The Kubernetes project is formally deprecating the kube-proxy IPVS backend starting with version 1.35, and it plans complete removal by version 1.40. Users relying on IPVS should prepare to transition to either the nftables or iptables backends, which are now the preferred and maintained solutions.
- IPVS backend deprecated in Kubernetes 1.35, removal planned for 1.40.
- Migration to nftables or iptables backends advised to maintain compatibility and support.
- IPVS removal addresses maintenance complexity and better aligns with Kubernetes service requirements.
Infrastructure signal
The IPVS backend for kube-proxy was originally developed to address performance bottlenecks in iptables by providing faster data plane connection establishment using hash tables and reducing control plane latency by avoiding full ruleset rewrites. However, its complex architecture, incompatibility with some Kubernetes Service edge cases, and dependence on the limited IPVS API have revealed scaling and maintenance issues over time.
In response, the Kubernetes project introduced a nftables backend beginning in version 1.29, achieving general availability by version 1.33. This backend is designed to consolidate the roles of both iptables and ipvs with better alignment to service APIs and improved maintainability. The deprecation and eventual removal of IPVS by Kubernetes 1.40 signals a strategic shift toward streamlined proxy infrastructure, impacting cluster networking design and toolchains dependent on kube-proxy.
Developer impact
Developers and platform teams using Kubernetes clusters must plan migrations from the IPVS backend to nftables or iptables to ensure ongoing support and compatibility. The IPVS backend’s incomplete service functionality implementation and architectural divergence made it harder to maintain, so transitioning will help reduce unexpected network behavior and improve reliability.
From a workflow perspective, teams may experience changes in how network rules and service load balancing are managed, with nftables offering more modern and efficient mechanisms. This also means updates to deployment configurations, observability tooling, and possibly network policy definitions to accommodate backend-specific behavior differences related to rule management and API capabilities.
What teams should watch
Infrastructure and cloud platform teams should monitor upcoming Kubernetes releases closely to avoid disruption from the planned removal of the IPVS backend in 1.40. Evaluating current usage of IPVS in clusters and assessing the complexity of migrating to nftables or iptables is critical to maintain cluster stability and service consistency.
Additionally, teams should watch for changes in observability and metrics related to service proxies because switching backends might alter how connection and update latencies appear. Understanding nftables’ capabilities and limitations versus the legacy IPVS setup will be important for tuning network performance and troubleshooting.