gcp-gke – skill DevOps & Hạ tầng cho agent AI

Google Kubernetes Engine (GKE)

Deploy, operate, and scale managed Kubernetes clusters on Google Cloud Platform.

When to Use

• Running containerized microservices at scale with automatic scaling and healing
• Workloads requiring fine-grained orchestration, service mesh, or custom scheduling
• Teams already invested in Kubernetes tooling (Helm, Argo CD, Flux)
• When Cloud Run's request-based model does not fit (long-running, stateful workloads)

Project, cost and identity boundaries

Verify gcloud account/project, cluster context, region and quota/budget before writes. Regional Standard node counts and autoscaling flags may be per zone, so compute total nodes/cost explicitly; Autopilot billing/limits depend on compute class and current pricing. Use a dedicated least-privilege node service account and bucket/resource-scoped workload permissions rather than broad project grants when possible. Private nodes do not alone make the control-plane endpoint private. Terraform below is an excerpt requiring provider, variables and network definitions; enabling a policy feature does not install the deny/allow rules or prove enforcement.

Prerequisites

• Google Cloud SDK (gcloud) and kubectl installed
• APIs enabled: Kubernetes Engine, Compute Engine
• IAM role roles/container.admin for cluster management

gcloud services enable container.googleapis.com compute.googleapis.com
gcloud components install kubectl

Standard vs Autopilot

| Feature | Standard | Autopilot |

|---------|----------|-----------|

| Node management | You manage node pools | Google manages nodes |

| Pricing | Pay per node (VM) | Pay per pod resource request |

| GPU/TPU | Full support | Supported (with limits) |

| DaemonSets | Allowed | Restricted |

| Best for | Full control, specialized HW | Hands-off, cost-optimized |

Create a Standard Cluster

gcloud container clusters create prod-cluster \
  --region=us-central1 --num-nodes=2 \
  --machine-type=e2-standard-4 --disk-size=100 \
  --enable-autoscaling --min-nodes=1 --max-nodes=5 \
  --enable-autorepair --enable-autoupgrade \
  --release-channel=regular \
  --workload-pool=${PROJECT_ID}.svc.id.goog \
  --enable-ip-alias --enable-network-policy \
  --enable-shielded-nodes \
  --logging=SYSTEM,WORKLOAD --monitoring=SYSTEM,WORKLOAD \
  --labels=env=production,team=platform

gcloud container clusters get-credentials prod-cluster --region=us-central1

Create an Autopilot Cluster

gcloud container clusters create-auto autopilot-prod \
  --region=us-central1 --release-channel=regular \
  --workload-pool=${PROJECT_ID}.svc.id.goog \
  --network=my-vpc --subnetwork=gke-subnet

Node Pools

# High-memory pool with taint
gcloud container node-pools create highmem-pool \
  --cluster=prod-cluster --region=us-central1 \
  --machine-type=n2-highmem-8 --disk-size=200 --disk-type=pd-ssd \
  --num-nodes=1 --enable-autoscaling --min-nodes=0 --max-nodes=4 \
  --node-labels=workload=memory-intensive \
  --node-taints=dedicated=highmem:NoSchedule

# GPU pool
gcloud container node-pools create gpu-pool \
  --cluster=prod-cluster --region=us-central1 \
  --machine-type=n1-standard-8 \
  --accelerator=type=nvidia-tesla-t4,count=1 \
  --num-nodes=0 --enable-autoscaling --min-nodes=0 --max-nodes=4 \
  --node-taints=nvidia.com/gpu=present:NoSchedule

# Spot pool for batch workloads
gcloud container node-pools create spot-pool \
  --cluster=prod-cluster --region=us-central1 \
  --machine-type=e2-standard-4 --spot \
  --num-nodes=0 --enable-autoscaling --min-nodes=0 --max-nodes=20 \
  --node-taints=cloud.google.com/gke-spot=true:NoSchedule

Workload Identity

# Create GSA and grant permissions
gcloud iam service-accounts create app-gsa
gcloud projects add-iam-policy-binding ${PROJECT_ID} \
  --member="serviceAccount:app-gsa@${PROJECT_ID}.iam.gserviceaccount.com" \
  --role="roles/storage.objectViewer"

# Create KSA and bind to GSA
kubectl create namespace myapp
kubectl create serviceaccount app-ksa --namespace=myapp
gcloud iam service-accounts add-iam-policy-binding \
  app-gsa@${PROJECT_ID}.iam.gserviceaccount.com \
  --role=roles/iam.workloadIdentityUser \
  --member="serviceAccount:${PROJECT_ID}.svc.id.goog[myapp/app-ksa]"
kubectl annotate serviceaccount app-ksa --namespace=myapp \
  iam.gke.io/gcp-service-account=app-gsa@${PROJECT_ID}.iam.gserviceaccount.com

Deploying Workloads

# deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: web-app
  namespace: myapp
spec:
  replicas: 3
  selector:
    matchLabels: { app: web-app }
  template:
    metadata:
      labels: { app: web-app }
    spec:
      serviceAccountName: app-ksa
      containers:
      - name: web
        image: us-central1-docker.pkg.dev/PROJECT_ID/repo/web-app:v1.2.0
        ports: [{ containerPort: 8080 }]
        resources:
          requests: { cpu: 250m, memory: 512Mi }
          limits: { cpu: 500m, memory: 1Gi }
        readinessProbe:
          httpGet: { path: /healthz, port: 8080 }
          initialDelaySeconds: 5
        livenessProbe:
          httpGet: { path: /healthz, port: 8080 }
          initialDelaySeconds: 15
      topologySpreadConstraints:
      - maxSkew: 1
        topologyKey: topology.kubernetes.io/zone
        whenUnsatisfiable: DoNotSchedule
        labelSelector:
          matchLabels: { app: web-app }
---
apiVersion: v1
kind: Service
metadata: { name: web-app, namespace: myapp }
spec:
  selector: { app: web-app }
  ports: [{ port: 80, targetPort: 8080 }]
  type: ClusterIP

Ingress with Managed SSL

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: web-ingress
  namespace: myapp
  annotations:
    kubernetes.io/ingress.class: "gce"
    networking.gke.io/managed-certificates: "web-cert"
    kubernetes.io/ingress.global-static-ip-name: "web-static-ip"
spec:
  rules:
  - host: app.example.com
    http:
      paths:
      - path: /
        pathType: Prefix
        backend:
          service: { name: web-app, port: { number: 80 } }
---
apiVersion: networking.gke.io/v1
kind: ManagedCertificate
metadata: { name: web-cert, namespace: myapp }
spec:
  domains: [app.example.com]

gcloud compute addresses create web-static-ip --global

Terraform Configuration

resource "google_container_cluster" "primary" {
  name     = "prod-cluster"
  location = "us-central1"

  release_channel { channel = "REGULAR" }
  workload_identity_config { workload_pool = "${var.project_id}.svc.id.goog" }

  network    = google_compute_network.vpc.name
  subnetwork = google_compute_subnetwork.gke.name

  ip_allocation_policy {
    cluster_secondary_range_name  = "pods"
    services_secondary_range_name = "services"
  }

  private_cluster_config {
    enable_private_nodes   = true
    master_ipv4_cidr_block = "172.16.0.0/28"
  }

  network_policy { enabled = true }
  logging_config { enable_components = ["SYSTEM_COMPONENTS", "WORKLOADS"] }
  monitoring_config {
    enable_components = ["SYSTEM_COMPONENTS", "WORKLOADS"]
    managed_prometheus { enabled = true }
  }

  remove_default_node_pool = true
  initial_node_count       = 1
}

resource "google_container_node_pool" "primary" {
  name     = "primary-pool"
  cluster  = google_container_cluster.primary.name
  location = "us-central1"

  initial_node_count = 2
  autoscaling { min_node_count = 1; max_node_count = 5 }
  management  { auto_repair = true; auto_upgrade = true }

  node_config {
    machine_type = "e2-standard-4"
    disk_size_gb = 100
    disk_type    = "pd-balanced"
    oauth_scopes = ["https://www.googleapis.com/auth/cloud-platform"]
    shielded_instance_config {
      enable_secure_boot          = true
      enable_integrity_monitoring = true
    }
    metadata = { disable-legacy-endpoints = "true" }
  }
}

resource "google_compute_subnetwork" "gke" {
  name          = "gke-subnet"
  ip_cidr_range = "10.0.0.0/20"
  region        = "us-central1"
  network       = google_compute_network.vpc.id

  secondary_ip_range { range_name = "pods";     ip_cidr_range = "10.4.0.0/14" }
  secondary_ip_range { range_name = "services"; ip_cidr_range = "10.8.0.0/20" }
}

Common Operations

gcloud container clusters list
gcloud container clusters upgrade prod-cluster --region=us-central1 --master
kubectl top nodes && kubectl top pods --namespace=myapp
kubectl scale deployment web-app --replicas=5 --namespace=myapp
kubectl autoscale deployment web-app --namespace=myapp --min=3 --max=20 --cpu-percent=70
kubectl logs -f deployment/web-app --namespace=myapp --all-containers

Troubleshooting

| Symptom | Cause | Fix |

|---------|-------|-----|

| Pods stuck in Pending | No nodes with enough resources | Check autoscaler; add larger node pool; verify resource requests |

| ImagePullBackOff | Wrong image path or missing AR access | Verify image URL; grant roles/artifactregistry.reader to node SA |

| Workload Identity wrong account | KSA annotation missing |

… (nội dung đầy đủ có trong gói tải về)