Home / Blog /

Deploy a Containerized Web Application on Google Kubernetes Engine

02 June 2020 • 7 min read

Deploying a containerized web application on a Kubernetes cluster may be daunting at first; at least that is how I felt initially. There are many managed Kubernetes service available today, such as Google Kubernetes Engine, AWS Elastic Kubernetes Service and Azure Kubernetes Service. You can even set up a Kubernetes cluster inside a Virtual Machine with your own laptop and Minikube .

For this article, I will be focusing on the steps to build a container image of the web application with Google Cloud Build, and deploy the containerized web application on a Google Kubernetes Engine (GKE) cluster.

Pre-Requisites

Web Application

I will be using my sample Node.js Express API as the web application. You may download it from my GitHub Repository , and push it to your own respository to try out the steps in this article. Alternatively, you may use your own application, but make sure that you have a Dockerfile included with your source code.

Google Cloud Account

Make sure you have a Google Cloud account with all the necessary priviledges granted and API enabled (e.g. Cloud Build, Container Registry, Kubernetes Engine).

Basic Understanding of Container Applications and Container Orchestration Tools

It is also beneficial to have a basic understanding of container applications and container orchestration tools, such as Kubernetes.

1. Automate the creation of container image with Google Cloud Build

In this section, we are going to link a GitHub repository with Google Cloud Build. A push trigger will also be created to build and store the image in Google Container Registry once there are changes pushed to the repository.

Navigate to Cloud Build > Trigger. Click "Connect Repository". Select "GitHub (Cloud Build GitHub App)" as the source and click "Continue". Cloud Build - Select GitHub as Source

Authorize Google Cloud Build to access your GitHub repository and click "Install Google Cloud Build". Cloud Build - Install Google Cloud Build

A new window should appear. Select the GitHub repository to install Google Cloud Build, and click install. Type in your GitHub account password. Cloud Build - Select GitHub repository to install Google Cloud Build

Cloud Build - Select GitHub repository to install Google Cloud Build

Tick all the necessary check boxes and click "Connect Repository". Cloud Build - Connect repository

Click "Create push trigger" and you will be redirected back to the Trigger list. You should be able to see the trigger which you have just created. Cloud Build - Create push trigger

In order to trigger the build only when there are changes pushed to the "master" branch, click on the name of the trigger and change value of "Branch" from ".*" to "^master$" on the Edit Trigger page. Click "Save". This is a good practice so that you can create a branch for development of features, and only merge to "master" once you are ready to build a newer version of the container image. Cloud Build - Edit Trigger

Click on "Run Trigger" on the Trigger page to trigger the build. In addition, whenever changes are pushed to the "master" branch of the repository, the build will be triggered as well. Navigate to Cloud Build > History to check whether the build is successful or not. If it is successful, you should see a similar record depicted in the following screenshot. Cloud Build - Build History

Navigate to Container Registry > Images. The container image should appear in the list. Click on the image name. Container Registry - Images

Remove the generated tag and add "1.0" as a tag. Click on the copy icon to copy the URL and keep it in the notepad. Append ":1.0" at the back of the URL. It will be used as the value of "image" in deployment.yaml later. Container Registry - Images

The URL should look similar to this:

gcr.io/[project-name]/sample-express-api:1.0

2. Create a GKE Cluster

In this section, we are going to create a GKE cluster, which will create the Pods to run the containerized web application. These Pods are running in a Node, which can be a physical or virtual worker machine (in GKE, Compute Engine is used).

Navigate to Kubernetes Engine > Clusters and click on "Create Cluster". Ensure that "Static Version" is selected and select the latest version in the dropdown list. Leave the other settings as their defaults. Kubernetes Engine - Cluster Basics

Navigate to the "default-pool" tab and change the number of nodes from 3 to 1. For production workloads, it should be 3 minimally, as recommended by Google. Leave the other settings as their defaults. Kubernetes Engine - Node Pool

Navigate to the "Nodes" tab. Select "N1" series and "g1-small" machine type and change boot disk size to 50gb (or lesser). This should minimize the cost for non-production workload. Leave the other settings as their defaults. Kubernetes Engine - Nodes

Click "Create" at the bottom of the page and wait for the cluster to be created.

3. Deploy the containerized web application to the GKE cluster

In Kubernetes, you need to create a deployment so that the cluster will "be informed" of the number of Nodes and Pods (replicas) to create, the container image to use, etc. These details will be stored in a YAML file (i.e. deployment.yaml).

You may download a sample deployment.yaml from my GitHub repository . Replace [URL] in deployment.yaml with the URL which was mentioned at the end of section 1. Click on the ellipsis icon located at the top right of the Cloud Shell window and click "Upload File". Select the deployment.yaml to upload. Cloud Shell - Upload File

Navigate to Kubernetes Engine > Clusters, click "Connect". Kubernetes Engine - Cluster List

Copy the command in the popup window and execute it in the Cloud Shell window to configure the kubectl command line access. The command should look similar to this:

gcloud container clusters get-credentials cluster-1 --zone asia-southeast1-a --project [project_id]

Next, execute the following command to create a deployment in Kubernetes:

kubectl create -f ./deployment.yaml

You should see a similar message like "deployment.apps/sample-express-api created".

Execute the following command to list the deployments:

kubectl get deployments

If the deployment is completed without errors, you should see a similar result as the following screenshot. Kubernetes Engine - Get Deployments in Cloud Shell

Kubernetes Engine - Get Deployments in Cloud Shell

Execute the following command expose the deployment via a Kubernetes Load Balancer Service. This will allow you to access the API via the internet.

kubectl expose deployment sample-express-api --type=LoadBalancer --name=lb-service

Navigate to Kubernetes Engine > Services & Ingress. There should be a record of the Kubernetes Load Balancer Service. Click on the endpoint. A new tab will be opened, which will redirect you to the IP address. You should see the values returned by the API. Kubernetes Engine - Services & Ingress

4. Updating the deployment in GKE cluster

When you update your source code in the repository which was linked to Cloud Build, Cloud Build will build a new container image automatically. In order to instruct the Kubernetes cluster to use the new image for all Pods, you need to tag the new container image with a tag, update the deployment.yaml and execute the "kubectl apply" command.

Make some changes to the values returned by the API and push these changes to the "master" branch. This should trigger a new build for the container image.

Navigate to Container Registry > Images. Click on the image name. You may tag the container images as follow: Container Registry - Image Tags

Assuming you tag your image as suggested above, you need to change the tag value of the image URL from "1.0" to "2.0" in deployment.yaml (remember to replace [project_id] with your project ID).

...
image: gcr.io/[project_id]/sample-express-api:2.0
...

You may edit the uploaded deployment.yaml with the Cloud Shell Editor. Cloud Shell - Edit deployment.yaml

Assuming that you have configured your Cloud Shell to access kubectl (if not, you may go back to section 3 and follow the steps to configure), execute the following command:

kubectl apply -f ./deployment.yaml

You should see the following message if the command is executed succesfully:

deployment.apps/sample-express-api configured

Verify that the value returned by the API is updated by accessing the Load Balancer Service endpoint.

TL;DR

Create a trigger in Cloud Build, which will build the container image from your Git repository when there are updates
Create a GKE cluster
Create a deployment in the GKE cluster
Update the deployment when a new version of the container image is built

GoogleContainersKubernetes