sachinmour / collaboration

Architecture Details

1. Using an AWS managed EKS cluster for deploying the kubernetes configuration. (kube folder)
2. It's a single node cluster using t3.a.small instance type with 2cpus and 2gb memory.
3. Running two pod replicas for redundancy.
4. Running an ELB service in front of the pods with a public subnet.
5. Backend application is completely stateless.
6. Using AWS managed Redis cluster for all database and stream needs.
7. Everything above is in a single VPC where everything is in a private subnet except ELB.
8. Using github action for CI/CD needs

Browser

Any browser with websocket is supported. If websockets are not working in your browser for any reason, the app will not work.

Need to know

To make it easier for testing, everytime you refresh your browser, you will be randomly assigned a new user. So if you switch between tabs or even refresh your current tab, you will be a new user

How to develop locally

1. Make sure you have following things already in the system
   • node version lts/fermium
   • redis 6.x
2. Run yarn install in backend and frontend folders
3. Start redis with default parameters redis-server
4. Run yarn start:dev in backend folder.
5. Create an .env file in frontend folder with value REACT_APP_BACKEND_ENDPOINT=localhost:5000 change value to whatever host and port backend is started on.
6. Run yarn start in frontend folder.
7. You should see your application at localhost:3000

Note: If you do not start redis with default parameters, you can create a .env file in backend folder with keys REDIS_HOST and REDIS_PORT

How to test the system with a local kubernetes setup

You definitely don't need the below if you just want to develop. It's only for those rare cases when you wanna see the whole architecture as it is.

1. Make sure you have the following things already in the system along with everything in How to develop locally _ minikube _ kubectl _ docker _ Your local machine IP Address

• If you do not know your local IP Address, you can check it by starting frontend and you'll see it in the terminal along with localhost

2. Start minikube minikube start
3. Start redis with command redis-server --protected-mode no
4. Building your local docker image
   • Make sure you are in root directory of the repo
   • Run eval $(minikube docker-env)
   • Run docker build -t clover .
5. Start kubernetes cluster along with pods and services
   • Run kubectl apply -f kube/clover.development.yaml
6. You can check the status of your pods and service by
   • kubectl get pods
   • kubectl get svc clover
7. Now, to access our cluster, we have to tunnel our network
   • minikube service clover --url
8. Access the link and hurray.
9. To stop and clean up
   • kubectl delete -f kube/clover.development.yaml
   • minikube stop

Considerations I would take for full production load

Because currently it's running on a pretty small instance which already contains 6 pods (4 system pods) + (2 application pods) it's not exactly a big node to have great performance. I have balanced the parameters to receive and push updates such that all updates are propogated within ~1sec. We can increase our performance by atleast allocating 1 cpu/pod and having a bigger redis cluster.