NonVPCnative

NonVPCnative, often described as non-VPC native, refers to a Kubernetes networking mode in which the pod network does not use the Amazon VPC for IP addressing. In this mode, pods receive IP addresses from a separate pod network CIDR managed by the cluster’s container network interface (CNI) plugin, such as kubenet, Flannel, or Calico, and pod-to-pod traffic is routed through an overlay or host-based mechanism rather than through the VPC’s routing table. As a result, pod IPs are not directly routable within the VPC and do not belong to the VPC’s IP space.

This contrasts with VPC native networking (also called IP per pod), where the AWS VPC CNI plugin assigns each pod a VPC IP. In VPC native mode, pods can be addressed from other VPC resources and are subject to VPC security groups and route policies, enabling closer integration with AWS networking features.

NonVPCnative is often associated with legacy Kubernetes networking configurations or with environments where a non-VPC CNI is preferred for compatibility with certain plugins, network policies, or operational constraints. It can offer simpler IP management in some contexts and may avoid consuming VPC IP addresses, but it also limits direct integration with AWS networking features such as VPC security groups on a per-pod basis and can introduce additional networking overhead for pod communication.

Migration and considerations: clusters currently configured for non-VPC native networking may require planning to switch to VPC native, including reconfiguring CNI, potential Pod/Service CIDR changes, and careful rollout to minimize disruption. Understanding the chosen CNI and its implications for traffic flow, security, and scalability is essential when working with nonVPCnative configurations.