Visualize Your Kubernetes Cluster

Abstract

This project contains 3 parts to visualize your k8s graph.

Docker container is already available on DockerHub so you can skip
part 1. and 2. and just create the k8s deployment.

1. Node server, that polls the information from k8s api and creates
   a graph using D3.
2. Dockerfile, that creates the container.
3. k8s yaml file, that will create a service and a deployment with two containers.
   1. kubectl started with proxy, so the node server can
      access the k8s api.
   2. Node server, that renders the graph using D3.

Node server

Relevant files for the node server are server.js, k8s.html
and package.json.

The node server polls - by default - every second the k8s api and
extracts the relevant information to create the graph visualization
which is implemented in the html file.

The following list shows the values,
that can be configured via env variables:

• masterSize: 30
• minionSize: 30
• podSize: 15
• linkSizePodToMinion: 500
• linkSizeMinionToMaster: 800
• dummyNodes: 0 - Create dummy nodes for debugging purposes
• podsApiUrl: ‘http://127.0.0.1:8001/api/v1/namespaces/default/pods‘
• pollingIntervalInSeconds: 1

Dockerfile

Yeah, this is the Dockerfile that creates the docker container with
the node server. Docker container is published at DockerHub rilleralle/k8s-graph

K8s Yaml File

Yaml file creates one service and one deployment - with two container.

Service:

By default the service type is NodePort.
If you are running k8s on a cloud service like AWS or Google
you can also use aLoadBalancer.
Read more about Service Types.

Deployment:

Creates a deployment with two container.

1. kubectl container, that starts with proxy command so
   the node server can access the k8s api to fetch the required
   information to render the graph.
2. Node server container. Docker image is available on DockerHub docker pull rilleralle/k8s-graph
   Set env variables to configure node server. See section Node server

Create deployment

$ kubectl apply -f k8s-graph.yaml

RBAC

With K8S 1.9.2 the default user does not have access to the API any more. Please create a service account that has read access for the endpoints nodes and pods. Use this account to execute the graph container.
You can also bypass the RBAC by making the default user a cluster-admin.
Please be aware, that this could be a secutiry issue!

kubectl create clusterrolebinding kube-system-default-admin   --clusterrole=cluster-admin --serviceaccount=default:default