Operational how-to: Longhorn distributed block storage

Sub-issue of #630.

Scope

A how-to for installing, configuring, and operating Longhorn distributed block storage on NKP.

New page under docs/docs/how-tos/.

Content:

• When to use Longhorn vs cloud-native volumes, including which providers install it
• Configure via the longhorn block and apply with nic deploy
• Node maintenance procedures (drain, uncordon)

Cross-links:

• Architecture explainer (#625)
• Back up and restore data (merged in #709) already covers Longhorn backups in depth. This page should link to it, not restate it.
• Provider pages that already mention Longhorn: docs/docs/how-tos/providers/aws.mdx, docs/docs/how-tos/providers/hetzner.mdx

Verified against source

Checked against nebari-infrastructure-core @ 199bc37 (2026-08-28). These supersede the design docs where they disagree:

• Provider support. Longhorn is installed by the AWS, Hetzner, and existing-cluster providers, which all call the shared pkg/storage/longhorn package. Azure, GCP, and the local provider hardcode LonghornEnabled: false.
• Chart version. Pinned to 1.11.2 in pkg/storage/longhorn/config.go.
• Config knobs. enabled, replica_count, dedicated_nodes, node_selector, instance_manager_cpu_percent.
• dedicated_nodes contract. When set, every storage node group must carry the labels from longhorn.StorageNodeLabels. NIC emits a loud warning when no node carries the create-default-disk label.
• CLI surface. deploy, destroy, validate, version, kubeconfig, outputs. There is still no nic status.

Acceptance criteria

• Page includes concrete commands, not just prose
• Page states which providers install Longhorn and which do not
• Page notes NKP-only scope (Classic equivalent stays under /classic/ if present)
• Backups are cross-linked to the existing guide rather than duplicated
• Cross-links wired

Reference

Repos:

• nic CLI: https://github.com/nebari-dev/nebari-infrastructure-core

[!WARNING]
These design docs drift from the implementation. Treat the nic source code as the source of truth and verify every claim against it before writing. Two confirmed drifts as of 2026-08-31: the node maintenance doc still says Longhorn "is currently only wired into the AWS provider" (it is wired for Hetzner and existing clusters too), and older guidance references a nic status command that does not exist.

Design docs (audit PR nebari-infrastructure-core#301 merged 2026-07-28; the worktree-design-doc-audit branch is deleted, so the links below point at main):

• Longhorn design decision: docs/adr/0002-longhorn-distributed-block-storage-for-aws.md. Note this was not part of the #301 audit; it was last touched 2026-06-08. It is titled "for AWS" but carries a 2026-05 amendment expanding the scope.
• Node maintenance: docs/design-doc/operations/longhorn-node-maintenance.md

Source of truth:

• Shared package: pkg/storage/longhorn/
• Provider callers: pkg/providers/cluster/aws/longhorn.go, pkg/providers/cluster/hetzner/provider.go, pkg/providers/cluster/existing/provider.go