Kubernetes – An Open-Source Container Management Framework

API Server

• The heart of any Kubernetes cluster.
• It is a REST API.
• It is also stateless.
• The default port it listens to is port 6443 (unlike 80 (http) and 443 (https) requests).
• The resources in Kube-apiserver are:
• Pod
• ReplicaSet
• PersistentVolume
• NetworkPolicy
• Deployment

Note: all are related to container management only.

Etcd

• Distributed NoSQL database
• The main memory of the Kubernetes cluster
• Independent project of Kubernetes
• No Alternative option to Etcd for data storage
• Apiserver calls will read or write data from or to Etcd
• Cluster state information
• Single point of failure - Etcd crashes then Kubernetes cluster is inaccessible
• Not an In-Memory DB. Data is stored on the Disk

Scheduler

• Responsible for electing a worker node.
• Scheduler continuously queries the Apiserver at regular intervals to list the pods that have not been scheduled.
• Each pod object stored in Etcd has a property called 'nodeName' - it is the name of the worker node where the pod will deploy (nodeName assigned means the pod is scheduled).
• The scheduler is responsible for assigning the name of the worker node for each pod.

Controller-Manager

• Tries to maintain the actual state of the cluster as updated in the Etcd datastore.
• It performs garbage collection of pods, nodes, events, etc on the cluster.
• To handle multiple responsibilities different controllers are called.
• List of the few controllers
• NodeController
• NamespaceController
• EndpointsController
• ServiceaccountController

Kubelet

• The most important component of the worker node
• Interacts with the local Docker daemon on the worker node
• Kubelet is required to run on the host machine
• Kubelet refers to the configuration file ~/kubernetes/kubelet.conf
• Updates the 2 parameters in the config file:
• The endpoint of the Kube-apiserver component
• The local Docker daemon UNIX socket
• Kubelet is a bridge between Apiserver and the local container (Docker) daemon
• Kubelet runs a GET request at an interval of 20 seconds to the checklist of the pods created on Etcd.
• The kubelet performs garbage collection
• Verifies for unused images every five minutes
• Verifies for unused containers every minute

Kube-Proxy

• Handles all the networking related tasks on Kubernetes worker node for containers
• Enables access to the running pods to other pods or external applications
• The Proxy uses a feature called Service
• Services route traffic to Pods
• Maintains network rules on nodes to manage and allow network communication

Container Engine

• Kubernetes is a container management tool
• The Default Container engine is Docker
• Common container runtimes:
• Docker Engine
• Contained
• CRI-O
• Mirantis Container Runtime

Kubernetes Installation on Windows

Installation Steps:

1: Check if Virtualization is supported in your machine

$ systeminfo

2: Download/Install kubectl, minikube utility & Add it to the Local ENV PATH

https://kubernetes.io/docs/tasks/tools/install-kubectl-windows/

https://github.com/kubernetes/minikube/releases

• a. Download kubectl.exe file and place it in a dedicated folder (minikube folder).

  Set the environment variables

  Test the kubectl

  open cmd prompt (or power shell)

  type: kubectl and enter

• b. Download the file from github link share above

  minikube-windows-amd64.exe

  place the file in the folder where we have placed kubectl file (minikube folder).

3: Download/Install a Hypervisor | ORACLE VM BOX

https://www.virtualbox.org/wiki/Downloads

4: Start the Minikube (Specify the Driver)

syntax: $ minikube start --driver=driver_name

$ minikube start

or

$ minikube start --driver=virtualbox

5: To verify the Minikube installation

$ minikube status

minikube

type: Control Plane

host: Running

kubelet: Running

apiserver: Running

kubeconfig: Configured

6: Create & Expose a container in the Minikube Cluster

syntax: $ kubectl create deployment --image= [options]

> kubectl get pod

No resources found in default namespace.

> kubectl get deployment

No resources found in default namespace.

> kubectl create deployment test1-v1 --image=spark

deployment.apps/test1-v1 created

> kubectl get deployment

NAME READY UP-TO-DATE AVAILABLE AGE

test1-v1 0/1 1 0 28s

> kubectl get pod

NAME READY STATUS RESTARTS AGE

test1-v1-56f4d5f5b4-qm7qg 0/1 ErrImagePull 0 41s

> kubectl get service

NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE

kubernetes ClusterIP 10.96.0.1 443/TCP 2d17h

7: Test or Access the Container

$ minikube service test1-v1 --url

8: Cleanup

$ kubectl delete pods test1-v1

$ kubectl delete service test1-v1

$ minikube stop

$ minikube delete

Refer to a few kubectl commands

kubectl get nodes

kubectl get pod

kubectl get services

kubectl create deployment nginx-depl --image=nginx

kubectl get deployment

kubectl get replicaset

kubectl edit deployment nginx-depl

### debugging

kubectl logs {pod-name}

kubectl exec -it {pod-name} -- bin/bash

### create mongo deployment

kubectl create deployment mongo-depl --image=mongo

kubectl logs mongo-depl-{pod-name}

kubectl describe pod mongo-depl-{pod-name}

### delete deployment

kubectl delete deployment mongo-depl

kubectl delete deployment nginx-depl

### create or edit config file

vim nginx-deployment.yaml

kubectl apply -f nginx-deployment.yaml

kubectl get pod

kubectl get deployment

### delete with config

kubectl delete -f nginx-deployment.yaml

### Metrics

kubectl top

# The kubectl top command returns current CPU and memory usage for a cluster’s pods or nodes, or for a particular pod or node if specified.