Skip to content

Longhorn vs central storage architecture — homelab / farm platform

Purpose

Doctrine — storage comparison — in Platform doctrine package — homelab / farm edge. Choose between Longhorn (distributed Kubernetes block storage) and central storage (NAS/NFS/object outside the cluster) for farmOS, telemetry stacks, and media—without prescribing a vendor SKU. Longhorn’s scope alone: Longhorn role in the homelab / farm platform.

Sources: Longhorn CSI on K3s, homelab Pi + Longhorn capture, Longhorn system backup. Compare also Backup strategy comparison.

When Longhorn fits

  • You want PVCs portable across k3s nodes with replication and snapshots in the Kubernetes control path.
  • Nodes have dedicated disk for Longhorn (Pi narrative: SSD per node).
  • You accept iSCSI + Longhorn operator burden (CSI on K3s).

When central storage fits

  • Large sequential files (video, bulk imagery) where NFS/SMB or object is simpler than block PVCs.
  • Single strong NAS with UPS and backup—good for read-mostly archives.
  • You want fewer moving parts inside Kubernetes for a specific dataset.

Hybrid (common)

  • Longhorn for stateful workloads (farmOS DB PVC, broker PVC ).
  • Central NFS or S3-compatible target for backup repositories (restic) and media—see Kubernetes platform backup / DR.

```mermaid flowchart LR subgraph k3s["k3s"] app["farmOS / stateful"] end lh["Longhorn PVCs"] nas["Central NAS / object"] restic["restic / off-cluster backup"]

app --> lh lh --> nas restic --> nas ```

“HA storage” in this wiki

  • Longhorn replication ≠ geographic HA—it is survive a node disk / host loss within a LAN.
  • Central NAS HA (dual-controller, etc. ) is a different product class—not assumed here.