name) 8182:80
and then access the GUI at
* http://localhost:8182

Alternatively to install a cluster where you are not interested in SD-RAN and only want onos-config, you could run

helm -n micro-onos install sd-ran sd-ran \
     --set onos-ric.store.controller=atomix-controller.micro-onos.svc.cluster.local:5679 \
     --set onos-ric.store.raft.backend.image=atomix/local-replica:latest \
     --set import.onos-config.enabled=true \
     --set import.onos-ric.enabled=false \
     --set import.onos-ric-ho.enabled=false \
     --set import.onos-ric-mlb.enabled=false \
     --set import.ran-simulator.enabled=false \
     --set import.nem-monitoring.enabled=false

Maintenance

To delete the deployment issue:

helm delete -n micro-onos sd-ran

If you make changes to one of the charts and want to re-deploy, please first issue

helm dependency update sd-ran

Deploy single services services

You can also deploy each service by itself. Please refer to each service's deployment file to get the exact command for each helm chart. Example for onos-config.