Hardware¶

The physical boxes, their upgrade paths, and the planned NAS swap.

Proxmox host — Dell OptiPlex 7060 MFF ("pve")¶

Component	Current	Notes
CPU	Intel i5-8500T (6C/6T)	Includes UHD 630 with QuickSync — the media-server enabler
RAM	16 GB DDR4-2666 SODIMM	Upgrade path: 2×16 GB or 2×32 GB DDR4-2666 SODIMM
Storage	256 GB NVMe	Local storage is deliberately small — real capacity is the NAS
Role	Proxmox VE, all VMs/LXCs	See topology for the guest list

Planned: Jellyfin LXC (after the RAM upgrade)

The plan of record for a media server: Jellyfin in an LXC with /dev/dri passed through so QuickSync handles transcoding — the i5-8500T's iGPU makes this box a capable media server without a dGPU. Blocked on the RAM upgrade first: 16 GB is already tight with four guests running.

Operational debt on this box

pve is running a ~140-day-old kernel and root@pam is still single-factor. Both are on the pending list in Security posture.

Current NAS — Synology (192.168.1.5)¶

7 TB, MAC 90:09:d0:0b:bf:42. Its entire role in the estate is one NFS export, /volume1/pbs, holding both backup datasets:

dump/ — nightly vzdump archives from pve
kopia/ — the kopia repository, mounted on ops at /mnt/nas-kopia

Layout details and retention in Backups overview.

Planned swap — UniFi UNAS Pro 4¶

The Synology will be replaced by a UNAS Pro 4. The backup stack was built so this is a one-file repoint, not a rebuild: every target reference lives in /etc/backup-stack/backup-targets.env on ops.

Migration steps¶

NAS-side prep (UNAS):
- Create the share and recreate the layout: a dump/ and a kopia/ directory.
- Enable NFS, export restricted by IP to pve (192.168.1.138) and ops (192.168.1.190) only — and this time don't carry over the stale PBS entry (192.168.1.201).
- Copy the data across (both dump/ and kopia/) while the Synology is still up.
On ops: edit /etc/backup-stack/backup-targets.env to point at the new NAS (IP / export path), then run backup-stack-repoint.sh. It rewires the kopia mount, the Proxmox NFS storage, and the sync/monitor config from that single env file.
Verify before trusting it (green exit codes have lied before — see Backups overview):
- Next 02:00 vzdump lands in the new dump/.
- A kopia snapshot from any endpoint succeeds and a restore drill passes.
- ntfy digest shows no stale sources the following morning.

Never host a PBS datastore on NAS NFS

Whatever the new NAS is: the PBS-datastore-on-NFS failure mode (silently empty backups, all monitoring green) is why PBS was retired. vzdump-to-NFS and kopia-on-NFS are fine; a PBS datastore is not.

Endpoint hardware quick reference¶

Device	MAC	WoL
pc1 — Elliott gaming PC	`34:5a:60:ce:33:17`	Yes — `wol-wake.sh pc1` on ops
pc2 — Rory gaming PC	`30:56:0f:44:c7:a7`	Yes — `wol-wake.sh pc2` on ops
tablet — diagnostic	`ec:4c:8c:96:da:4d`	WiFi-attached → WoL unreliable; plan is Veeam wake-from-sleep

Automated wake timers are gated off (WOL_ENABLED=0 on ops) until Veeam is configured — see Onboarding.