What’s New in Kubernetes 1.35? Vertical Scaling and 60+ Enhancements Explained

New Feature: Vertical Scaling

Kubernetes 1.35 introduces in‑place vertical scaling, allowing users to adjust CPU and memory resources of a pod without restarting it. This capability benefits stateless workloads that can simply add more replicas, while stateful workloads face additional complexity because data and business logic reside on the node being upgraded.

Previously, vertical scaling required creating a new pod and migrating the workload, which could cause service interruption. The new feature eliminates the need for pod recreation, enabling seamless resource adjustments.

Other Notable Additions in v1.35

KYAML (Beta) : A concise subset of YAML optimized for Kubernetes, making configuration easier for operators and improving clarity for tools like kubectl. (KEP #5295)

Group Scheduling (Alpha) : Provides an all‑or‑none scheduling strategy for large, inter‑dependent AI workloads, ensuring a cluster has sufficient resources to host the entire group before scheduling any pod. (KEP #4671)

Variable Container Resources for Pending Jobs (Alpha) : Allows users to modify resource requests and limits of jobs that are pending due to insufficient CPU or memory, preserving execution history instead of deleting and recreating the job. (KEP #5440)

Deprecations and Removals

The release deprecates the long‑standing Ingress NGINX controller in favor of the Gateway API, prompting many operators to migrate. Support for Linux cgroup v1 is also removed, with the system now requiring cgroup v2.

Logo Design – “Planting a Tree”

Each Kubernetes release features a new logo designed by the release manager. For v1.35, engineer Drew Hagen chose a tree to symbolize the resilience of the Kubernetes community.