Azure Kubernetes Fleet Manager Gets Cilium Cross-Cluster Networking
Cloud Platform EngineeringAzure Kubernetes Fleet Manager now delivers Cilium-based cross-cluster networking — a managed, high-performance network that can span an entire fleet of AKS clusters without requiring custom overlay solutions or third-party service mesh installations.
What Fleet Manager Cross-Cluster Networking Solves
Multi-cluster Kubernetes architectures on Azure have historically required each cluster to manage its own network boundary, with cross-cluster communication routed through external load balancers or ingress controllers. This architecture works but creates operational friction: services cannot discover each other across clusters without DNS hacks, network policies cannot span cluster boundaries, and workload portability between clusters requires application-level awareness of the cluster topology.
Azure Kubernetes Fleet Manager addresses the fleet orchestration layer — propagating configurations, managing upgrades, and now extending the network plane across clusters. The Cilium-based cross-cluster networking capability uses eBPF rather than iptables for packet processing, which reduces latency and eliminates the kube-proxy overhead that accumulates at scale across large node pools.
The Cilium Architecture
Cilium’s eBPF networking provides several properties relevant to multi-cluster deployments: kernel-level packet processing without sidecar proxies, identity-aware network policy (based on workload identity rather than IP addresses), and native support for cross-cluster service load balancing. When Fleet Manager deploys Cilium-based networking across the fleet, services can be exposed to other clusters with consistent policy enforcement — the same network policy that governs internal cluster traffic governs cross-cluster traffic.
For platform teams already running ingress-level distributed tracing with W3C Trace Context, cross-cluster Cilium networking adds the condition that trace propagation can work natively across cluster boundaries — the traceparent header follows requests through the eBPF layer without custom instrumentation at each cluster edge.
The So What
Platform teams managing multiple AKS clusters for environment isolation, regional distribution, or blast radius reduction now have a managed path to fleet-wide governance and networking in a single control plane. The practical sequence: migrate existing multi-cluster setups to Fleet Manager governance first, then layer on cross-cluster networking as the eBPF plane stabilizes across clusters. The alternative — maintaining custom Submariner or Skupper overlays — has a higher operational cost and no Azure SLA backing. Cross-cluster Cilium through Fleet Manager is the managed path forward for AKS-first platform teams.
Source: Azure Blog
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →