kubernetes-stroke-prediction

This repository will walk you throught the steps of deploying a containerized application in the Azure Cloud with Azure Kuebernetes Service, testing the endpoint, load testing the deployment, and automating the deployment and load testing.

Part 0: Test the Application Locally

Step 1 - Create a virtual environment and install the requirements

python3 -m venv ~/.venv
source ~/.venv/bin/activate
pip install -r requirements.txt

Step 2 - Run the application locally

python app.py

Step 3 - From a separate terminal, query the local host endpoint

python utilscli.py payload-predict

default host for utilscli.py is http://localhost:8080/predict - use python utilscli.py --help for more info

Part 1: AKS/ACR Setup and Application Deployment

Step 0 - Create a new resource group from the Azure dashboard

Ours is called ids721final

Step 1 - Create your aks cluster w/ desires specifications

az aks create --resource-group ids721final --name ids721final \
--generate-ssh-keys \
--node-count 3 \
--vm-set-type VirtualMachineScaleSets \
--load-balancer-sku standard \
--enable-cluster-autoscaler \
--min-count 1 \
--max-count 5

Step 2 - Merge aks credentials between kubectl and your aks cluster

az aks get-credentials --resource-group ids721Final --name ids721final

Step 3 - Create an acr respository

az acr create --resource-group ids721final --name ids721final --sku Basic --admin-enabled true

Step 4 - Attach acr repository to your aks cluster

az aks update --resource-group ids721final --name ids721final --attach-acr ids721final

Step 5 - Build your container image in you acr repository

az acr build --registry ids721final --image stroke-predict .

Step 5 - Deploy your application on your aks cluster

kubectl apply -f k8s/deployment.yaml

Step 6 - Add a load balancer in front of your cluster

kubectl apply -f k8s/loadbalancer.yaml

Step 7 - Find your load balancer IP

kubectl get services

Part 2: Query your Kubernetes Application

Step 1 - (if you haven't already) Create a virtual environment and install the cli requirements

python3 -m venv ~/.venv
source ~/.venv/bin/activate
pip install -r requirements.txt

Step 2 - Use the utility cli to query the load balancer endpoint:

python  utilscli.py payload-predict -host http://<Load-Balacer-External-IP>/predict

Part 3: Continuously Deliver you Kubernetes Application

Azure makes this process very simple :)

Navigate to the Azure Kubernetes Service console and select your deployment cluster.
In the panel on the left select "Deployment Center"
Follow the prompts to link the kubernetes deployment to your application's source code repository and set up CD

Part 4 : Clean-up Resources

The simplest way is to delete the resource group which contains the project

az group delete --name ids721final

Tips & Troubleshooting

Be careful with naming resources, Azure has strict requirements. For example, DNS-1035 label must consist of lower case alphanumeric characters or '-', start with an alphabetic character, and end with an alphanumeric character (e.g. 'my-name', or 'abc-123', regex used for validation is 'a-z?')

This means that even though you can name a resource '721final', it would be an invalid name. However 'final721' would be acceptable