Getting Started with Kubernetes

February 3, 2021

Kubernetes is a container orchestration platform. It was designed to manage the lifecycle of containers and services, while providing scalability, predictability, and availability.

Kubernetes architecture

Kubernetes is built in layers. Each higher layer abstracts away the complexity that is associated with the lower layer. Kubernetes cluster is made up of the following components:

  1. Master node acts as the gateway for the Kubernetes cluster. It performs the following tasks.

    • Exposes the API that users and client applications can connect to and communicate with the cluster.
    • It checks the health of other nodes within the cluster.
    • It schedules and assigns work to other nodes in the cluster.
  2. Worker nodes accept and run workloads assigned to them by the master node using the external and local resources.

  3. etcd is a distributed key-value store that persists cluster configuration, the status of objects, nodes on the cluster, and the nodes that the objects are scheduled to run on.

  4. kubelets communicate with the API server to determine which container workload has been assigned to the node. It is also responsible for spinning up pods to run the container workloads.

  5. kube-proxy makes it possible for containers to communicate with each other across various nodes within the cluster.

Note: For isolation, flexibility, and management of the cluster, Kubernetes runs applications and services in containers using the container runtime that is available within the node i.e Docker.

Kubernetes objects

Kubernetes objects are defined using YAML or JSON files that are referred to as the manifests.


The pod is the fundamental building block of Kubernetes. It comprises one or more tightly coupled containers with a shared networking layer and file system volume. Higher-level components manage pods therefore, they are not explicitly defined in the manifest file.



A deployment comprises a collection of pods defined by a template and a replica count. A specific value for replica count can be set, or a separate Kubernetes resource (horizontal pod autoscaller) can be used to control the replica count based on the CPU utilization.


If three pods are running in the deployment crash due to machine failure, three more pods will be scheduled to run on a different machine. Deployments are best for deploying stateless applications where pods can be replaced without causing a cluster problem.

Below is a YAML file that shows a simple definition for a deployment.

apiVersion: apps/V1
kind: Deployment
  name: rest-api-server
    app: api-server
  replicas: 10 # Defines the number of replicas in the deployment
      app: api-server
  template: # Defines how a pod should look like
        app: api-server
      containers: # Defines the containers that should be running in the pod
      - name: rest-api-server
        image: api-server
        - containerPort: 80


A service provides a static endpoint that can direct traffic to the desired pod even if the pods change due to updating or scaling.

The below YAML file shows how we can wrap a service around the deployment we created earlier.

apiVersion: v1
kind: service
  name: rest-api-service
    app: rest-api
  type: ClusterIP # Defines how we expose our endpoints
    app: rest-api # Defines the pods to direct traffic to.
  - protocol: TCP # Defines how clients communicate with our service.
    port: 80


Ingress makes it possible to expose our application running in a Kubernetes cluster to external traffic. Ingress makes it possible to define the services that should be public and the services that should remain private. The service for the user interface for our application can be public while the service running our backend application remains private.

Below is an example of an Ingress configuration that exposes the user interface service.

kind: Ingress
  name: product-ingress
  annotations: "nginx" / # Defines the configuration for the ingress controller.
    - http: # Defines how external traffic should access the service
        - path: /app
          backend: # Defines the service that the traffic is directed to.
            serviceName: user-interface-service
            servicePort: 80

Other Kubernetes objects

Volumes are used to manage the file system within the pods. Secret can be used when storing credentials, that is, passwords. ConfigMap is used to specify application configurations that are mounted into a file. HorizontalPodAutoscaler is used to scale pods based on resource utilization and availability automatically. StatefulSet is similar to deployment but used to deploy stateful applications such as databases to the Kubernetes cluster.

Installing Kubernetes locally

Kubernetes is bundled with Docker desktop which is available for both Windows and MacOS. One can find a tutorial on how to install Docker on Linux here.

To verify if your Kubernetes installation is working, run the command below.

$ kubectl version --client


Now that you have learned various components found in Kubernetes, run a Kubernetes cluster locally using minikube. The next article will cover deploying a Django application that we created here to a Kubernetes cluster running locally.

Peer Review Contributions by: Briana Nzivu