Future Plans¶

Planned¶

Nothing is planned right now. The previous round of plans wrapped up in Q3 2026 (see below).

Upgrading to more powerful hardware: replace the Lenovo M920q units and the self-built server with three Minisforum MS-01 units as Proxmox VE hosts.

Done differently: the cluster now runs on 2 × Lenovo M920x and 1 × Lenovo M90q Gen 5, going from 15 to 48 CPU cores. See Hardware.
Expanding network capacity: add an aggregation switch, because the existing 10G SFP+ ports were at capacity. This also follows networking best practices.

Done: a UniFi USW Aggregation now connects the cluster nodes and the NAS.
Dedicated NAS hardware: move TrueNAS from a virtualized setup with hardware passthrough to bare metal on the existing 3U server.

Done: TrueNAS SCALE runs bare metal. See Hardware → NAS.
Better power management: upgrade to a more powerful UPS and add a managed PDU for better power distribution and management.

Done: a UniFi UPS 2U (1500 VA) is in the rack and monitored by nut-exporter.

Building a distributed storage foundation: run Ceph directly on the Proxmox VE cluster, with Kubernetes consuming it through rook-ceph-operator only.

No more virtualization: the cluster is bare metal now, and Rook-Ceph runs inside Kubernetes. Proxmox and Ceph competing for disk I/O was... not great. See Storage.
Optimizing inter-node connectivity: 20G Thunderbolt networking between the cluster nodes, plus dedicated 10G SFP+ links for virtualized Kubernetes nodes.

Not possible with the current hardware.