Kubernetes Autoscaling

介紹

1. 用 Metrics Server 取出指標，比如可以根據 CPU、Memory 的使用率調整。
2. 用 HPA CRD 告知 Kubernetes 要根據什麼指標調整 Deployment.replicas 的數量。
3. 用 CA CRD 告知 Kubernetes 當 Worker Node 資源不足時（Worker Node 沒有足夠的 CPU, Memory 可以運行 Pod），新增 Worker Node。

Metrics Server

helm repo add metrics-server https://kubernetes-sigs.github.io/metrics-server/
helm upgrade --install metrics-server metrics-server/metrics-server

❯ kubectl top node
NAME                             CPU(cores)   CPU%   MEMORY(bytes)   MEMORY%
gke-default-pool-4d64e7e1-o5h7   158m         16%    2029Mi          77%
❯ kubectl top pod
NAME                                                             CPU(cores)   MEMORY(bytes)
haproxy-ic-kubernetes-ingress-7ddcc5db-kkjdl                     5m           109Mi
haproxy-ic-kubernetes-ingress-default-backend-56844d87f8-gjfbz   0m           2Mi
haproxy-ic-kubernetes-ingress-default-backend-56844d87f8-r2vmm   0m           2Mi

Pod Resource Request

apiVersion: v1
kind: Pod
metadata:
  name: frontend
spec:
  containers:
  - name: app
    image: images.my-company.example/app:v4
    resources:
      requests:
        memory: "64Mi"
        cpu: "250m"
      limits:
        memory: "128Mi"
        cpu: "500m"

HorizontalPodAutoscaler（HPA）

1. scaleTargetRef - 對哪個 Deployment 調整 replicas 參數。
2. minReplicas, maxReplicas - Pod 的數量範圍。
3. metrics - 根據什麼指標做調整，這個範例是當全部 Pod 累計的 CPU 使用率超過 80% 時，增加 Pod 的數量。

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: my-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-deployment
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
    name: cpu
    target:
      type: Utilization
      averageUtilization: 80

apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
  name: my-app-vpa
spec:
  targetRef:
  apiVersion: "apps/v1"
  kind: Deployment
    name: my-app
  updatePolicy:
    updateMode: "Auto"

Cluster Autoscaling（CA）

module "gke" {
  source = "terraform-google-modules/kubernetes-engine/google"
  version = "~>17.1.0"
  ...
  cluster_autoscaling = {
    enabled = true
    min_cpu_cores = 1
    max_cpu_cores = 20
    min_memory_gb = 3
    max_memory_gb = 20
    gpu_resources = []
  }
  node_pools = [
  {
    ...
    autoscaling = true
    min_count = 1
    max_count = 10
  },]
}