Kubernetes 1.35 Releases In-Place Pod Resize as Stable Feature
The Kubernetes project has announced the general availability (GA) of the In-Place Pod Resize feature in version 1.35, marking a significant milestone six years after its initial conception. Previously, adjusting CPU and memory resources for containers required deleting and recreating Pods, causing disruptions for stateful or latency-sensitive workloads. This new stable feature allows users to modify resource requests and limits for running Pods without restarting containers, thereby enhancing resource efficiency and operational flexibility. Key improvements in this release include the ability to decrease memory limits, provided current usage is below the new threshold, and the implementation of prioritized resize requests based on PriorityClass and QoS when node resources are constrained. Additionally, the update supports Vertical Pod Autoscaler (VPA) integration for seamless automatic scaling and introduces alpha support for Pod Level Resources. This development enables use cases such as dynamic scaling for game servers, efficient bin-packing, and handling transient resource spikes like JIT compilation during startup, significantly reducing downtime and improving workload management.
Wire timeline
Kubernetes 1.35 Releases In-Place Pod Resize as Stable Feature
The Kubernetes project has announced the general availability (GA) of the In-Place Pod Resize feature in version 1.35, marking a significant milestone six years after its initial conception. Previously, adjusting CPU and memory resources for containers required deleting and recreating Pods, causing disruptions for stateful or latency-sensitive workloads. This new stable feature allows users to modify resource requests and limits for running Pods without restarting containers, thereby enhancing resource efficiency and operational flexibility. Key improvements in this release include the ability to decrease memory limits, provided current usage is below the new threshold, and the implementation of prioritized resize requests based on PriorityClass and QoS when node resources are constrained. Additionally, the update supports Vertical Pod Autoscaler (VPA) integration for seamless automatic scaling and introduces alpha support for Pod Level Resources. This development enables use cases such as dynamic scaling for game servers, efficient bin-packing, and handling transient resource spikes like JIT compilation during startup, significantly reducing downtime and improving workload management.
Kubernetes Blog