Free Preview

Question 1 of 5 · Sample from the AZ-305 bank · Read-only

Azure Kubernetes Fleet Manager for Multi-Cluster Hub Topology

A platform team operates 22 AKS clusters across 6 regions for different business units. They need to: (1) coordinate Kubernetes version + node-image upgrades across the fleet in waves with bake time between rings, (2) propagate a baseline set of Kubernetes objects (namespaces, NetworkPolicy, ResourceQuota, Flux GitRepository) from a single source of truth to all member clusters, and (3) load-balance an HTTP workload across multiple member clusters with a single multi-cluster service.

Question: What should you recommend?

  • A.

    Use Azure Policy for Kubernetes initiatives plus Azure Front Door per workload — no fleet abstraction is needed
  • B.

    Deploy GitOps (Flux) independently to each cluster and write a custom Logic App to sequence AKS cluster upgrades
  • C.

    Provision an Azure Kubernetes Fleet Manager hub cluster, join all 22 AKS clusters as members, and use Fleet Manager update runs / update strategies for staged upgrades, ClusterResourcePlacement (CRP) for workload + config propagation, and the Fleet multi-cluster L4 load balancer for cross-cluster traffic

     ✓ correct
  • D.

    Migrate everything to a single very large AKS cluster with multiple node pools per region
EXPLANATION

Azure Kubernetes Fleet Manager is purpose-built for this exact set of problems. A Fleet hub (with or without a hub cluster) registers member AKS clusters and provides: update runs / update strategies / update stages for orchestrated, ringed Kubernetes + node-image upgrades with bake/wait gates between waves; ClusterResourcePlacement (CRP) to propagate Kubernetes objects from the hub to selected members based on labels (the GitOps-of-config layer Microsoft recommends for fleets); and a multi-cluster Layer-4 load balancer that exposes a service across multiple member clusters from a single VIP. Option A doesn't handle staged upgrades or workload propagation. Option B re-implements Fleet Manager poorly and doesn't give multi-cluster services. Option D collapses a fleet into one cluster, losing the regional/BU isolation that drove the design. See: https://learn.microsoft.com/en-us/azure/kubernetes-fleet/overview

Purchase access $25 Log in ← Home