Learn all about Kubernetes with this quick reference.

Kubernetes Intro

What is Kubernetes?

• Kubernetes is an open-source container orchestration platform.
• Kubernetes abstracts the underlying infrastructure.
• Kubernetes runs on infrastructure and the application runs on Kubernetes.
• Automates the deployment and management of containerized applications at scale
• Kubernetes exposes required API’s for controlling containerized workloads.
• Kubernetes operates at container level not at hardware level
• Containers enables you to package your application as a microservice but to manage these containers(ex. when the container goes down) in a production environment, you need orchestration tools like Kubernetes.
• Kubernetes originated from Greek word, meaning “helmsman” or “pilot.” and Google open-sourced the project in 2014
• Kubernetes is sometimes referred as “k8s” where “8” represents the 8 characters between “K” and “s” in Kubernetes word.

Why to use Kubernetes and its Key offerings?

• It takes care of scaling and failures of your containerized application
• Self-healing – Kubernetes can restart or replace the faulty container and it will not expose the containers until it is ready
• Kubernetes exposes container IP address or DNS name
• Kubernetes provides load balancing, so that traffic is distributed evenly across all the containers
• Effective resource utilization on the nodes based on the available compute resources
• Kubernetes provides automated rollouts or creation of new containers for your deployment
• Kubernetes allows you to mount storage of your choice
• Kubenetes lets you manage secrets and configurations without rebuilding container images
• Loosely coupled and highly available infrastructure
• Extensible open-source community built plugins to add additional capabilities such as logging, monitoring, and security.
• Highly portable – you can move containerized apps across environments without worrying about the configuration.
• Kubernetes can able to span multi-cloud – i.e Master and Nodes can live on different cloud provider environments but this may not be a good option considering performance and reliability.

Good to Know

• Kubernetes releases uses Semantic Versioning Specification – Major.Minor.Patch
• The “minor” changes are (1.xx at present) delivered every three to four months.
• Patch versions are released to address the critical bug and security fixes to the latest minor version.

Kubernetes compared to other Container Orchestrators

• Highly complex ecosystem compared to other container orchestrators such as Nomad and Docker Swarm etc.
• Kubernetes is supported by majority of cloud providers which is not the case for others.
• Kubernetes has got extensive community support.

Kubernetes Architecture

Kubernetes Core Components

• You get cluster when you deploy Kubernetes – which aggregates all the server resources into a usable pool
• Kubernetes follows master-slave architecture
• Kubernetes cluster comprises Control Plane / Master and Worker / Slave node 
• Kubernetes cluster usually runs on multiple nodes / servers to provide high availability and fault tolerance.

1. Control Plane Components

• Collection of components which are used to make global decisions like scheduling and responding to the events.
• Manages all the worker node and the Pods
• Setup scripts usually start the control plane on the same server and do not run user containers/workloads on it.
• Control Plane Components – kube-apiserver, etcd, kube-scheduler, kube-controller-manager, cloud-controller-manager

kube-apiserver

kube-apiserver is a front end for your Kubernetes Control Plane

• bube-apiserver responsibility is to authenticate and keep the data within etcd upto date
• It exposes Kubernetes API’s for other components in the cluster

etcd

• etcd is a distributed and highly available key-value store.
• etcd stores configuration information about the Kubernetes cluster componentes such as Nodes, Pods, Configs, Secrets etc
• Accessible only by Kubernetes API server
• The result you get when you run “kubectl get” is from etcd
• You will need to have backup plan in place if you are using etcd as datastore

kube-scheduler

• kube-scheduler is the decision maker for all the newly created pods. It decides on which node the pod has to run.
• kube-scheduler sends the decision information to kubelet on the worker node, which then takes care of creating Pod.

Kube-controller-manager

• There are many controller managers available which task is to maintain the desired functioning state of the components within the cluster.
• Node Controller: Responsible for monitoring and responding to nodes failures.
• Replication Controller: Responsible for ensuring the desired number of Pods for every replica set.
• Endpoints Controller: Responsible for joining endpoints for services and pods.
• Service Account: Creates default accounts.
• Token Controllers: Creates API access tokens for new name spaces.

cloud-controller-manager

• cloud-controller -manager responsibility is to link the kubernetes cluster to cloud provider API.
• Cluster doesn’t have cloud-controller-manager if you are running Kubernetes on your local machine or on premise.
• Runs controllers that are specific to cloud providers.
• Node Controller: Responsible for monitoring and responding to nodes failures on cloud
• Route Controller: Responsible to set up routing logic inside cloud infrastructure.
• Service Controller: Responsible for creating, updating and deleting load balancers on the cloud.

2. Worker Node Components

• Every cluster has at least one worker node – it may be a virtual or physical machine.
• Worker node is managed by Control Plane
• Contains necessary services to run the Pods
• Host the containerized application workload in a component called Pods
• Worker Node Components – kubelet, kube-proxy, container runtime.
• Node components run on every node that you have on the cluster.
• Maintains the running pods and provides the runtime container environment.

kubelet

• kubelet is an agent which runs on all worker nodes.
• kubelet responsible for creating pods on the worker node and ensures containers are running on pods

kube-proxy

• Kube-proxy is a network proxy which runs on all worker nodes.
• Responsible for implementing service concept
• Maintains network rules on worker nodes, which allows network communication with pods.

Container runtime

• Container runtime responsible for running the containers
• Kubernetes supports container runtime such as Docker, Conainerd.