Disks — four array drives, the NVMe root, the rescue stick

Identify by serial or UUID. Letters reshuffle across boots and cable swaps; /etc/fstab mounts by UUID so a reshuffled letter is harmless, a missing UUID is not.

Why this shape

Three data + one parity is SnapRAID's minimum single-parity layout; parity must be ≥ the largest data disk (all four are 4 TB). The mix of models is history, not design — the HGST is 3.2× older than the rest.

Shape

Dev (09-08)	Mount	Serial	Model	Notes
sda1	/mnt/disk1 (d1)	Z4F0100N	HP MB4000GEFNA	3.66M inodes (rsnapshot forest); Feb-2026 defect band near block 702.5M — swept read-only 2026-09-09, 0 unreadable; HP firmware exposes no attr 197/198/199
sdb1	/mnt/parity	PN1334PEKE2DWS	HGST HUS724040ALA640 7200 rpm	78k+ power-on hours; lifetime max 80 °C; hd-idle spins it down (rests 31–34 °C); replace-on-schedule
sdc1	/mnt/disk2 (d2)	WD-WCC7K7ZVKKLY	WD40EZRZ (Blue)	idle3 timer disabled 09-03, latched 09-05; Load_Cycle_Count frozen at 450,430 (was +460/day)
sdd2	/mnt/disk3 (d3)	WD-WCC4E0TJN09N	WD40EZRX (Green)	same fix, frozen at 182,461; sdd1 is a stray 200 MB EFI partition — leave it
nvme0n1p2	/	placeholder	512 GB DRAM-less Realtek	every container config, precious/, 4th content file; 54 of 73 power cycles were unsafe
sde1	/mnt/rescue	0701859E39D3CF74	Verbatim Store'n'Go 14.4 GB	ext4, nofail; sentinel .rescue-usb-id must match its UUID

• fstab: all four array lines defaults,noatime,errors=remount-ro (/etc/fstab); backups /etc/fstab.bak.2026-09-03*.
• ext4 reserve: parity 0 %, data disks 1 % (item 07, 633 GiB returned).
• Fill (09-08): d1 47 %, d2 41 %, d3 43 %, parity 65 %, pool 44 % (4.6 of 11 TB).
• Chassis: one 930 rpm fan, ~32 °C ambient, no drive-bay fan ever existed.

Connected to

• owns: snapraid-array, mergerfs-pool
• owned-by: host/machine
• looks-like-but-is-not: ~/disk-uuids.txt (Feb-14 pre-reformat UUIDs: lists sda1 as d649dac7…, live is bc6af84b…; lists sde as a vfat QBIT-BACKUP stick) — leftover. ~/disk-layout-before-replacement.txt, ~/fstab-backup.txt — leftover.

If you change this

• Hits: the fstab UUID line · data dN path in /etc/snapraid.conf · hd-idle by-id (parity only, /etc/default/hd-idle) · /etc/smartd.conf uses letters /dev/sdX — re-check after any cable or drive change · ~/disk-map-*.txt · rescue kit host/disks.txt.
• Does not hit: docker-compose.yml (mounts /mnt/storage, never a disk) · the mergerfs fstab line (paths, not devices).

Surfaces

Surface	Role
smartd	polls every 30 min, appends to /var/lib/smartmontools/attrlog.<model>-<serial>.ata.csv (world-readable: read history here, no root, no spin-up)
scrutiny	5 devices incl. /dev/nvme0; web UI
hd-idle	parity only, by serial
snapraid-precheck	SMART health + pending/offline sectors on sda–sdd before every sync

See

• Source: /home/angel/disk-map-20260908.txt · /etc/fstab · /etc/smartd.conf