Kubernetes
Making Kubernetes the Network Control Plane: Public Service IPs with Cilium, BGP and ExternalDNS
Moving the public identity of a service into Kubernetes, from address allocation and route advertisement to packet forwarding and DNS.
Kubernetes
Moving the public identity of a service into Kubernetes, from address allocation and route advertisement to packet forwarding and DNS.
Network
An anonymized engineering case study of an XDP/eBPF multi-exit router, focusing on partial TCP observation, bounded policy execution, fast/slow-path consistency, and failures involving verification, VLAN forwarding, neighbor resolution, and concurrent state updates.
Kubernetes
This article explores the challenges of deploying Calico in eBPF mode for Kubernetes, focusing on issues like the need for an external load balancer, VIP certificate configurations, and circular dependencies when the control plane fails. Solutions for these challenges are also discussed.
Kubernetes
MetalLB (controller-only) with Calico BGP does not bind VIPs to any node interface, causing ARP failures. Services stay unreachable externally. Solution: locally bind VIPs, enable strictARP, and disable rp_filter to restore correct traffic flow.
Kubernetes
This article examines how to build a self-hosted Kubernetes cluster on bare-metal infrastructure, focusing on networks, high availability, DNS, and storage. It targets academic infrastructure engineers managing clusters under 50 nodes, contrasting with managed cloud solutions.
Kubernetes
Deploy MetalLB as controller-only and let Calico’s BGP advertise a LoadBalancer pool. The guide covers Helm setup, IP pool config with avoidBuggyIPs, route flow analysis, source-IP-preserving Local policy, and failover tests for production-grade bare-metal LB.