SeriesA guided journey · 5 parts

Upgrade to NKP 2.18

A complete, step-by-step path for taking a Nutanix Kubernetes Platform fleet to 2.18 in production: prerequisites and registry, Kommander and platform apps, the Loki storage migration, the node roll, and the GitOps cutover. Follow it in order, or jump to the step you are on.

New here? Start with the overview →

Parts

28 min

Total read

Advanced

Level

NKP 2.18 / CAPI / Flux

Stack

Track your progress as you read each part

1Prerequisites, Artifacts

2Kommander, Platform Apps

3Loki v2

4Kubernetes Node Roll

5GitOps Migration

The path

Each part is a self-contained article. Together they form the runbook we use to move a managed NKP fleet from 2.x to 2.18 without a dreaded maintenance weekend.

Not started Part 1 Jul 4 ·4 min

Prerequisites, Artifacts and the Registry Push

Everything you need in place before typing nkp upgrade: compatibility gates, cluster readiness checks, the air-gapped bundle layout, and pushing 400+ images to your private registry without surprises.

nkpkubernetesnutanix

Read Part 1 →

Not started Part 2 Jul 5 ·5 min

Kommander, Platform Apps and the Workspaces

The platform phase of an NKP 2.18 upgrade: what nkp upgrade kommander actually does, why its timeouts are normal, how to find the app it is silently waiting for, and the three failure classes our workspaces produced.

nkpkubernetesnutanix

Read Part 2 →

Not started Part 3 Jul 6 ·10 min

The Loki v2 to v3 Migration with Custom S3 Storage

Our NKP 2.18 upgrade froze at 26/27 platform apps because the automated Loki v2-to-v3 migration could not convert our custom Nutanix Objects storage override. Here is the converter's undocumented input contract, mapped one error message at a time.

nkpkubernetesnutanix

Read Part 3 →

Not started Part 4 Jul 7 ·3 min

The Kubernetes Node Roll (and the PDB That Froze It)

Upgrading the management cluster from Kubernetes 1.34 to 1.35.2 on NKP 2.18: the new preflight checks, a cluster-name gotcha from the docs, how to actually watch a node roll, and the 67 minutes a single PodDisruptionBudget cost us.

nkpkubernetesnutanix

Read Part 4 →

Not started Part 5 Jul 8 ·6 min

GitOps Migration to the NKPCluster API

After upgrading to NKP 2.18, a routine git commit to scale a Flux-managed workload cluster gets rejected by an admission webhook. Your repo has lost write access to its own cluster. Here is why, what the rejection looks like, and the migration that gives GitOps its power back.

nkpkubernetesnutanix

Read Part 5 →

Upgrade to NKP 2.18

The path

Prerequisites, Artifacts and the Registry Push

Kommander, Platform Apps and the Workspaces

The Loki v2 to v3 Migration with Custom S3 Storage

The Kubernetes Node Roll (and the PDB That Froze It)

GitOps Migration to the NKPCluster API

More series