hpa-prescaler-controller

Kubernetes Controller for prescaling HPA replicas of ArgoCD Apps

Kubernetes Deployment

Create ArgoCD Account and API Token

Login to ArgoCD Server with CLI

# define without https:// or http://
export ARGOCD_ENDPOINT="a4c345906685b46...........eu-west-1.elb.amazonaws.com"

argocd login $ARGOCD_ENDPOINT

Create hpaprescaler local account

kubectl patch configmap argocd-cm -n argocd --type merge -p '{"data":{"accounts.hpaprescaler":"apiKey","accounts.hpaprescaler.enabled":"true"}}'

Create hpaprescalerrole role and permissions

kubectl edit -n argocd cm argocd-rbac-cm

Add the following lines to policy.csv:

policy.csv: |
  p, role:hpaprescalerrole, applications, *, */*, allow
  p, role:hpaprescalerrole, projects, *, *, allow
  g, hpaprescaler, role:hpaprescalerrole

Restart argocd-server deployment

kubectl -n argocd rollout restart deployment/argocd-server

Generate hpaprescaler Api Token

argocd account list

argocd account generate-token --account hpaprescaler

Create CRDs

kubectl apply -f https://raw.githubusercontent.com/hepapi/hpa-prescaler-controller/main/deploy/crds/HpaPrescaler.yaml
kubectl apply -f https://raw.githubusercontent.com/hepapi/hpa-prescaler-controller/main/deploy/crds/HpaPrescalerProfile.yaml

Monitoring & Prometheus Alarms Setup

• Do this step if you want to enable Prometheus Rules in the values.yaml to generate Alarms .
• We'll use kube-state-metrics to export the CRDs object info as Prometheus Metrics.
• We need to customize the kube-state-metrics metrics installation with new helm values.

Assuming you have kube-prometheus-stack deployed on the monitoring namespace:

• Get the current values of your Helm release:

  helm -n monitoring get values kube-prometheus-stack -o yaml > current-stack-values.yaml

• Find out the current chart version used and make a note of it:

  helm -n monitoring list -o yaml

• Add/Merge the new kube-state-metrics values to current-stack-values.yaml file:

  kube-state-metrics:
    prometheus:
      # deploy a ServiceMonitor so the metrics are collected by Prometheus
      monitor:
        enabled: true
  
    rbac:
      create: true
      extraRules:
      - apiGroups: ["hepapi.com"]
        resources: ["hpaprescalers", "hpaprescalerprofiles", "hpaprescalers/status", "hpaprescalerprofiles/status"]
        verbs: ["list", "watch"]
  
    customResourceState:
      enabled: true
      config: 
        kind: CustomResourceStateMetrics
        spec:
          resources:
          - groupVersionKind:
              group: hepapi.com
              version: v1
              kind: HpaPrescaler
            labelsFromPath:
              name: ["metadata", "name"]
              namespace: ["metadata", "namespace"]
            metrics:
            - name: hpaprescaler_info
              help: "Exposes details about HpaPrescaler objects"
              each:
                type: Info
                info:
                  labelsFromPath:
                    argoAppName: ["spec", "argocdAppName"]
                    targetProfileName: ["spec", "targetProfileName"]
                    targetTime:  ["spec", "timeStart"]
                    state: ["status", "state"]
                    message: ["status", "message"]
                    processedAt: ["status", "processedAt"]

• Helm upgrade with new values yaml, using the same version:

  helm upgrade --install \
    kube-prometheus-stack prometheus-community/kube-prometheus-stack \
    --version "51.2.0" \
    -f current-stack-values.yaml \
    --namespace monitoring --atomic

Helm Deployment

update Values.yaml

vim deploy/helm-chart/values.yaml

Helm Chart Installation

helm upgrade --install hpaprescaler deploy/helm-chart

Uninstall

[optional] Remove objects

kubectl get hpaprescaler -o json | jq '.items[] | del(.metadata.finalizers)' | kubectl replace -f -
kubectl delete hpaprescaler --all

kubectl get hpaprescalerprofiles -o json | jq '.items[] | del(.metadata.finalizers)' | kubectl replace -f -
kubectl delete hpaprescalerprofiles --all

Helm Uninstall

helm uninstall hpaprescaler deploy/helm-chart

kubectl delete crd hpaprescalers.hepapi.com
kubectl delete crd hpaprescalerprofiles.hepapi.com