admin
920c1aaf84
A fresh one-disk install (512 GB VM, installer 1.33.0, agent 0.158.1 + golden 0.318.0 vouched): data volume 100 GiB, data area 213 GiB registered as the default drive with no press, Jellyfin with no extra step, nothing of the area in a whole-box copy; partitions formatted and used; the honest backup offer on a VM; /mnt/sys_drive never offered (no box had it). 16 §13 build B status; capability map D15; poster facts (drawing needs a refresh, STATUS); the backup page HU+EN now that it is true; REPORT-storage-identity-buildB-2026-10-16.md. Claude-Session: https://claude.ai/code/session_012qRErfCoiTkvDK9N5XHbzb
Storage identity, build B — live evidence (2026-10-16)
Design: architecture/16-storage-physical-identity.md §5–§7, §9 (§13 = what was built). Releases: agent v0.158.1
(v0.158.0 superseded by a security review), controller v0.318.0, installer 1.33.0 (published), golden 0.318.0
(documentation/tests/golden-0.318.0-2026-10-16/). The operator vouched agent 0.158.1 + golden 0.318.0 and raised the
global floor to 0.318.0 / MinAgent 0.158.1.
Method: endpoint level (no browser on DooPlex) — the dashboard requests the pages make (/api/stacks/<app>/deploy,
/api/storage/init, /api/storage/attach, /api/disks/candidates, /backups, /claim, /api/server/poweroff), the agent's
local API with the guest's own token (/disks, /disks/candidates, /disks/same-disk, /backup), the VM console
(screendumps) for the Proxmox install. Times are UTC.
Part 1 — the D3 space check
SPACE-CHECK.md + space/ — ruled by the operator 2026-10-16: data volume 50/100/130 GiB, reserve max(16 GiB, 10 %) ≤ 48 GiB.
Part 2 — live
| What | Result | File |
|---|---|---|
| A fresh one-disk install (scratch VM 342 on demo-hp, 512 GB disk, ISO 1.29.0, the public installer 1.33.0, customer Tester4) | pool 383 GiB → data volume 100 GiB, internal place created, 213 GiB; day-0 SUCCESS in 3 min | 05-vm342-create.txt, 07-fresh-install.log |
| The agent mounts it, the controller registers it | /mnt/belso (by fs-UUID) → bound at /mnt/felhom-drives/belso; the guest config has no mount point for it; controller 0.318.0: „registered /mnt/felhom-drives/belso (default=true)" |
15-box-logs.txt |
| Jellyfin with no extra step | deploy with DOMAIN + SUBDOMAIN only → 202, healthy in 30 s, /media = /mnt/felhom-drives/belso/userdata/media (the OMV run on 0.314.0: 400 „HDD_PATH … kötelező") |
08-jellyfin-deploy.txt |
| Not in the whole-box copy | a 64 MB marker in the place → 0 entries in the archive (control: 29,512 var/lib/felhom/docker entries); vzdump „excluding bind mount point mp8" |
09-… |
| The place is the system disk | same-disk internal vs sys_drive: same disk, proven (kind internal, unknown/vm — a VM; build A never counts such a copy) |
10-… |
| R-1082 | the attach list is empty while the controller's own mount table holds /mnt/sys_drive (0.314.0 offered it) |
10-… |
| Partitions (a 16 GB scratch disk, two parts, part 2 ext4) | both listed, formattable; part 1 formatted + registered (/mnt/felhom-drives/resz1); then nothing of the disk is offered for formatting, part 2 offered for USE only and attached (resz2); same-disk part 1 vs part 2: same disk, proven; formatting the whole disk: 409 in words |
11-…14- |
| R-1084 | the offer on a VM: „Ezt a meghajtót kijelölheted a rendszermentés helyéül. A Felhom nem tudja ellenőrizni … ezért nem ígérhetjük, hogy egy lemezhiba után is vissza tudod állítani a rendszert." | 14-… |
| Delivery to the three test boxes | Tester 1, demo-hp, N100: agent 0.158.1 + bundle 0.158.1 (through the step bundle 0.158.1-step1), controller 0.318.0 by the floor |
— |
Not as the design said (filed or recorded)
- An empty part asks for the typed confirmation („device has a partition table"):
lsblkreports the parent disk's table for a partition, so the agent's probe reads an empty part as data-bearing. Safe direction, wrong friction → R-1093. - The whole-disk refusal sentence ends „Csak a szabad részét formázhatod" — not true while another part is mounted (the free part can then only be used) → R-1094.
- A 200 GB disk gets no internal place under the ruled sizes (pool 113 → 113 − 16 − 82 = 15 GiB < 20); the brief's
200 GB VM could not show it, so the VM was rebuilt with 512 GB (
05-…). - Tester 1's dashboard password in the restored credentials file does not open Tester 1 — the dashboard checks ran on the Tester4 scratch box instead (also a VM, same shape).
Teardown
Machine: VM 342 switched off the household way (/api/server/poweroff, alarms held) and destroyed; Tester 1's 16 GB scratch
disk unlinked and deleted. Host: the helper agentq.sh removed from Tester 1's host (VM 342 is gone with its copy). Hub: the
three per-customer floors of build A cleared (303 floor_set); Tester4's host is off — the operator deletes the throwaway
customer. Secrets: scan of this folder for the phrase, the dashboard password, the setup code and the token prefix = none.