Nutanix Kubernetes Platform (NKP) 2.19

Nutanix Kubernetes Platform (NKP) 2.19 expands Kubernetes deployment and management with bare-metal support, simplified cluster upgrades, and new capabilities for AI workloads. The release also adds Kubernetes 1.36 support, AHV virtual machine management, and improvements to AI Navigator.

Administrators upgrading existing environments should review changes to Prism Central projects, container registry authentication, and deprecated components before proceeding.

Action Required: Prepare for NKP 2.19

Several changes require attention before installation or upgrade.

• Prism Central Projects 2.0: Clusters must use a single Prism Central project across control plane and worker nodes. Before upgrading NKP 2.18 clusters spanning multiple projects, consolidate all nodes into one project.

• DockerHub authentication: Management clusters without an existing Nutanix DockerHub registry token must provide one during upgrade.

• Catalog migration: GitRepository-based catalogs are no longer supported. Migrate remaining AppDeployment resources to OCIRepository-based catalogs.

• Removed CLI flag: The --kommander-applications-repository flag is no longer supported for nkp install kommander or nkp upgrade kommander.

• Kubernetes Dashboard: NKP replaces Kubernetes Dashboard with Headlamp. Existing Dashboard instances remain running after upgrade, but Nutanix recommends migrating to Headlamp as soon as possible.

The nkp upgrade workspace command is also deprecated. Use the supported cluster or platform application upgrade workflows instead.

What's New and Improved

Deploy Kubernetes Directly on Bare Metal

NKP 2.19 introduces NKP Metal, enabling Kubernetes deployment and management on physical servers.

Using Foundation Central and the NKP host inventory, administrators can image and register hosts, assign them to workspaces, and create bare-metal clusters. Physical worker node pools can also be added to VM-based clusters.

This provides direct access to compute, networking, storage, and accelerators for AI, GPU computing, and edge workloads.

Simplified Cluster Upgrades

ClusterClass workload clusters, including Nutanix and EKS clusters, now support per-cluster upgrades through the NKP UI or a single CLI command.

The workflow upgrades platform applications before Kubernetes, cluster infrastructure, and related add-ons.

For non-ClusterClass clusters, the new nkp upgrade platform-apps command upgrades platform applications independently.

The management cluster must first be upgraded to NKP 2.19.

Manage Virtual Machines Alongside Containers

NKP 2.19 introduces AHV virtual machine management through VM Operator.

Platform teams can provision, update, monitor, and manage Linux and Windows VMs using Kubernetes APIs and declarative workflows.

This brings VM and container lifecycle management into a common Kubernetes operational model.

Expanded AI and GPU Capabilities

NKP 2.19 adds a curated AI application catalog for Pro and Ultimate environments, including KAI Scheduler, Slurm Operator, Kueue, Kubeflow Pipelines, JupyterHub, and Milvus Operator.

The release also supports GPU scheduling and slicing through Kubernetes Dynamic Resource Allocation for NVIDIA and AMD GPUs. NVIDIA multi-instance GPU partitioning is not included.

AI Navigator and MCP Improvements

The NKP Model Context Protocol (MCP) Server is now enabled by default on management clusters with Pro or Ultimate licenses in internet-connected environments.

AI Navigator can access additional Kubernetes, Flux GitOps, and NKP resources through the MCP Server, with Kubernetes RBAC controlling access.

Additional improvements include dynamic resource discovery, configurable resource exposure, session metrics, and PostgreSQL connection recovery.

Fixes and Reliability

NKP 2.19 resolves several issues affecting cluster operations:

• Cluster readiness: Fixes etcd learner promotion failures that prevented new clusters from becoming ready.

• Upgrade reliability: Resolves control plane upgrades stalled by validating webhooks and registry preflight checks that ignored --skip-preflight-checks=Registry.

• Node provisioning: Fixes expired Prism Central session tokens that left new workload cluster nodes in the NotReady state.

• Certificates: Corrects ACME certificate issuance failures involving the Traefik ingress class.

• Application management: Resolves workspace application update failures involving more than 10 clusters.

NKP Insights 1.9 also improves security alert forwarding, restores tracked resource state after restarts, and introduces image-tag CVE reports.

Availability and Compatibility

NKP 2.19 supports Kubernetes 1.36.2, installed by default, along with the supported Kubernetes 1.35.x release family.

Additional compatibility includes:

• Operating systems: Rocky Linux 9.8 and Red Hat Enterprise Linux 9.8.

• Prism Central: pc.7.5 and pc.7.6.

• AOS: Versions 7.5 and 7.6.

• Nutanix Data Services for Kubernetes: NDK 2.3.

• Nutanix Enterprise AI: NAI 2.8.0 and 2.8.1.

Administrators should review known limitations before upgrading. In particular, certain single-instance Nutanix AI configurations can prevent worker nodes from draining during cluster upgrades.

Learn More

For complete installation guidance, compatibility requirements, and upgrade procedures, see:

• Nutanix Kubernetes Platform Upgrade Guide

• NKP Compatibility and Interoperability Matrix

• Generating a Docker Hub Access Token

• Workspace Catalog Applications

• Upgrading Nutanix Enterprise AI to 2.8.x on NKP

Review the complete NKP 2.19 release notes before upgrading production environments.