/hpa

Horizontal Pod autoscaler

Install Metrics server -

Clone this repository and install metrics server. Please do note that this setup is good for dev/qa environment. A lot of considerations must be put while installing metrics server in production environment. The official metrics-server repository is kept at https://github.com/kubernetes-incubator/metrics-server and we are using the same repo with few changes.

The file metrics-server-deployment.yaml is edited with the below statements -

        command:
        - /metrics-server
        - --kubelet-insecure-tls
        - --kubelet-preferred-address-types=InternalIP

The above command has the flag --kubelet-insecure-tls set which ignores strict check of kubelet certificates. In production you will definitely have a single CA private key that will be used to sign all kubelets. Please check the official metrics server documentation on other flags.

Install the metrics server

cd metrics-server

kubectl create -f .

Create nginx deployment

It is mandatory to set requests on cpu utilization as HPA requires CPU metrics.

kubectl create -f nginx.yaml

Create HPA resource

kubectl autoscale deploy nginx --min=3 --max=5 --cpu-percent=40

This might take a minute or two to show up -

kubectl get hpa 
NAME    REFERENCE          TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
nginx   Deployment/nginx   0%/40%    3         5         3          55s

Test the HPA using apache bench

apt-get install apache2-utils

kubectl expose deploy nginx --port=80 --type=ClusterIP

Get the service IP address using kubectl get svc

ab -n 500000 -c 1000 http://10.97.161.152/