In this project, we developed a full-stack web app using the microservices architecture, and deployed it to a Kubernetes cluster. We also setup a monitoring system for the cluster with Prometheus and Grafana.
You can find the full demo for this project here.
The web app used for deployment is a simple blogging website, with the back-end divided into 3 microservices. The architecture of the app is as following:
- Front-end: ReactJS, handle UI and user interactivity.
- API Gateway: NodeJS, redirects requests/responses from front-end to back-end and vice versa.
- Back-end: ExpressJS, consists of 3 microservices:
- Auth service: Responsible for user authentication
- Post service: Handles creation, deletion and information of blog posts
- Comment service: Manages comments on each post
- Database: MongoDB, one for each microservice
The K8S deployment for the app consists of the following components:
- 3 PVC and 3 PV: Persistent volumes for each database.
- 8 Pod: Run front-end, API gateway, 3 microservices and their respective databases.
- 8 Service: One for each pod, only the service for front-end is a LoadBalancer, the rest are ClusterIP.
- 1 ConfigMap & 1 Secret: Contains configuration URL for connecting components and secrets.
We have used Minikube and Amazon EKS for the deployment.
To run the app on a Kubernetes cluster, we run the kubectl apply command on each resource YAML file resides here.
kubectl apply -f file-name.yaml
The order of deployment is as follows:
- ConfigMap & Secret
- Data volume
- Databases
- Microservices & API gateway
- Front-end
If you use Minikube, you must expose the front-end's LoadBalancer service to access it.
minikube service frontend-svc
If you use EKS, simply access the LoadBalancer's external IP provided by Amazon.
To setup Prometheus and Grafana, simply use the commands provided in monitor-cmd.txt.
Note: The installation commands require Helm, make sure you install it first.