AutoScale K8s REST example

A REST api example developed in Go, based on Echo framework. backed by Redis, running on K8s, featured with Ingress and HPA.

Get Started

git clone https://github.com/csimplestring/autoscale-k8s-rest-example.git
minikube start (you need to install minikube before)
make build-docker

Deploy on Minikube

cd k8s
kubectl create -f namespaces.yaml
kubectl create -f volumes.yaml
kubectl create -f ingress.yaml
kubectl create -f ingress-controller.yaml
kubectl create -f redis.yaml
kubectl create -f api.yaml

Because we are using the local Minikube cluster, it is not possible to expose the Ingress Controller as a 'LoadBalancer' Service(usually you should) on your cloud provider.

In order to access the api via Ingress, you have to run:

kubectl describe svc --namespace=kube-system traefik-ingress-service

you need to find the exposed port, for example. 31321, use this port as {api-port}

Test

// create
Post http://{minikube-ip}:{api-port}}/api/customer
Host: api.minikube
Content-Type: application/json
{
  "name": "1",
  "address": "b"
}

// get
GET http://{minikube-ip}:{api-port}}/api/customer/1
Host: api.minikube
Content-Type: application/json

Run CronJob

// this cron job will back up the redis data every 5 minutes

kubectl create -f jobs.yaml

Enable HorizontalPodAutoscaler

before this, you need to enable heapster and metric-server in Minikube.

	minikube addons enable metrics-server
	minikube addons enable heapster
	kubectl autoscale deployment api --cpu-percent=10 --min=2 --max=5 --namespace=testing

which means when the average CPU usage is more than 10%, the api pod will be automatically created, the maximum pod number will be 5. If the average CPU usage is less than 10% later, the api pods will be automatically deleted in order to save resources.