storage

storage follows the same product thinking as compute: the kubernetes proof path is zfs on the hypervisor, longhorn inside the cluster, and object storage for backups and terraform state. capacity nas experiments stay secondary to that practiced path.

hypervisor storage

the proxmox host uses a mirrored zfs pool for vm disks, snapshots, and related lab data — checksums, compression, and single-drive failure tolerance without turning the lab into a fragile museum.

—zfs mirror
—vm disks + snapshots
—proxmox storage pool

kubernetes volumes

longhorn provides the default storage class for stateful workloads, with enough replication to survive node churn in a six-node lab cluster.

—longhorn default storageclass
—stateful pvc path

backup & state

velero sends cluster backups to self-hosted minio. terraform remote state also lives on that object store — recovery and infrastructure history share a deliberate backend.

—velero → minio
—terraform state bucket
—restore drills

broader capacity ideas

truenas-style nas and larger media datasets remain part of the wider lab conversation — performance vs capacity, dedicated vs virtualized nas — without confusing them for the kubernetes platform’s current storage path.

← back to homelab·ai lab →