smolify - Load Balancer for URL Shortner

Demo

merged.mp4

Docker Containerization

building and the Dockerfile itself

docker build --target dev . -t flasker

redis clustering

• making use of redis sentinel for high availability

how to setup

• ensure you are using the same network across to do this

docker network create redis

• you will need redis:7.4.2-alpine so pull that image

docker pull redis:7.4.2-alpine

• for data persistence let us create volumes for redis db (/data)

docker volume create redis0-data
docker volume create redis1-data
docker volume create redis2-data

• now run the redis.sh in storage folder, this should get 3 containers up and running

cd storage/
bash redis.sh
cd ..

• to find the ip address of the sentinels run for all sentinels

docker inspect -f '{{range.NetworkSettings.Networks}}{{.IPAddress}}{{end}}' sentinel-0

• to run redis container interactively

docker exec -it redis-0 redis-cli -h 172.18.0.2 -p 6379

• running the python container

docker run -t -v ${PWD}:/work -p 5000:5000 --net redis -e REDIS_SENTINELS="172.18.0.5:5000, 172.18.0.6:5000, 172.18.0.7:5000" -e REDIS_MASTER="mymaster" -e REDIS_PASSWORD="admin" -e DB_PASS="vhOEgZiXgTfd9Z43" flasker

---

Container Orchestration - Kubernetes

• install kind (Kubernetes IN Docker)
  kind is a tool for running local Kubernetes clusters using Docker container “nodes”.

• Create a redis cluster

kind create cluster --name redis --image kindest/node:v1.23.5

• Apply the kind config so that flask-app ports get exposed properly

kind create cluster --config kind-config.yaml

• Create a namespace

kubectl create ns redis

• We will be using the default storage class which is standard, you can check this by

kubectl get storageclass

• Deployment of pods

  • Redis nodes (1 master 2 slaves)

    kubectl apply -n redis -f ./kubernetes/redis/redis-configmap.yaml
    kubectl apply -n redis -f ./kubernetes/redis/redis-statefulset.yaml

  • Redis Sentinel (3 instances)

    kubectl apply -n redis -f ./kubernetes/sentinel/sentinel-statefulset.yaml

  • Verify all pods are running

    kubectl -n redis get pods

  You should see 3 redis pods and 3 sentinels so a total of 6 pods should be visible

  • Flask app (3 replicas)

    kubectl apply -f ./kubernetes/flask/flask-app-configmap.yaml
    kubectl apply -f ./kubernetes/flask/flask-app-secret.yaml
    kubectl apply -f ./kubernetes/flask/flask-app.yaml

    kubectl get pods # note that this is not inside redis namespace

  You should see 3 flask-app-{random hash}

• Setting up metrics-server (the below command might take sometime so give it ~5mins)

kubectl apply -f .kubernetes/metrics-server/metrics-server-deployment.yaml

• To check if it got deployed correctly

kubectl get deployments -n kube-system

• Display resource (CPU/memory) usage of pods

kubectl top pods

kubectl describe hpa flask-app

• Port Forward local machine's port to the flask-app service

kubectl port-forward service/flask-app 5000:5000 # (make sure to access app from http://127.0.0.1:30007/)

• Manual Monitoring

  • To check if flask app is receiving requests

    kubectl logs <flask-pod-name> # run for all pods one of them will be receiving requests

  • Interactively checking the redis db

    kubectl -n redis exec -it redis-1 -- sh
    -> redis-cli
    -> auth admin
    -> keys *

• Logs of all flask-app pods

kubectl logs -l app=flask-app --prefix --follow --max-log-requests 10

Stress Testing

• Hitting the / end-point

bash stress-testing/stress_GET.sh

• Url-shortening

bash stress-testing/stress_POST.sh

Future Work

• Secondary db storage- MongoDB
• Setup Flunetd for better Log collection
• Setup Prometheus and Grafana for monitoring
• Setup CI/CD pipeline or Github Actions
• Improve Docs, add References