Kubernetes 1.37: What’s New, Key Features and Upgrade Guide

Kubernetes 1.37 “Garhwal” introduces important improvements for production workloads, autoscaling, observability, security, resource management and cluster administration. Learn what changed and how to prepare your cluster for the upgrade.

Kubernetes continues to evolve rapidly, and the latest release, Kubernetes v1.37, brings several changes that are especially relevant for DevOps engineers, platform teams, and Kubernetes administrators.

Released on August 26, 2026, Kubernetes v1.37 is named Garhwal and contains 67 enhancements, including 16 Stable, 23 Beta, and 27 Alpha enhancements. This release also includes important changes around autoscaling, metrics, memory management, security, storage migration, and node configuration.

In this guide, we will look at the most important Kubernetes 1.37 changes and the practical things you should check before upgrading a production cluster.

What Is Kubernetes 1.37?

Kubernetes 1.37 is a new minor release of the Kubernetes container orchestration platform. It is actively supported and was released on August 26, 2026.

The release focuses on improving Kubernetes for modern production environments, including better workload scaling, resource management, observability, security and cluster operations.

Kubernetes 1.37 is particularly interesting for teams running production workloads because several long-running features have moved to Beta or Stable status.

Key Kubernetes 1.37 Features

1. HPA Can Scale Workloads Down to Zero

One of the notable changes in Kubernetes 1.37 is the improved HorizontalPodAutoscaler (HPA) support for scaling workloads down to zero replicas.

This can be useful for workloads such as queue consumers, batch processors, and event-driven services where keeping an idle Pod running is unnecessary.

In Kubernetes 1.37, the HPAScaleToZero feature is enabled by default. Scaling to zero requires a suitable object or external metric.

kubectl describe hpa queue-worker

Keep in mind that scaling to zero can introduce cold-start latency. It is therefore more suitable for workloads where a short startup delay is acceptable.

2. Metrics API Graduates to Stable

The Kubernetes Resource Metrics API has graduated to the stable metrics.k8s.io/v1 API in Kubernetes 1.37.

This API provides CPU and memory usage information for Nodes and Pods and is used by tools such as kubectl top and resource-based HorizontalPodAutoscaling.

kubectl top nodes
kubectl top pods

The stable API improves the long-term stability of clients that consume Kubernetes resource metrics. It is important to note that this API is not a replacement for a complete monitoring solution such as Prometheus and Grafana.

3. Memory QoS Moves to Beta

Memory QoS has graduated to Beta in Kubernetes 1.37 and is enabled by default.

On Linux nodes using cgroup v2, Memory QoS allows Kubernetes to provide the Linux kernel with additional information about how container memory should be handled.

This is particularly relevant for production environments where memory pressure and workload isolation are important.

Teams upgrading to Kubernetes 1.37 should verify that their Linux nodes are using cgroup v2 and review their existing resource requests and limits.

4. Pod-Level Resource Managers Move to Beta

Kubernetes 1.37 also promotes Pod-Level Resource Managers to Beta. This allows the CPU Manager, Memory Manager and Topology Manager to work with Pod-level resource declarations.

This can be useful for workloads that require predictable CPU and memory placement, especially on systems where NUMA topology and dedicated resources matter.

The feature is currently disabled by default, so production teams should evaluate it carefully before enabling it.

5. Rootless Kubelet Moves to Beta

The KubeletInUserNamespace feature, also known as rootless mode, has graduated to Beta in Kubernetes 1.37.

When enabled, Kubernetes node components can operate using a Linux user namespace instead of requiring root privileges on the host.

This feature can be interesting for environments where reducing host-level privileges is an important security requirement.

6. Storage Version Migration Is GA

Storage Version Migration has graduated to General Availability in Kubernetes 1.37.

Storage Version Migration helps Kubernetes migrate stored API resources to their current storage versions. This is particularly useful when APIs evolve and older versions need to be removed safely.

For platform teams operating clusters with many CustomResourceDefinitions (CRDs), understanding storage version migration can help reduce problems during API lifecycle changes.

7. Pod Certificates and Cluster Trust Bundles

Kubernetes 1.37 also brings important improvements to workload identity. Pod Certificates and Cluster Trust Bundles provide built-in mechanisms for certificate-based workload identity.

This is particularly relevant for applications that use TLS or mutual TLS (mTLS) to communicate securely with other services.

Important Change: cgroup v2

One of the most important upgrade considerations in Kubernetes 1.37 is the continued move away from cgroup v1.

Kubernetes 1.37 defaults the kubelet configuration option failCgroupV1 to true. Nodes that still rely on cgroup v1 can therefore fail to initialize unless a temporary configuration override is applied.

The recommended long-term approach is to migrate Linux nodes to cgroup v2 rather than relying on the compatibility override.

apiVersion: kubelet.config.k8s.io/v1beta1
kind: KubeletConfiguration
failCgroupV1: false

The configuration above is intended as a temporary compatibility option. Production environments should plan their migration to cgroup v2 instead.

Kubernetes 1.37 Upgrade Checklist

Before upgrading a production cluster, it is important to verify both the Kubernetes components and the workloads running on the cluster.

Check the current Kubernetes version.
Confirm that your current version is supported for the upgrade path.
Review Kubernetes 1.37 release notes.
Check deprecated and removed APIs.
Verify that worker nodes support cgroup v2.
Check CNI compatibility.
Check CSI and storage plugin compatibility.
Verify ingress controller compatibility.
Review monitoring and logging integrations.
Back up important cluster data.
Test the upgrade in a non-production environment first.
Check PodDisruptionBudgets.
Review workloads with strict resource requirements.
Verify custom controllers and operators.
Plan a maintenance window for production.

Checking Your Current Kubernetes Version

Before starting an upgrade, check the Kubernetes client and server versions.

kubectl version
kubectl get nodes -o wide

You can also check the versions of all nodes:

kubectl get nodes

Basic Cluster Health Checks

Before upgrading, make sure the cluster is healthy. Start by checking Nodes and Pods.

kubectl get nodes
kubectl get pods -A
kubectl get events -A --sort-by=.lastTimestamp

Look for Nodes in NotReady state, Pods stuck in Pending, repeated container restarts, failed probes, or unexpected warning events.

Upgrading a kubeadm Cluster

The exact upgrade process depends on how your cluster was installed. For clusters created using kubeadm, the Kubernetes project provides a dedicated upgrade procedure.

At a high level, a Kubernetes upgrade involves upgrading the control plane, upgrading worker nodes, updating client tools such as kubectl, and addressing API or manifest changes.

kubectl drain <node-name> --ignore-daemonsets --delete-emptydir-data

After maintenance or replacement of a node, verify that it returns to the Ready state:

kubectl get nodes

Do not blindly copy upgrade commands into a production environment. The exact commands depend on your Kubernetes installation method, operating system, package repositories, container runtime, CNI and other cluster components.

Common Problems to Check Before an Upgrade

cgroup v1 Nodes

Older Linux nodes using cgroup v1 are an important consideration for Kubernetes 1.37. Verify the cgroup configuration of every node before upgrading.

Deprecated APIs

Applications, Helm charts and operators can continue using older Kubernetes APIs even when the cluster itself has been upgraded. Search your manifests and installed applications for APIs that are deprecated or removed in the target version.

Third-Party Controllers

Operators, admission controllers, ingress controllers, monitoring agents and storage plugins may have their own Kubernetes version requirements. Always verify compatibility before upgrading.

Resource Configuration

Review CPU and memory requests and limits before the upgrade. Resource configuration becomes increasingly important as Kubernetes introduces more advanced resource management capabilities.

Production Validation After the Upgrade

After upgrading the cluster, do not immediately assume that everything is working correctly. Perform a structured validation.

Verify all control-plane components.
Verify that all Nodes are Ready.
Check system Pods.
Check application Pods.
Verify ingress and load balancers.
Test DNS resolution.
Check persistent volumes.
Verify monitoring and alerting.
Check application logs.
Run smoke tests against critical applications.

kubectl get nodes
kubectl get pods -A
kubectl get deployments -A
kubectl get pvc -A

Should You Upgrade to Kubernetes 1.37?

Kubernetes 1.37 is an actively supported release and introduces several changes that are useful for modern production environments.

However, a production upgrade should be based on compatibility testing rather than simply upgrading because a new version is available.

Start by testing Kubernetes 1.37 in a development or staging environment. Validate your workloads, CNI, CSI, ingress, monitoring, operators and automation before moving the same upgrade process into production.

Conclusion

Kubernetes 1.37 is an important release for DevOps and platform engineering teams. The release includes improvements across autoscaling, metrics, memory management, resource management, security and storage.

Features such as HPA scale-to-zero, the stable Metrics API, Memory QoS, Storage Version Migration and improved workload identity make Kubernetes 1.37 particularly relevant for teams operating modern cloud-native infrastructure.

The most important point for administrators is preparation. Check cgroup compatibility, deprecated APIs, third-party components, workloads and cluster backups before starting the upgrade.

A successful Kubernetes upgrade is not just about upgrading the control plane. It is about validating the entire application platform around it.