AKS Cluster Configuration

Once you understand the fundamentals of AKS, the next step is configuring your cluster for production use. This lesson covers node pools, autoscaling, upgrades, identity management, and the configuration options that turn a basic cluster into a production-ready platform.

---

Node Pools

Node pools are groups of virtual machines with the same configuration. AKS supports two types:

System Node Pools

System pools run essential Kubernetes components (CoreDNS, metrics-server, tunnelfront). Every cluster must have at least one system pool.

Best practices:

• Use a small, reliable VM size (e.g. Standard_D4s_v5)
• Set a minimum of 2–3 nodes for high availability
• Apply a CriticalAddonsOnly=true:NoSchedule taint to prevent application workloads from running on system nodes

User Node Pools

User pools run your application workloads. You can create multiple pools with different characteristics:

# Add a GPU node pool for ML workloads
az aks nodepool add \
  --resource-group rg-aks \
  --cluster-name my-aks-cluster \
  --name gpupool \
  --node-count 2 \
  --node-vm-size Standard_NC6s_v3 \
  --labels workload=gpu

# Add a spot node pool for batch processing
az aks nodepool add \
  --resource-group rg-aks \
  --cluster-name my-aks-cluster \
  --name spotpool \
  --node-count 0 \
  --min-count 0 \
  --max-count 10 \
  --enable-cluster-autoscaler \
  --priority Spot \
  --eviction-policy Delete \
  --spot-max-price -1 \
  --node-vm-size Standard_D4s_v5

Node Selectors and Taints

Use labels and taints to control which workloads run on which node pools:

# Deployment targeting the GPU pool
spec:
  nodeSelector:
    workload: gpu
  tolerations:
    - key: "kubernetes.azure.com/scalesetpriority"
      operator: "Equal"
      value: "spot"
      effect: "NoSchedule"

---

Autoscaling

AKS provides three layers of autoscaling:

1. Horizontal Pod Autoscaler (HPA)

Scales the number of pod replicas based on CPU, memory, or custom metrics.

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: my-app-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  minReplicas: 2
  maxReplicas: 20
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 70

2. Cluster Autoscaler

Adds or removes nodes from a node pool based on pod scheduling demand.

# Enable cluster autoscaler on a node pool
az aks nodepool update \
  --resource-group rg-aks \
  --cluster-name my-aks-cluster \
  --name userpool \
  --enable-cluster-autoscaler \
  --min-count 2 \
  --max-count 10

When pods cannot be scheduled due to insufficient resources, the cluster autoscaler adds nodes. When nodes are underutilised, it removes them.

3. KEDA (Kubernetes Event-Driven Autoscaler)

KEDA scales based on external events — queue length, database connections, HTTP traffic, cron schedules, and many more.

# Install KEDA add-on
az aks update \
  --resource-group rg-aks \
  --cluster-name my-aks-cluster \
  --enable-keda

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: queue-processor
spec:
  scaleTargetRef:
    name: queue-worker
  minReplicaCount: 0
  maxReplicaCount: 50
  triggers:
    - type: azure-servicebus
      metadata:
        queueName: orders
        messageCount: "5"

---

Cluster Upgrades

AKS supports both automatic and manual Kubernetes version upgrades.

Checking Available Versions

az aks get-upgrades \
  --resource-group rg-aks \
  --name my-aks-cluster \
  --output table

Manual Upgrade