Skip to main content

Start working with Kubernetes Today!

Kubernetes is quickly becoming the new standard for deploying and managing software in the cloud. With all the power Kubernetes provides, however, comes a steep learning curve. As a newcomer, trying to parse the official documentation can be overwhelming. There are many different pieces that make up the system, and it can be hard to tell which ones are relevant for your use case. This blog post will provide a simplified view of Kubernetes, but it will attempt to give a high-level overview of the most important components and how they fit together. We have used Kubernetes while hosting our Travel APIs like Flight API and Hotel API
Currently, there is a plenty of various courses/playgrounds, which can help you start working with the Kubernetes, like official Kubernetes tutorials or katacoda. I also went through them, but in this article, you will find not only theory but also examples, which help you implement your Kubernetes resources. We will deploy a complete application stack, consisting of a database and backend, frontend parts. In the end, you will find some exercises to do. I hope you will like it 🙂.
For the sake of simplicity, in this article, I will use the short name for Kubernetes: k8s.
#k8s #k8s cluster #k8s objects #deploy with kubectl #scaling & rollback #exercises


All the required tools, which have to be installed are listed in my Github repository, under the Prerequisites section. Prepare your environment, on the other hand, contains all commands, which have to be executed, before we start deploying our applications on Kubernetes.

Kubernetes and its components

#Kubernetes #Master #Nodes #kubelet #node processes

Kubernetes cluster

Master is the cluster orchestrator, which exposes the k8s API (docs). Every time, when we are deploying the app containers, we are telling something like: “Hey Master! Here is the docker image URL of my application. Please, start the app container for me.”. And then Master schedules the app instances (containers) to run on the Nodes. Kubernetes will choose where to deploy the app based on Nodes’ available resources.
Master manages the cluster. It scales and schedules app containers and rolls out the updates.
Nodes are k8s workers, which run app containers. A Node consists of the following processes:
  • kubelet — it’s an agent for managing the Node. It communicates with the Master using k8s API. It manages the containers and ensures that they are running and are healthy.
  • other tools — Node contains additional tools like Docker, to handle the container operations like pulling the image, running and so on.
Nodes are workers, which run application containers. They consist of the kubelet agent, which manages the Node and comunicates with the Master.

Kubernetes resources


Pod in Kubernetes cluster

Pod is the smallest resource in k8s. It represents a group of one or more application containers and some shared resources (volumes). It runs on a private, isolated network, so containers can talk to each other using localhost. Normally, you would have one container per Pod. But sometimes, we can run multiple containers in one Pod. Typically it happens, when we want to implement side-car (read more about this pattern in Designing Distributed Systems or Dave’s post).
To deploy a single Pod, you should run kubectl create pod-file.yaml or kubectl apply -f pod-file.yaml. Both commands are quite similar, but in comparison to create (which creates a resource), apply modifies an existing k8s object or creates new if it doesn’t already exist. To see how it works, let’s clone this repository and run the following command:
# deploy single Podkubectl apply -f example-pod.yaml# see running Podkubectl get pods
💡 Kubectl is a command-line client for k8s. It communicates with kube-apiserver (REST server), which then performs all operations.
💡 Most of the k8s commands consist of an action and a resource, on which this action is performed:
# <action> represents a verb like create, delete. 
# <resource> represents a k8s object.kubectl <action> <resource> <resource-name> <flags># e.g.
kubectl get pods my-pod
kubectl get pods
kubectl create deployment.yaml
kubectl apply -f deployment.yaml
👉 Other commands: cheatsheet and kubectl overview.
💡 You can use short names for the k8s resources:
pods (or pod): po, 
services (or service): svc, 
deployments (or deployment): deploy, 
ingresses (or ingress): ing, 
namespaces (or namespace): ns, 
nodes (or node): no# example (means kubectl get pods (or kubectl get pod))kubectl get po
Pod is the smallest resource in k8s. It runs on a private, isolated network and hosts app container instance. One Pod can contain multiple containers.


Deployment in Kubernetes cluster

Deployment is a k8s abstraction, which is responsible for managing (creating, updating, deleting) Pods. To deploy your application, you can always use Pods as in the previous example, however, using Deployments is a recommended way, which brings a lot of advantages:
  • you don’t have to worry about managing Pods. If one of the Pods terminates, the Deployment controller will create another Pod immediately. Deployments always take care of having a proper number of running Pods,
  • you have only one file, where you “define” Pod specification and desired number of running Pods. Pod specification is under spec.template key, whereas number or running Pods is under spec.replicas,
  • it provides a self-healing mechanism in case of machine failure. If the Node hosting an instance goes down or is deleted, the Deployment controller replaces the instance with an instance on another Node in the cluster,
  • it provides an easy way to rolling updates, etc. If you want to apply a change to your Pods, Deployment will update all Pods gradually, one by one.
Let’s create Deployments!
Before we start deploying backend and frontend, we have to make sure that the database is up and running. Our web server depends on the database’s health, so in case of issues with database connection, it will throw an error. Clone the Github repository and run the following commands to deploy the database:
# deploy databasekubectl apply -f database/deployment/database-deployment.yaml
kubectl apply -f database/deployment/database-service.yaml
Now, let’s build the image and deploy backend:
# build backend docker imagedocker build -t backend:v1 ./backend# deploy backendkubectl apply -f ./backend/deployment/backend-deployment.yaml
Repeat the above steps to deploy the frontend. The source code with the Dockerfile is under thefrontend/ directory:
# build frontend docker imagedocker build -t frontend:v1 ./frontend# deploy frontendkubectl apply -f ./frontend/deployment/frontend-deployment.yaml
As a result, you should see running Deployments and Pods inside the cluster:
kubectl get po,deploy

Screenshot from iTerm — kubectl get po, deploy

Deployment is responsible for creating and updating instances of your application. It monitors the application instance and provides a self-healing mechanism in case of machine failure.


Service in Kubernetes cluster

We already have Deployments, which started Pods with Docker containers inside. Even though they are deployed and ready to use, we can’t access them. Now it’s time to introduce a Service resource!
Service is another k8s object. Using it, we can make our Pods accessible from the inside, or outside k8s cluster. Service matches a set of Pods using selector defined in the Service under spec section and labels defined in the Pods’ metadata. So if the Pod has the label my-app, then the Service uses this label as a selector to know, which Pods should it expose.
Currently, Service has only 4 types:
  • ClusterIP — the default one. It allocates a cluster-internal IP address, and it makes our Pods reachable only from the inside the cluster.
  • NodePort — built on top of the ClusterIP (ClusterIP on steroids 💪). Service is now reachable not only from the inside of the cluster through the service’s internal cluster IP, but also from the outside: curl <node-ip>:<node-port>. Each Node opens the same port (node-port), and redirects the traffic received on that port to the specified Service.
  • LoadBalancer — build on top of the NodePort (NodePort on steroids 💪). Service is accessible outside the cluster: curl <service-EXTERNAL-IP>. Traffic is now coming via LoadBalancer, which then is redirected to the Nodes on a specific port (node-port).
  • ExternalName — this type is different than the previously mentioned. Here, you can have access to an external reference (web services/database/…) deployed somewhere else. Your Pods running in the k8s cluster can access them by using name specified in the Service YAML file. 

Let’s create Services!
👉 Links to the Github repository .
# deploy backend Servicekubectl apply -f ./backend/deployment/backend-service.yaml# deploy frontend Servicekubectl apply -f ./frontend/deployment/frontend-service.yaml
As a result, you should see running Services inside the cluster:
kubectl get svc

Screenshot from iTerm — kubectl get svc

Service defines the policy, by which we are able to access the Pods. Service matches a set of Pods using his selector and Pod’s labels. There are 4 types of Services: ClusterIP, NodePort, LoadBalander and ExternalName. Connections to the Service are load-balanced across all the backing pods.


Ingress in Kubernetes cluster

Ingress is a simple proxy, which routes traffic to the Services in the cluster. In one Ingress you can specify multiple Services to which it will redirect the traffic. Potentially, we don’t have to use Ingresses, but using it brings some advantages like having virtual hosts, SSL, CORS settings and so on.
Let’s create Ingresses!
👉 Links to the Github repository.
# deploy backend Ingresskubectl apply -f ./backend/deployment/backend-ingress.yaml# deploy frontend Ingresskubectl apply -f ./frontend/deployment/frontend-ingress.yaml
As a result, you should see running Ingresses inside the cluster:
kubectl get ing

Screenshot from iTerm — kubectl get ing

We have already deployed a complete application stack. Now, let’s check how does it look in the browser.
Ingress is a simple proxy, which routes the traffic to the Services in the cluster. It is easily extensible to take care of eg. CORS settings or SSL. It’s possible to have one Ingress for all Services in your cluster.

Scaling, rollback and quick updates


# scale the resource to specific number of replicaskubectl scale --replicas=REPLICAS_NUMBER -f your-yaml-file.yaml# example: scale backend-deployment.yamlkubectl scale --replicas=3 ./backend/deployment/backend-deployment.yaml# scale up to 3, when the current number of replicas of deployment is 2kubectl scale --current-replicas=2 --replicas=3 deployment/DEPLOYMENT_NAME# autoscale deployment between 2 - 10kubectl autoscale deployment DEPLOYMENT_NAME --min=2 --max=10
👉 More scaling commands: cheatsheet.


# rolling update "c" containers of "DEPLOYMENT_NAME" deployment, updating the IMAGE_NAME imagekubectl set image deployment/DEPLOYMENT_NAME c=IMAGE_NAME:v2# rollback to the previous deployment  
kubectl rollout undo deployment/DEPLOYMENT_NAME# watch rolling update status of deployment until completion      
kubectl rollout status -w deployment/DEPLOYMENT_NAME

Quick updates

# update a single-container pod's image version (tag) to v4kubectl get pod POD_NAME -o yaml | sed 's/\(image: IMAGE_NAME\):.*$/\1:v4/' | kubectl replace -f -# add a labelkubectl label pods POD_NAME new-label=awesome# add an annotationkubectl annotate pods POD_NAME icon-url=


Good luck! 🚀


Kubernetes is a very powerful platform. When you understand the basic concepts, you can already do a lot, but you will be still hungry to learn and play more with it. Until now, I will leave you with some resources, which I encourage you to read one by one.

Resources. Nice to read

  • Kubernetes docs — always very helpful.
  • Kubernetes API reference — I very often look into the API. Sometimes it’s easier to understand resources (their specification, what they have, what they need) from API than from the docs.
  • Kubernetes NodePort vs LoadBalancer vs Ingress?” blog post written by Sandeep Dinesh, with a good explanation about exposing your Services, with nice pictures.


  1. There you need to just register yourself for a free account, and on the next page, you can see the 30 days the Semrush trial will be activated. You do need to also remember that your card will be charged after the 30 days of free trial ends. Get help from

  2. This is a very nice one and gives in-depth information. I am really happy with the quality and presentation of the article. I’d really like to appreciate the efforts you get with writing this post. Thanks for sharing.
    3Ds max training in pune


Post a Comment

Popular posts from this blog

30 Amazing Machine Learning Projects for the Past Year (v.2018)

For the past year , we’ve compared nearly 8,800 open source Machine Learning projects to pick Top 30 (0.3% chance). This is an extremely competitive list and it carefully picks the best open source Machine Learning libraries, datasets and apps published between January and December 2017. Mybridge AI evaluates the quality by considering popularity, engagement and recency. To give you an idea about the quality, the average number of Github stars is 3,558. Do visit our Hotel price comparison api which compares more than 200 hotel websites to get you the best possible price of your dream hotel. Python Projects of the Year (avg. 3,707 ⭐️): Here (0 duplicate) Learn Machine Learning from Top Articles for the Past Year: Here (0 duplicate) Open source projects can be useful for data scientists. You can learn by reading the source code and build something on top of the existing projects. Give a plenty of time to play around with Machine Learning projects you may

Cheat Sheets for AI, Neural Networks, Machine Learning, Deep Learning & Big Data

Over the past few months, I have been collecting AI cheat sheets. From time to time I share them with friends and colleagues and recently I have been getting asked a lot, so I decided to organize and share the entire collection. To make things more interesting and give context, I added descriptions and/or excerpts for each major topic. This is the most complete list and the Big-O is at the very end, enjoy… If you like this list, you can let me know  here Neural Networks Neural Networks Cheat Sheet Neural Networks Graphs Neural Networks Graphs Cheat Sheet Neural Network Cheat Sheet Ultimate Guide to Leveraging NLP & Machine Learning for your Chatbot Code Snippets and Github Included Machine Learning Overview Machine Learning Cheat Sheet Machine Learning: Scikit-learn algorithm This machine learning cheat sheet will help you find the right estimator for the job which is the most difficult part. Th

Building a Game using JavaScript : Part-1

Introduction I really wanted to write a tutorial about a game technology I like to use, so here it is. In this story, we will start making a little shoot’em up game with PixiJS, a really simple and cool Javascript library. What we are going to do exactly is to make a spaceship able to move and shoot, enemy waves coming through and a beautiful animated background with moving clouds. The first part (this story) will focus on the background. Ready guys? Let’s nail it! Getting started Let’s start by setting up our project: I uploaded a code structure already set so we are all working with the same base. However if you want to make it yourself, I put a picture of my folder just below : Click here to download the starter project Project folder structure We will need a local server to run the game: I invite you to download WAMP if you’re working with Windows, or MAMP for macOS, they are free and easy to use. Let’s put your game folder in the server one ( htdo