Kubernetes: Pragmatic Overview

Introduction to Kubernetes

Kubernetes or k8s is an open source platform that is used to configure and automate container operations. It was developed by Google in 2014. It is used for scaling and deploying containerized applications. Automating container operations usually include orchestration of networking, computing and storage infrastructure on behalf of user workloads. A group of hosts running containers are clustered together and Kubernetes manages those clusters.

Why to use Kubernetes?

Real production applications span multiple containers which need to be deployed across multiple server hosts. Kubernetes manages to scale and deploy these containers for the workloads. Kubernetes orchestration allows us to build application services that span multiple containers, schedule and scale those containers across a cluster and manage to spawn up instances in case any one of the instance from a cluster fails. Kubernetes can be thought of as:

• a container platform
• a microservice platform
• a portable cloud platform

Containers can be grouped to form a "pod" and then schedule workload while providing services like storage and networking to the containers inside. Kubernetes also allows load balancing across pods.

Kubernetes Architecture

• Kubernetes follows a client-server arcitecture. There is 1 master which controls1 or more nodes. It is possible to have multiple master nodes for high availability.
• There is a Docker Registry which is a kind of central repository which containsDocker images.
• Master node contains 5 main components:
  Kube API Server: It is a central management entity that receives REST API requests and serves as a front-end for the cluster. It is the only component that communicates with the etcd cluster and makes sure that cluster state and data is stored in it. Replication Controller: It is a type of kube-controller-manager that is responsible for pod-replication to maintain shared state of cluster and to check if all tasks are performed. When a change in service configuration occurs or if a node fails, it sees to it that a new node is spawned up and the cluster state is changed to the desired state. Kube Scheduler: It is responsible for scheduling a pods on various nodes based on resource utilization. It considers the request requirements and assigns it the best-fit node. For ex, if an application needs 2GB ram, then it will look for nodes that satisfy this requirement with available resources. The scheduler must know the total available as well as the number of utilized resources. Kubeadm: It is responsible for pod administration as well as for initializing clusters. Etcd: etcd is a cluster which stores other clusters data and API objects. It isaccessible only from API Server for security reasons.
• Node contains 3 main components as follows:
  Kubelet: It is an agent that runs on each node in a cluster. It makes sure that the containers are running inside pods. Kube-proxy: It enables service abstraction by maintaining network rules on the host and performing connection forwarding. It also exposes services to the external world. Pods: It encapsulates containers, resources, configurations of how to run containers. Each pod has a unique ID associated to it. A continer contains simply a docker image.
• Whenever a user hits a requsest, it first goes to a Load Balancer. Here we have used Nginx as our Load Balancer. It is the duty of the load balancer to look after the traffic distribution of the requests. For more information, refer this.
• Nginx is a kind of a load balancer which sits in front of your server and distributes client requests across a group of services. The main advantage of using Nginx is that it maximizes speed and capacity utilization. For more information on Nginx, please refer this.

Conclusion

On a concluding note, we can find answers to 2 crucial questions based on k8s architecture as follows:

Why do we build a microservice using Kubernetes?

• We can run pre-built and configured components in containers as a part of every release. With every deployment, a new container is deployed which saves the overhead of doing manual configurations.
• We provide all the kubernetes cluster configurations through .yaml files, whichis usually called as the desired state of the cluster.
• Kubernetes maintains the desired and actual states of cluster and sees to it if they are always in synchronization. Thus, if any instance fails, the replication controller(which is a kind of kube-controller-manager) replicates the pod of thefailed instance, thus running the application successfully. It makes k8s maintain reliability of the system.
• Kubernetes offers autoscaling of workload based on CPU utilization and memory consumption.
• It allows vertical scaling-increasing the no. Of CPU's for pods and horizontal scaling-increasing the no. of pods.
• Thus, microservices can run better with kubernetes.

Why is it best practice to keep Database out of kubernetes cluster?

• In care of Kubernetes volume, when a pod is deleted, all the data is lost too.
• Thus, we use Persistent Volumes, which will keep the data even if a pod is spawned.
• It stores the data on our local storage.
• There is one Persistent Volume for one mysql. So, if the data inside the database will increase, the size of the local storage will also be needed to be increased.
• Thus, it is a best practice to keep database outside the kubernetes cluster.