Deploying vFw and EdgeXFoundry Services on Kubernets Cluster with ONAP

vFW Helm Chart link:

https://github.com/onap/multicloud-k8s/tree/master/kud/demo/firewall

EdgeXFoundry Helm Chart link:

https://github.com/onap/multicloud-k8s/tree/master/kud/tests/vnfs/edgex/helm/edgex

Create CSAR with Helm chart as an artifact

The CSAR is a heat template package with Helm chart in it. Basic package consists of an environment file, base_dummy.yaml file (for the sake of example) and MANIFEST.json and the tar.gz file (of Helm chart). We need to zip all of these files before onboarding. 

One thing to pay much attention to is the naming convention which must be followed while making the tgz.

NOTE: The Naming convention is for the helm chart tgz file.

Naming convention follows the format:

<free format string>_cloudtech_<technology>_<subtype>.extension

1.       Cloudtech is a fixed pattern and should not be changed if not necessary
2.      Technology: k8s, azure, aws, ….
3.       Subtype:  charts, day0, configtemplate, ...
4.      Extension: zip, tgz, csar, …”

NOTE: The .tgz file must be a tgz created from the top level helm chart folder. I.e a folder that contains a Chart.yaml file in it. For vFW use case the content of the tgz file must be following

Listed below is an example of the contents inside a heat template package

vfw_k8s_demo.zip file is a zip of the 4 other files( manifest.json, base_dummy.env,  base_dummy.yaml, vfw_cloudtech_k8s_charts.tgz)  gets on boarded onto SDC.

MANIFEST.json

Key thing is note the addition of cloud artifact

type: "CLOUD_TECHNOLOGY_SPECIFIC_ARTIFACTS"

base_dummy.yaml

Designed to be minimal HEAT template.

base_dummy.env

Onboard the CSAR

For onboarding instructions please refer to steps 4-9 from the document here.

Distribute the CSAR

On onboarding, a service gets stored in SDC and as a final action, it is distributed to SO and other services. When distribution happens it takes tar.gz file and uploads to k8s plugin.

Steps for installing KUD Cloud

Follow the link to install KUD Kubernetes Deployment.  KUD contains all the packages required for running vFw Usecase.

     Kubernetes Baremetal deployment setup instructions

REGISTER KUD CLOUD REGION with K8s-Plugin

API to support Reachability for Kubernetes Cloud

The command to POST Connectivity Info

This is a multipart upload and here is how you do the POST for this.

#Using a json file (eg: post.json) containing content as above;

Command to GET Connectivity Info

Command to DELETE Connectivity Info

Command to UPDATE/PUT Connectivity Info

Update K8sConfig

Workaround for R4 Dublin. This step will not be needed from R5.

Edit the configMap helm-release-name-multicloud-k8s for K8s plugin to make changes to the config like below to add ovn-central-address:

{
"ca-file": "/opt/multicloud/k8splugin/certs/root_ca.cer",
"server-cert": "/opt/multicloud/k8splugin/certs/multicloud-k8s.pub",
"server-key": "/opt/multicloud/k8splugin/certs/multicloud-k8s.pr",
"password": "c2VjcmV0bWFuYWdlbWVudHNlcnZpY2VzZWNyZXRwYXNzd29yZA==",
"database-type": "mongo",
"database-address": "multicloud-k8s-mongo",
"etcd-ip": "multicloud-k8s-etcd",
"plugin-dir": "/opt/multicloud/k8splugin/plugins",

 "ovn-central-address": "<IP address of the Kubernetes controller>:6641"
}

(the configMap is based on  oom/kubernetes/multicloud/charts/multicloud-k8s/resources/config/k8sconfig.json )

Restart the Multoclod-K8s Plugin for the changes to take effect.

Register KUD Cloud region with AAI

With k8s cloud region, we need to add a tenant to the k8s cloud region.  The 'easy' way is to have the ESR information (in step 1 of cloud registration) point to a real OpenStack tenant (e.g. the OOF tenant in the lab where we tested).  

This will cause multicloud to add the tenant to the k8s cloud region and then, similar to #10 in the documentation here , the service-subscription can be added to that object.

NOTE: use same name cloud-region and cloud-owner name

An example is shown below for K8s cloud but following the steps 1,2,3 from here. The sample input below is for k8s cloud type.

Step 1 - Cloud Registration/ Create a cloud region to represent the instance.

Note: highlighted part of the body refers to an existing OpenStack tenant (OOF in this case).  Has nothing to do with the K8s cloud region we are adding.

Step 2 – add a complex to the cloud

Note:  just adding one that exists already

Step 3 - Trigger the Muticloud plugin registration process

This registers the K8S cloud with Multicloud – it also reaches out and adds tenant information to the cloud (see example below – you’ll see all kinds of flavor, image information that is associated with the OOF tenant).

If we had not done it this way, then we’d have to go in to AAI at this point and manually add a tenant to the cloud region.  The first time I tried this (k8s region one), I just made up some random tenant id and put it in.)

The tenant is there so you can add the service-subscription to it:

Making a Service Type:

Add subscription to service type to the customer (Demonstration in this case – which was already created by running the robot demo scripts)

Add Service-Subscription to the tenant (resource-version changes based on actual value at the time):

4. Addition of Cloud Region to SO Catalog DB

See starting around 4:45 of the video -  https://wiki.onap.org/download/attachments/64006768/vfwk8s_cloud_registration_720.mp4

Additional SO Configuration

There is a configuration needed for SO – it is described here in the docs (in step 4): https://onap.readthedocs.io/en/latest/submodules/integration.git/docs/docs_vfwHPA.html#docs-vfw-hpa

But also replicated below for convenience:

Modify the SO bpmn configmap to change the SO vnf adapter endpoint to v2

oom-rancher#    kubectl -n onap edit configmap dev-so-so-bpmn-infra-app-configmap

- vnf:

          endpoint: http://so-openstack-adapter.onap:8087/services/VnfAdapter

          rest:

            endpoint: http://so-openstack-adapter.onap:8087/services/rest/v1/vnfs

+ vnf:

          endpoint: http://so-openstack-adapter.onap:8087/services/VnfAdapter

          rest:

            endpoint: http://so-openstack-adapter.onap:8087/services/rest/v2/vnfs

Then delete the bpmn pod

oom-rancher#  kubectl delete <pod-name> -n onap

Create Profile Manually

K8s-plugin artifacts start in the form of Definitions. These are nothing but Helm Charts wrapped with some metadata about the chart itself. Once the Definitions are created, we are ready to create some profiles so that we can customize that definition and instantiate it in Kubernetes.

(NOTE: Refer this link for complete API lists and documentation: MultiCloud K8s-Plugin-service API)

A profile contains the following:

1. manifest.yaml
1. Contains the details for the profile and everything contained within
2. A HELM values override yaml file.
1. It can have any name as long as it matches the corresponding entry in the manifest.yaml
3. Any number of files organized in a folder structure
1. All these files should have a corresponding entry in manifest.yaml file

Creating a Profile Artifact

The manifest file contains the following

Note: values: "values_override.yaml" can be empty file if you are creating a dummy profile

Note: A dummy profile does not need any customization. The following is optional in the manifest file.

We need to read the name of the Definition which was created while distribution of the service from SDC.

Command to read the Definition name and its version

On the ONAP K8s Rancher host execute following statement 

From the output read the name of the definition which is "rb-name" and "rb-version" respectively

Command to read (GET) Definition

With this information, we are ready to upload the profile with the following JSON data

 Command to create (POST) Profile

Command to UPLOAD artifact for Profile

Command to GET Profiles

Command to DELETE Profile

Instantiation

Instantiation is done by SO. SO then talks to Multi Cloud-broker via MSB and that in turn looks up the cloud region in AAI to find the endpoint. If k8sregion one is registered with AAI and SO makes a call with that, then the broker will know that it needs to talk to k8s-plugin based on the type of the registration.

This video shows the whole sequence of instantiation using VID:

https://wiki.onap.org/download/attachments/64006768/vfwk8s_deploy_delete_720.mp4

Create User parameters

In the VID user parameters are created in the following format during vfModule creation:

Known Issues:

• Use the vFw Helm chart from the Master branch - (https://github.com/onap/multicloud-k8s/tree/master/kud/demo/firewall)
• Artifact broker version issue (Libo to update this section)
  • please refer to the description in  MULTICLOUD-749 - Getting issue details... STATUS
• Recommended way tar the resource bundle image is tar.gz. helm package is not supported in Dublin.
• In Dublin Get all Definitions is not supported in K8s Plugin API's. Bug is filed for this.

Commands for instantiation using K8s API:

These commands can be used to interact directly with K8s Plugin with SO. 

Command to Instantiate a Profile

Delete Instantiated Kubernetes resources

The id  field from the returned JSON can be used to DELETE the resources created in the previous step. This executes a Delete operation using the Kubernetes API.

GET Instantiated Kubernetes resources

The id field from the returned JSON can be used to GET the resources created in the previous step. This executes a get operation using the Kubernetes API.