IPv6 First, Not Just Ready: Kubernetes Without IPv4 Using Cilium
summary
ESnet runs an IPv6‑only on‑prem Kubernetes cluster using K3s + Cilium and proves production traffic works but needs specific underlay and Cilium features: DNS64/NAT64 or a NAT46x64Gateway to reach IPv4 services, BGP + load‑balancer IPAM for stable ingress addresses, and Hubble for flow debugging. Early design pins pods to nodes for predictable egress; cluster‑pool/native routing works but causes masquerade pain. multiple IPAM per‑namespace /64 arrives in 1.19 (usable in 1.18) and removes many of those tradeoffs.
AI-written summary · about
chapters (10)
- 0:00ESNet and Kubernetes IPv6 journey
- 3:10IPv6-only Kubernetes motivation
- 6:01Bootstrapping K3S IPv6-only
- 9:14IPv6 app deployment pitfalls
- 13:53Predictable IPv6 egress design
- 16:00Cilium 1.18 IPv6-only improvements
- 18:11Key takeaways and lessons
- 23:10Network observability with Cilium
- 24:35Hardcoded IPv4 in apps
- 25:17Q&A
related talks
40mOur Experiences Deploying Kubernetes with IPv6
26mSupercharge Kubernetes Networking with Cilium and IPv6
28m100Gbit/S Clusters With Cilium: Building Tomorrows Networking
18mThe Road to IPv6 Support in...
35mNext-level Kubernetes Networking with Cilium - Nico Meisenzahl and Philip Welz
32mIPv6 / Dual-Stack in Kubernetes - Why, When, Where and How?
22mNavigating the Kubernetes Networking Maze
24mThe Cilium Story - Why We Created Cilium