Kubernetes 1.37, codenamed “Garhwal,” has been released, with the headline change being that KYAML is now Stable. It is a stricter and less ambiguous subset of YAML for Kubernetes configurations and remains fully compatible with standard YAML. The kubectl get -o kyaml output option is also now Stable.
The metrics.k8s.io API has also finally reached Stable after nearly nine years in Beta. It provides CPU and memory usage data for Pods and nodes and is used by features such as HorizontalPodAutoscaler and the kubectl top command.
Speaking of autoscaling, HPA scale-to-zero support is now Beta and enabled by default. Workloads using external or object metrics can set minReplicas to zero. This allows Kubernetes to scale down idle applications completely and restart them when demand returns.
Kubernetes 1.37 also expands Dynamic Resource Allocation, with several DRA capabilities reaching Stable and additional support for workload-level resource claims.
Another notable Beta addition is gang scheduling, which allows groups of related Pods to be scheduled together only when enough resources are available for the entire workload. The feature is aimed particularly at distributed AI, machine learning, and HPC jobs.
On the security side, Pod certificates and ClusterTrustBundles are now Stable, providing native mechanisms for delivering certificates and trust information to workloads.
The release also makes the StorageVersionMigration API Stable and enables improved SELinux volume handling by default, while Prometheus native histogram support moves to Beta.
Administrators should note some deprecations. The old kube-dns add-on is formally deprecated in favor of CoreDNS, and kube-proxy’s IPVS mode has also entered the deprecation path.
Finally, Kubernetes continues its move away from cgroup v1, with users strongly encouraged to migrate to cgroup v2 ahead of its eventual removal.
Kubernetes 1.37 is available now. For more details, refer to the official release announcement and full changelog.
