The Road to IPv6 Support in...
summary
kOps adds IPv6 cluster support by using single‑stack pods with dual‑stack at the edge, initially using unique‑local addresses then switching to globally routable IPv6, and implements a node prefix controller that writes per‑VM /64 prefixes into Node objects so CNIs can assign pod IPs. Workarounds include enabling dns64 and nat64, patching the AWS cloud provider and launch templates, fixing an Ubuntu systemd bug that drops addresses after five minutes, and noting GCE auto‑assigns a */96 prefix.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 0:41Why IPv6
- 1:42Single Stack
- 2:30Unique Local Addresses
- 3:07Ready, but not quite
- 5:07Pivot to global addresses
- 9:01Custom node IPAM controller
- 10:54Public Topology
- 11:37Private Topology
- 12:53Subnet assignments
- 13:21Remaining work
- 14:09Status of GCE support
- 15:53CNI Evolution for IPv6?
related talks
17mThe Long Road to IPv4/IPv6 Dual-stack Kubernetes
26mIPv6 First, Not Just Ready: Kubernetes Without IPv4 Using Cilium
32mIPv6 / Dual-Stack in Kubernetes - Why, When, Where and How?
40mOur Experiences Deploying Kubernetes with IPv6
06mKeynote: And Here We Go: Dual-stack Networking in Kubernetes
22mkOps: Recent Advances and the Journey to an OSS Kubernetes Distribution
31mDual-stack Kubernetes with Direct server Return
36mAdapting Kubernetes to Constrained IP Address Environments