Requirements For HPA To Work For Pravega

[Prabhaker24]

Description

Requirements For HPA To work for Pravega

Importance

nice-to-have

Suggestions for an improvement

1. Read-up on HPA - how it works, what metrics can it use for scaling.
2. Deploy a sample application with HPA configured and get it to scale up/down based on resource and external metrics.
3. List down requirements for HPA to work.

[Prabhaker24]

As part of Experimenting, I have tried setting up HPA for a sample php-server with the resource as metrics (both CPU and Memory) and here are the steps involved for doing that:-

1. We need to install the Metrics Server below is how we can do that:-
   kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml

2. Because of an issue in K8 ( x509: certificate signed by unknown authority kubernetes-sigs/metrics-server#146) we need to edit the metric server before installing hpa for hpa to be able to fetch the metrics
   Note:- this is needed With Kubernetes 1.18 and Metrics v0.3.7 or above

args:
  - --kubelet-insecure-tls
  - --kubelet-preferred-address-types=InternalIP
  - --cert-dir=/tmp
  - --secure-port=4443

3. Install php-server deployment

4. Create HPA based on cpu utilization and memory utilization

5. Try creating load on the server and we should see the pod getting scalled up

For points from 3 to 5 following link can be referred (https://medium.com/faun/kubernetes-horizontal-pod-autoscaler-hpa-bb789b3070e4)