Horizontal Pod Scaler

Introduction

If you have read through my blog from the top to the bottom, you will understand the concepts of MicroServices. To annex the learning goal of Scalability I have decided to use a technique called Horizontal Scaling. Before understanding Horizontal Scaling, Vertical Scaling must be first explained as it can paint a clear picture. Vertical Scaling refers to increasing the system resources based on demand. This setup has a few clear disadvantages; no single system can be finitely scaled vertically. Horizontal scaling gets around this problem by introducing multiple pods, meaning the same microservice hosted multiple times.

Theory

Execution

I created my first Microservice for the User component. This microservice uses mySQL; for simplicity reasons & the user microservice being a lesser used component in my application. I used my previous Spring Java REST API codebase to quickly set it up.

Implementation

Up until this point I implemented the deployment of my Ramses application in my Rancher cluster.

Rancher is an Kubernetes Management Interface; easily allowing me to deploy my pods in the way I configured them to. One of the services it provides is the implementation of a Loadbalancer; as can be seen here:

In it is my LoadBalancer configuration. Most of it is auto-generated by Rancher. I’ve made my own adjustments to the cpu resource, making Rancher scale my pod when the mCPU is met.

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  creationTimestamp: "2022-05-17T11:11:20Z"
  managedFields:
  - apiVersion: autoscaling/v2
    fieldsType: FieldsV1
    fieldsV1:
      f:spec:
        f:maxReplicas: {}
        f:metrics: {}
        f:minReplicas: {}
        f:scaleTargetRef:
          f:apiVersion: {}
          f:kind: {}
          f:name: {}
    manager: rancher
    operation: Update
    time: "2022-05-17T11:11:20Z"
  - apiVersion: autoscaling/v2
    fieldsType: FieldsV1
    fieldsV1:
      f:status:
        f:conditions:
          .: {}
          k:{"type":"AbleToScale"}:
            .: {}
            f:lastTransitionTime: {}
            f:message: {}
            f:reason: {}
            f:status: {}
            f:type: {}
          k:{"type":"ScalingActive"}:
            .: {}
            f:lastTransitionTime: {}
            f:message: {}
            f:reason: {}
            f:status: {}
            f:type: {}
          k:{"type":"ScalingLimited"}:
            .: {}
            f:lastTransitionTime: {}
            f:message: {}
            f:reason: {}
            f:status: {}
            f:type: {}
        f:currentMetrics: {}
        f:currentReplicas: {}
        f:desiredReplicas: {}
        f:lastScaleTime: {}
    manager: k3s
    operation: Update
    subresource: status
    time: "2022-05-17T11:12:35Z"
  name: lb-gameservice
  namespace: ramses
  resourceVersion: "418564"
  uid: a10d1017-32d7-4921-95d5-51fc4ebf228a
spec:
  maxReplicas: 10
  metrics:
  - resource:
      name: cpu
      target:
        averageValue: 500m
        type: AverageValue
    type: Resource
  minReplicas: 1
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: gameservice
status:
  conditions:
  - lastTransitionTime: "2022-05-23T06:17:13Z"
    message: recommended size matches current size
    reason: ReadyForNewScale
    status: "True"
    type: AbleToScale
  - lastTransitionTime: "2022-05-22T17:18:38Z"
    message: the HPA was able to successfully calculate a replica count from cpu resource
    reason: ValidMetricFound
    status: "True"
    type: ScalingActive
  - lastTransitionTime: "2022-05-17T11:12:50Z"
    message: the desired count is within the acceptable range
    reason: DesiredWithinRange
    status: "False"
    type: ScalingLimited
  currentMetrics:
  - resource:
      current:
        averageValue: 1m
      name: cpu
    type: Resource
  currentReplicas: 1
  desiredReplicas: 1
  lastScaleTime: "2022-05-22T14:26:40Z"

Monitoring

And lastly, I’ve implemented monitoring metrics to my Rancher, enabling with the help of Grafana to see the metrics granting insight to the pods.