Home / Azure

AzureChecklistIntermediate

High availability for apps on AKS: zones, replicas, spread constraints and disruption budgets

Cluster and workload settings that let an AKS application survive node drains, upgrades and a full zone outage, including topology spread, PodDisruptionBudgets, probes and zone-aware storage.

Deepak Bhardwaj2 min read

Applies toAzure Kubernetes Service (Standard tier, zone-enabled regions)

Cluster level

  • Standard tier (API server SLA; 99.95% with zones).
  • Node pools across zones 1 2 3, minimum 3 nodes per pool.
  • Cluster autoscaler with enough maximum to absorb a lost zone.
  • Separate system pool with CriticalAddonsOnly taint.
  • Planned maintenance windows and maxSurge on node pools.

Workload level

  • At least 3 replicas for anything user-facing.
  • Spread replicas across zones and nodes:
spec:
  replicas: 3
  template:
    spec:
      topologySpreadConstraints:
        - maxSkew: 1
          topologyKey: topology.kubernetes.io/zone
          whenUnsatisfiable: DoNotSchedule
          labelSelector: { matchLabels: { app: orders-api } }
        - maxSkew: 1
          topologyKey: kubernetes.io/hostname
          whenUnsatisfiable: ScheduleAnyway
          labelSelector: { matchLabels: { app: orders-api } }
  • PodDisruptionBudget so drains and upgrades keep enough pods running:
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata: { name: orders-api }
spec:
  minAvailable: 2
  selector: { matchLabels: { app: orders-api } }

Warning

A PDB with minAvailable equal to the replica count (or maxUnavailable: 0) blocks node drains forever and makes AKS upgrades fail. Leave room for at least one pod to move.

  • Readiness probes (traffic only to ready pods) and liveness probes with generous thresholds; startup probes for slow starters.
  • Requests and limits set, so the scheduler places pods realistically.
  • Graceful shutdown: handle SIGTERM, preStop sleep of a few seconds so the load balancer stops sending traffic first.

Storage

  • Azure Disk PVs are zonal (LRS disks): a pod with a disk can only run in that zone. For stateful apps use ZRS disks (skuName: Premium_ZRS StorageClass) or app-level replication.
  • Azure Files (ZRS) for shared storage that must follow pods across zones.
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata: { name: managed-csi-premium-zrs }
provisioner: disk.csi.azure.com
parameters: { skuName: Premium_ZRS }
volumeBindingMode: WaitForFirstConsumer
allowVolumeExpansion: true

Ingress and egress

  • Ingress controller (or Application Gateway for Containers) with 2+ replicas across zones.
  • Load balancer Standard SKU, zone-redundant frontend.
  • Egress via hub firewall (zone-redundant) or one NAT gateway per zone.

Verify

kubectl get pods -l app=orders-api -o wide
kubectl get nodes -L topology.kubernetes.io/zone
kubectl get pdb -A

Run a drain on all nodes of one zone in a test window (kubectl drain per node) and confirm the app keeps serving.

Related fixes