- Go to EC2 console: https://us-east-1.console.aws.amazon.com/ec2/home?region=us-east-1
- Create EC2 instance
- Pick deep learning ami
- Pick instance type: At least p3.2xlarge
- Create key-pair
- Download key
- Edit network
- Enable IPV4 address
- Open ports 8000-8004 from anywhere
- Launch Instance
- Get the ip address of the instance
- Change key permissions to 400 (
chmod 400 key.pem) - SSH into the machine
ssh -i key.pem [email protected] - Install git if needed (
sudo apt install gitfor ubuntu based distros,sudo yum install gitfor amazon linux) - Install Docker (
sudo apt install dockerfor ubuntu based distros,sudo yum install dockerfor amazon linux) - Start Docker (
sudo systemctl start docker) - Add user to docker group (
sudo usermod -aG docker ${USER}) - Logout and Login again through SSH to take the group changes into account
- Check if docker installed correctly (
docker run hello-world) - Install minikube
curl -LO https://storage.googleapis.com/minikube/releases/latest/minikube-linux-amd64
sudo install minikube-linux-amd64 /usr/local/bin/minikube
- Start a minikube cluster (
minikube start) - Set an alias for kubectl (
alias kubectl="minikube kubectl --") - For triton we need to add the credentials as secrets
kubectl create secret generic aws-env --from-literal='AWS_ACCESS_KEY_ID=YOUR_ACCESS_KEY' --from-literal='AWS_SECRET_ACCESS_KEY=YOUR_SECRET_ACCESS_KEY' --from-literal='AWS_DEFAULT_REGION=us-east-1'
- Use minikube's docker (
eval $(minikube docker-env)) - Pull the triton image (
docker pull nvcr.io/nvidia/tritonserver:22.06-py3) - Build all the docker images
docker build -t main .
docker build -t face-emotion face-emotion/
docker build -t pet-bokeh pet-bokeh/
- Load all the kubernetes resources (
kubectl apply -f K8s/) - Forward the main port (
kubectl port-forward svc/main 8004:8004 --address 0.0.0.0)
- Configure credentials (
minikube addons configure registry-creds) - Enable the addon (
minikube addons enable registry-creds) - Edit K8s/ECR/*.yaml to use your ECRs
- Load all the kubernetes resources (
kubectl apply -f K8s/) - Forward the main port (
kubectl port-forward svc/main 8004:8004 --address 0.0.0.0)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent