Kubernetes (K8s)

Need for K8s

• High Availability (HA):
  • When we run our applications in docker container and if the container fails, we need to manually start the container
  • If the node i.e. the machine fails all the containers running on the machine should be re-created on other machine
  • K8s can do both of the above
• Autoscaling
  • Containers don’t scale on their own.
  • Scaling is of two types
    • Vertical Scaling
    • Horizontal Scaling
  • K8s can do both horizontal and vertical scaling of containers
• Zero-Down time Deployments
  • K8s can handle deployments with near zero-down time deployments
  • K8s can handle rollout (new version) and roll back (undo new version => previous version)
• K8s is described as Production grade Container management

History

• Google had a history of running everything on containers.
• To manage these containers, Google has developed container management tools (inhouse)
  • Borg
  • Omega
• With Docker publicizing containers, With the experience in running and managing containers, Google has started a project Kubernetes (developed in Go) and then handed it over to Cloud Native Container Foundation (CNCF)

Competetiors

• Apache Mesos
• Hashicorp Nomad
• Docker Swarm
• But K8s is clear winner

Terms

• Distributed System
• Node
• Cluster
• State
• Stateful Applications
• Stateless Applications
• Monolith
• Microservices
• Desired State
• Declarative vs Imperative
• Pet Vs Cattle
  

K8s is not designed only for Docker

• Initially k8s used docker as a main container platform and docker used to get special treatment, from k8s 1.24 special treatment is stopped.
• k8s is designed to run any container technology, for this k8s expects container technology to follow k8s interfaces.

K8s Architecture

• Official Architecture image
  
• Other easier representations
• Master Node
  
• Node
  
• Clients
  • kubectl
  • any rest based client
• Logical view
  
• Actual view
  

Kubernetes Components

• Refer Here for the k8s components article
• Control plane components (Master Node Components)
  • kube-api server
  • etcd (*)
  • kube-scheduler
  • controller manager
  • cloud controller manager
• Node Components
  • kubelet
  • kube-proxy
  • Container run time (*)

kube-api server

• Handles all the communication of k8s cluster
• Let it be internal or external
• kube-api server exposes functionality over HTTP(s) protocol and provides REST API

etcd

• Refer Here for etcd
• This is memory of k8s cluster

scheduler

• Scheduler is responsible for creating k8s objects and scheduling them on right node

Controller

• Controller Manger is responsible for maintaining desired state
• This reconcilation loop that checks for desired state and if it mis matches doing the necessary steps is done by controller

Kubelet

• This is an agent of the control plane

Container Runtime

• Container technology to be used in k8s cluster
• in our case it is docker.

Kube-Proxy

• This component is responsible for networking for containers on the node

kubectl

• This is command line that can be installed on the machine from which you communicate to k8s cluster.
• This tool is created to make communication with api-server simplified.
• Kubectl has a config file (KUBECONFIG) which contains
  • api-server information
  • keys to communicate with api server
• Kubectl allows to communication with cluster to create resources
  • imperatively: Type commands
  • declartively: Write manifests (YAML files)
• Reads manifests and connects to api server. Converts the manifest into REST API calls over JSON

What is k8s manifest

• This is a yaml file which describes the desired state of what you want in/using k8s cluster

CI/CD Workflow

• Basic Workflow
  
• Jenkins
  
• Azure DevOps
  

IDEAL K8s HA-Cluster

Kubernetes as a Service

• All popular clouds are offering k8s as a service
  • AKS (Azure K8s service)
  • EKS (Elastic K8s Service)
  • GKE (Google K8s engine)
• All cloud providers manage control plane for you and they charge hourly. For nodes we pay the similar costs of virtual machines

K8s Installations

• Single Node Installations
  • minikube
  • kind
• On-prem installations
  • kube-admin
• k8s as a Service
  • AKS
  • EKS
  • GKE
• Playground (for learning): Refer Here

Installing k8s cluster on ubuntu vms

• Create 3 ubuntu vms which are accesible to each other with atlest 2 vCPUS and 4 GB RAM
• Installation method (kubeadm) which is something we will be using in on-premises k8s.
• Refer Here for kubeadm installation on single master node

Steps

• Install docker on all nodes
• Install CRI-Dockerd Refer Here
• Run the below commands as root user in all the nodes

# Run these commands as root
###Install GO###
wget https://storage.googleapis.com/golang/getgo/installer_linux
chmod +x ./installer_linux
./installer_linux
source ~/.bash_profile

git clone https://github.com/Mirantis/cri-dockerd.git
cd cri-dockerd
mkdir bin
go build -o bin/cri-dockerd
mkdir -p /usr/local/bin
install -o root -g root -m 0755 bin/cri-dockerd /usr/local/bin/cri-dockerd
cp -a packaging/systemd/* /etc/systemd/system
sed -i -e 's,/usr/bin/cri-dockerd,/usr/local/bin/cri-dockerd,' /etc/systemd/system/cri-docker.service
systemctl daemon-reload
systemctl enable cri-docker.service
systemctl enable --now cri-docker.socket

• Installing kubadm, kubectl, kubelet Refer Here
• Now create a cluster from a master node Refer Here
• use the command kubeadm init --pod-network-cidr "10.244.0.0/16" --cri-socket "unix:///var/run/cri-dockerd.sock"
• Setup kubeconfig
• install flannel kubectl apply -f https://github.com/flannel-io/flannel/releases/latest/download/kube-flannel.yml
• AS a root user run kubeadm join commands (need to pass crisocket)
• Now from manager execute kubectl get nodes