Homelab

The NAS as storage for the cluster

Four 4 TB disks in RAID 10, two NFS exports, and a place for Proxmox’s weekly dumps that every node sees.

The NAS is a Debian 13 box that held one large disk until this month. Now it has four 4 TB disks and two jobs for the cluster: a place for Proxmox’s own backups, and the media library for Jellyfin. Both go over NFS.

The array

Four disks in RAID 10 give 7.3 TiB and survive a failed disk in each pair. RAID 5 would give more space, and a rebuild that reads every disk for a day.

$ sudo mdadm --create /dev/md0 --level=10 --raid-devices=4 /dev/sda /dev/sdb /dev/sdc /dev/sdd
$ sudo mkfs.ext4 -L srv /dev/md0
$ sudo mount /dev/md0 /srv
$ df -h /srv
Filesystem      Size  Used Avail Use% Mounted on
/dev/md0        7.3T   28K  6.9T   1% /srv

Two exports

/srv/dump   10.0.10.11(rw,no_root_squash) 10.0.10.12(rw,no_root_squash) 10.0.10.13(rw,no_root_squash)
/srv/media  10.0.10.0/24(ro)

no_root_squash is needed because Proxmox writes its dumps as root, so that export names the three nodes and nobody else. The media share is read-only for the whole network. Files reach it over SMB from the laptops.

$ sudo exportfs -ra
$ showmount -e nas
Export list for nas:
/srv/media 10.0.10.0/24
/srv/dump  10.0.10.11,10.0.10.12,10.0.10.13

Storage in Proxmox

Storage added on one node is there on all three, because it lives in /etc/pve/storage.cfg.

$ pvesm add nfs nas-dump --server 10.0.10.20 --export /srv/dump --content backup

A backup job under Datacenter, Backup now writes every guest to nas-dump on Thursdays at 22:00. I left retention at its default.

The rack: switch, pve-1 to pve-3, the NAS, a blank panel and the UPS at the bottom.
The NAS sits below the mini PCs. The blank panel is where the tower’s shelf was.

Comments

Buy Logbuch