Kubernetes 1.37 Released: Stable Metrics API and Rootless Kubelet in Beta

The Cloud Native Computing Foundation (CNCF) has announced the release of Kubernetes 1.37, named Garhwal, with a focus on stability, security, and AI/ML workload optimization. One of the notable features in this version is the general availability of the Metrics API, which offers a standardized method for monitoring and measuring node and pod health, supporting auto-scaling features like Horizontal Pod Autoscaler (HPA) and Vertical Pod Autoscaler (VPA), as well as kubectl top commands.

Additionally, Kubernetes 1.37 introduces workload-aware scheduling, enabling the scheduler to free up resources on overloaded nodes by preempting low-priority workloads, allowing high-priority applications to be scaled in-place. Another feature in this release is the ability to create pod-level checkpoints and restore pods from them, aiding in debugging and security analysis by allowing practitioners to snapshot a container's memory and process trees.

The Recreate strategy for StatefulSet rollouts, which deletes all existing pods, is also entering the alpha phase. Kubelet operation in the user namespace or Rootless mode has been promoted to beta in Kubernetes 1.37, reducing the risk of container escape attacks. This feature is enabled by default, allowing core node components to run as non-root users on the host using Linux user namespaces.

Resilient watchcache initialization has also become generally available in Kubernetes 1.37, preventing excessive requests to etcd during kube-apiserver restarts, and reducing the potential for control plane outages. Furthermore, pod certificates have graduated to stable status, providing native support for secure pod-to-pod communication using mTLS without external tools.

A total of 67 enhancements, 27 alpha features, 23 graduating to beta, 16 becoming generally available, and 1 deprecation were included in Kubernetes 1.37. Users can find detailed release information in the official release notes and documentation.