autoscale-on-ocp-4.md

Autoscale in OCP 4.x

Step 1: Set resource limit on Deployment or Deployment Config

This step is for existing service which not yet set resource limit. Edit deployment config yaml or deployment yaml. set this variable

• spec.containers[].resource.limits.cpu -> max limit
• spec.containers[].resource.request.cpu -> initial vcpu

    spec:
      containers:
        - name: training-rest-notif-jms
          image: >-
            image-registry.openshift-image-registry.svc:5000/erfin-demo/training-rest-notif-jms@sha256:30807906a9d7d6419c3d4219a6f8469728e1bfba0410009532277d798108bcbf
          ports:
            - containerPort: 8080
              protocol: TCP
            - containerPort: 8443
              protocol: TCP
            - containerPort: 8778
              protocol: TCP
          resources:
            limits:
              cpu: 200m
            requests:
              cpu: 100m

set request cpu 100 milicore max 200 milicore

Step 2: Set Autoscaler

Run below command if using dc (DeploymentConfig)

> oc autoscale dc/training-rest-notif-jms --min 2 --max 10 --cpu-percent=80

Or, run below command if using deployment -> Deployment

> oc autoscale deployment/training-rest-notif-jms --min 2 --max 10 --cpu-percent=80

this oc command will generate HPA (Horizontal Pod Autoscaller)

kind: HorizontalPodAutoscaler
apiVersion: autoscaling/v2beta1
metadata:
  name: training-rest-notif-jms
  namespace: erfin-demo

...


...
spec:
  scaleTargetRef:
    kind: DeploymentConfig
    name: training-rest-notif-jms
  minReplicas: 2
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: cpu
        targetAverageUtilization: 80

We will set minimum replica 2, max replica 10. autoscale when cpu reach 80% (targetAverageUtilization) or 160 milicore

Notes

beware of your application cpu consumption during startup. if you set request cpu to small, hence it will scale out your pod during apps startup process.