docs: two designs, proposed and not built — the off-site copies of a removed app end after 30 days (D72), the installer below 200 GB (D75); hub: the account record never reaches the box's config
gates / gates (push) Successful in 2m9s

Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
This commit is contained in:
admin committed 2026-11-14 18:44:26 +01:00
1 parent 73d99d06c5
commit a044eb8c16
7 files changed
+182 -3

No files matched your search

@@ -1835,6 +1835,89 @@ minutes). **[FACT]** 9202 rc.6: an image that exits after 20 s → not started,
unit back → success, copy released 3 minutes later. Tester 1 on 0.343.0: a unit restore and a set-aside restore, each
„ran stably for 3m0s … the restore copy is released" (`audits/release-2026-10-26/`).
### 8.2f The off-site copies of a removed app end 30 days after „delete backup data” `[DESIGN — proposed, not built]` (D72, `09` §3 decision 292, R-1374 part 1; written 2026-11-14)
**The ruling.** When the household removes an app and presses „delete backup data” (`remove_backups`), that app's
off-site copies end 30 days later, at the next clean-up after that day — the number of a restore copy (D62). When the
household keeps the app's backups, the off-site copies stay; no end is named for them, by ruling. The page must show
the end date BEFORE anything is deleted.
**Today (read in source, 2026-11-14).** „Delete backup data” deletes the unit, every second-drive mirror and the
app's backup preferences at once, and never touches an off-site snapshot (R-474). The weekly clean-up runs
`forget --group-by host,tags` with keep 7 daily / 4 weekly / 6 monthly per (host, app tag) — and an app that is gone
makes no new snapshots, so its group keeps its last ones for ever (`backup/offbox_window.go:80-82`). Every off-site
snapshot carries the marker tag `felhom-offbox` and one tag per app it belongs to (`offbox_inventory.go:64`). The box
can delete off-site only inside the weekly window the hub opens, by snapshot id, behind the guard and under the
window's cap (decisions 68–69).
**The design, in five parts.**
1. **The record.** At a removal with „delete backup data” the controller writes an *off-site end record* for the app:
`{app, removed_at, ends_at = removed_at + 30 days}`, in its settings (it survives a controller restart). No record
= no deletion, ever. A guest rebuild loses the record: the copies then stay (the safe direction).
2. **The household sees the date before it presses, and can take it back.** The remove dialog names the date under
the tick box. Until the copies are gone, the off-site list on the Backups page (which is keyed on the store, not on
what is installed — R-237) shows the app with „marked for deletion: <date>” and one button to keep it after all;
the button deletes the record. Nothing else in the product starts or shortens the clock.
3. **The deletion rides the weekly window, nothing else.** At the first window on or after `ends_at`, the box adds to
the window's plan the ids of the snapshots **whose app tags are all ended apps** (a snapshot that also belongs to
a living app stays until that app ends too). The removal is `forget <ids>` and the window's own `prune`, the calls
the window makes today. The window's result names them apart (`removed_for_ended_apps`), so the hub's 07:10 audit
explains the vanished files and raises nothing.
4. **The hub agrees first — the box cannot use this to delete a living app's history.** The box reports its pending
end records. The hub allows an app in a window only when ITS OWN records agree: the app has been absent from that
box's reported app list for 30 days by the hub's clock (not the box's). A box that claims an end the hub cannot
see gets the app refused, and the window says so. This keeps decision 68's custody: a taken-over box gains no new
way to remove copies.
5. **Local copies are unchanged.** They are deleted at once, as today. The text says both.
**Edge cases, each with its rule.**
| Case | Rule | Why |
|---|---|---|
| The app is installed again before the date | The install deletes the end record. The old copies are restore sources again („use my kept data”) | the household uses the app; deleting its history a few days later would be a surprise |
| The app is installed again after the copies are gone | Nothing special: a new history starts | — |
| The household kept the backups at removal | No record. The copies stay with no end | the ruling: it follows the household's own choice |
| The household changes its mind | „Keep it after all” until the window has run; after that the copies are gone and the page says so | a deletion is not undoable off-site |
| The box is off at the date, or the off-site store cannot be reached | The next window that runs does it | the record does not age |
| The off-site tier is switched off before the date | The record stays and does nothing; if the tier comes back, the next window does it | no call to a store that is off |
| The household is deleted or RESET before the date | Nothing to do: the account goes with all copies | `05` §14.1 |
| A snapshot carries two apps' tags | It stays until BOTH have ended | a living app's copy is never removed by another app's end |
| The `_shares` pseudo-app and the infrastructure stacks | Never eligible | they are not apps a household removes |
| The window's cap (half the store; R-1294) | **Open — a decision:** the ended apps' snapshots count toward the cap, or have their own allowance | see below |
| The box's clock is wrong | The hub's clock decides (part 4) | R-1384 |
| An app removed with „delete backup data” BEFORE this is built | No record exists, so nothing is deleted. The off-site list shows the app with a button „delete in 30 days”: the household starts the clock itself | never a deletion the household was not told about |
| The guard refuses the window (a future-dated snapshot, a young removal) | The ended apps wait with everything else | one guard, no second path |
| The app's name is used by another catalog app later | The record is keyed on the app name AND `removed_at`; an install deletes it (first row) | same tag, new owner |
**What it costs if nothing is built:** the off-site copies of every removed app stay for ever; the paid space only
grows; the household that pressed „delete backup data” is told a thing that is not true off-site.
**The open decision (for the operator).** Do the ended apps' snapshots count toward the window's weekly cap?
*CC's pick: no — their own allowance, only for ids the hub agreed to (part 4).* Why: the cap exists to stop a
taken-over box from deleting a living history fast; the hub's own 30-day check does that job for this path, and under
the cap a household that removes one big app could stop every clean-up for weeks (R-1294's shape). The other option —
they count — needs no new rule, and costs a refused window, then the operator's one-shot grant, at each large removal.
**The texts — a proposal, NOT published.** No suffix on the date, so one sentence serves every date.
| Where | Hungarian | English |
|---|---|---|
| The remove dialog, under „Mentési adatok törlése” | „A helyi mentések most törlődnek. A távoli másolatok még 30 napig megmaradnak. Törlésük napja: {dátum}. Addig a Mentések oldalon visszavonhatod.” | "The local backups are deleted now. The off-site copies stay for 30 more days. They are deleted on: {date}. Until then you can take this back on the Backups page." |
| The off-site list, the app's line | „Törlésre kijelölve. Törlés napja: {dátum}.” | "Marked for deletion. Deleted on: {date}." |
| The button | „Mégis megtartom” | "Keep it after all" |
| After the button | „A távoli másolatok megmaradnak.” | "The off-site copies are kept." |
| After the deletion, in the list's history line | „A távoli másolatok törölve: {dátum}.” | "The off-site copies were deleted on {date}." |
| An app removed before this existed | „Ennek az alkalmazásnak a távoli másolatai megmaradtak.” + button „Törlés 30 nap múlva” | "The off-site copies of this app were kept." + "Delete in 30 days" |
No mail to the household at the deletion: it can do nothing with it (`08` §8.1). One operator-only event per app
(`offsite_ended_app_removed`: app, snapshots, bytes).
**Not measured, and what must be measured before a build:** nothing in this design is a new mechanism on the box —
the window removes by id today. To measure in the build: how long `prune` takes after a whole app's history goes
(the largest removal a window has made is 209 snapshots, refused), and that restic 0.19.1 frees the space in the same
window.
### 8.2c Decisions in code that no document recorded [DESIGN — confirmed by the operator 2026-10-25 (D44 = all picks, R-1265)]
Read at `file:line` on `main` (controller `321df1f`, hub `44d19042`) by the main session:
@@ -412,9 +412,58 @@ below 200 GB — refuse, or warn (R-1370). demo-hp (128 GB) stays as a test box.
| Helper images carry no version; the guest's journal has no stated cap; off-site restores have no limit | R-1372 |
| Old kernels: the clean-up is built for ring-0 boxes (agent v0.174.0); a ring-1 box still keeps every kernel | `11-os-updates.md` §5.11.1 |
| The parts of the host disk with no limit and no cleaner (H4, H5, H6, H13) | R-1087, R-1368 |
| The smallest disk is ruled (200 GB, `09` §3 decision 295). Nothing enforces it: what the installer does below 200 GB (refuse, or warn) is open | R-1370 |
| The smallest disk is ruled (200 GB, `09` §3 decision 295). Nothing enforces it: what the installer does below 200 GB is a written design with a pick (refuse, with one flag for the operator) — §14.5; it waits for the operator's yes | R-1370 |
| What to do by hand when a disk is full | `runbooks/disk-full.md` (R-1369) |
### 14.5 What the installer does on a system disk below 200 GB `[DESIGN — proposed, not built]` (the open point of D75, `09` §3 decision 295, R-1370; written 2026-11-14)
**The ruling.** The smallest system disk is 200 GB, with or without a backup drive. Open: what the installer does
below it — refuse, or warn.
**Today (read in source, 2026-11-14).** The installer reads the free space of the thin pool and WARNS below 120 GiB
(`RECOMMENDED_MIN_LVM_GIB`, „a recommendation, not a floor”; R-130); the install goes on. It reads no disk size.
A 75 GiB box installed fully on 2026-07-03.
**What „the system disk” is, and how its size is read.** The disk that holds the `pve` volume group — the one the
Proxmox installer wrote. Its size is a kernel fact (`/sys/block/<disk>/size`, in 512-byte sectors), read before
anything is changed. A disk sold as „200 GB” has 200,049,647,616 bytes by the makers' common formula; a virtual disk
made as „200G” has 200 GiB. So the line is **200 × 10⁹ bytes**: both pass, a 128 GB or 180 GB disk does not. NOT
MEASURED: no disk's sector count was read for this design; the three test boxes' disks are 128, 200 and 512 GB by
their labels (`audits/space-list-2026-11-10/`).
**The two options.**
| | A — refuse (CC's pick) | B — warn |
|---|---|---|
| What the household sees | The install stops before it changes anything: „This machine's system disk is 128 GB. Felhom needs 200 GB.” | The install goes on; one line in the installer's output, and one line on the hub's page for the box |
| What it protects | The promise is one number, and it is true for every box in a household | Nothing by itself: on the public image the installer runs with nobody watching, so a warning in its output is read by nobody |
| What it costs | A household with a 128 GB machine cannot join. That is the ruling's own cost | A 128 GB box gets a pool of 54 GiB for a guest that was promised 82 GiB (`16` §14.3); the guest's own space guards read the promise, so only the hub's thin-pool alarm stands before a full pool |
| The way round | One named flag for the operator, `--accept-small-disk`, that prints the cost and records it (`small_disk: true` in the install state and on the hub) | none needed |
**CC's pick: A, with the flag.** The reasons: the ruling chose one number so that it can be promised; a warning that
nobody reads keeps no promise; and the cost of B is paid by the household months later, as a box that stops its local
copies. The flag keeps demo-hp (128 GB, a test box by the ruling) and a reinstall of an existing small box possible —
by a person who was told.
**Edge cases, each with its rule.**
| Case | Rule |
|---|---|
| A BYO host (the owner's own Proxmox) | **Warn only, never refuse.** The disk layout is the owner's; there may be no `pve` volume group to read. The line names the 200 GB |
| A reinstall of a box that was accepted earlier | The same rule as a new install. The operator's flag is the way |
| The size cannot be read (no `pve` group, an unusual layout, a RAID) | Appliance: refuse with „cannot read the system disk's size” and name the flag. A refusal that can be explained is better than an install on an unknown disk |
| Two disks: a small system disk and a large second one | The rule reads the SYSTEM disk. The large disk becomes the internal place or a drive; it does not make the system disk larger |
| Exactly 200 GB by the label | Passes (the line is 200 × 10⁹ bytes) |
| A virtual machine (the OMV and test installs) | The same rule on the virtual disk; the tester makes the disk 200 GB |
| The public image, nobody watching | The refusal must reach a person: the box still pairs, and the hub's page for it reads „install refused: system disk 128 GB, needs 200 GB”. **Open in the build:** today a failed install's reason does not reach the hub's page — to be read before the build |
| The existing 120 GiB pool warning | Stays, as the second check (a 200 GB disk with a small pool is a broken install, not a small disk) |
**Where the refusal sits.** In the installer's preflight, with the other checks that run before any change, so a
refused machine is exactly as it was. One test per row above in `scripts/test_hostinstall.py`, the sizes fed through a
seam (no real disk).
**If nothing is decided:** the installer accepts any disk, as today; the 200 GB is a sentence in a document.
## Decisions in code no section records [DESIGN — confirmed by the operator 2026-10-25 (D44 = all picks, R-1265)]
Read at `file:line` on `main` by the main session:
@@ -42,3 +42,9 @@
account_record_test.go:40: a config that carries BOTH a ticked tier and a record: the record goes (the descriptor is the record): a record &{Type:shared User:u1-sub3 Host: UntickedAt:}, want none ({"offsite":{"enabled":true,"type"
1 mutations, 0 not a red-proof
== MUTATION: the hub's account record is merged into the box's config
RED on the mutant, GREEN on the real file
--- FAIL: TestGenerate_TheAccountRecordNeverReachesTheBox (0.00s)
offsite_account_record_test.go:23: the box's config carries the hub's account record ("offsite_account"):
1 mutations, 0 not a red-proof
+2 -2
View File
@@ -219,7 +219,7 @@ stopping line that lies.
| **R-1263** | Backup & restore | P4 | **Small items from review round 2:** a LOCAL restore's put-back mails the operator „its off-site restore FAILED” (`felhom-controller/controller/cmd/controller/main.go:1514-1518`, shared via `restore_unit.go:517-518`); Remove — even „keep my data” — deletes undeclared volumes labelled with the app's project, which no tier holds (`internal/stacks/delete.go:763-764`, R-1117); the capped copy's `--delete` reads the source in the daemon's mount view, the emptiness check in the controller's (`tier2.go:834` vs `:464`); stale comments — `felhom-agent/internal/localapi/backup_cap.go:15` („keep-last”) and `restore_crash_putback.go:32` (names a test file that does not exist); the qBittorrent password goes to whatever answers the DNS name `qbittorrent` (`upwatch/explain.go:94,108-113`); kept-copy Remove has a check-to-remove gap (`tier2_kept.go:92-111`); the release-copy restore check trusts hashes stored on ep0 (`runbooks/total-loss-of-dooplex.md`); rrsync without `-munge`/`-wo` (`audits/day-2026-10-23/08-release-copy-D33.txt:6`). | OPEN | — | fix in each repo's next attended pass | CC |
| **R-1294** | Backup & restore | P3 | **demo-hp's weekly off-site clean-up was refused by its own guard on 2026-10-26 (window 16): the plan would remove 209 of 359 snapshots, above the week's cap (179) — nothing was deleted, and each refused week adds to the next plan.** Seen in the operator mailbox during the 2026-10-26 release run (03:17Z, `offsite_prune_guard_refused`, error; details `count_before 359, count_after 359, outcome guard-refused`). The cap refusal is kept on purpose (R-824, R-867 closed the false ones); the open question is whether a store that once exceeds the cap can ever prune again by itself, or refuses every week while it grows. demo-hp is the Tier-0 box with many hand-made backups (sweeps, probes). Earlier cap refusals: demo-hp 2026-10-03 (window 1), demo-felhom 2026-10-05 (window 3). **2026-11-11:** the End-dates list leans on this row (`documentation/audits/end-dates-list-2026-11-11/TABLE.md` DB05). | **OPEN — the grant is SET (D55 = 260, `09` §3 decision 268)**: 2026-10-26 19:10:31Z through the operator's one-purpose token `cc-d55-window`, read back `max:260`; it is used at demo-hp's next window request (after its nightly off-site run, ~03:15Z). Close when that window prunes; then the operator revokes the token. **Earlier:** WAITING-ON-OPERATOR — 2026-10-26 evening: CC could NOT set the D54 grant — `POST /offsite/window-grant/demo-hp` (`max_remove=220`) with CC's hub token answered **403 „missing scope offsite:write"** (the hub log shows the refusal; nothing granted). The plan also grows: restic's dry-run 2026-10-26 18:56Z still 359 copies / plan 209 / default cap 179 (before tonight's run); a simulation of the keep rule adds ~14 per night (14 of 18 groups back up nightly), so ~223 at the night of 10-27 and ~237 at 10-28 — a cap of 220 would be refused again. STATUS D55: the operator grants it, with the cap. The grant opens the box's NEXT window request, which comes after its next nightly off-site run. Ruled D54 = prune once (`09` §3 decision 266). **Earlier:** WAITING-ON-OPERATOR — 2026-10-26 (read only, `audits/r1294-offsite-cap-2026-10-26/`): the plan is HONEST (restic's own dry-run: keep 150, remove 209; every removal 7–30 days old; every app keeps ≥ 9 or its last copies; the bulk is one test day, 2026-10-16, 152 snapshots). It can NOT recover by itself: the cap is half the store and grows ~49/week while the plan grows ~84/week. Only demo-hp has this shape (demo-felhom and Tester 1 pruned normally 10-26). The pick is STATUS D54. **Earlier:** OPEN **2026-11-10 (space list, `documentation/audits/space-list-2026-11-10/TABLE.md`):** place O1, listed from documents only (the off-site box was not read in this run — second pass). **2026-11-10:** the endings list leans on this row (`documentation/audits/endings-list-2026-11-10/TABLE.md` EC23-7). | — | read the guard's cap rule (controller off-site prune) against demo-hp's snapshot list: is the plan honest (old, superseded) or not; if honest, say how such a store gets back under the cap (an attended one-time prune, or a cap that follows the store) — a design question for the operator if it changes what a box may delete | CC |
| **R-1362** | Backup & restore | P3 | **With the internet down, the Backups page takes over 2 minutes (135 s and 145 s measured; 5 s with only the hub down).** Walked 2026-11-03 on Tester 1: every other page answers at once on the home network. The page waits on something off-site inside its handler; a household that lost its line and opens exactly this page sees a hang. `audits/security-2026-11-02/part5/outage-wan.log` | OPEN — filed 2026-11-03 | — | name the wait with a timing, then give it a short deadline and show „the off-site state could not be read" | CC |
| **R-1374** | Backup & restore | P3 | **Four kinds of copy are kept with no end named, and nobody is reminded of them.** Found 2026-11-10 (the endings list, `documentation/audits/endings-list-2026-11-10/TABLE.md` endings EC02, EC23, EC18, EC27; read in source; 9202 has no off-site store, so not measured). (1) **Off-site snapshots of a removed app**: the removal keeps them by ruling (closed R-474), and the weekly clean-up keeps the last snapshots of each app for ever (`internal/backup/offbox_window.go:80-82`), also for an app that is gone. No rule says when they end. (2) **A whole restore from the second drive** moves the old files beside the restored ones (`internal/backup/tier2_whole.go:52`, `:132`, `:345`); nothing removes them later and no page lists them. (3) **Exported app files** stay in the export folder of the drive; the pages can start, list and read an export, none deletes one, and nothing ages it (`internal/web/handler_export.go:30-87`). **Members (answers marked N):** EC02-4, EC23-4, EC18-4, EC27-4. | OPEN — filed 2026-11-10 (endings list) **2026-11-10 15:19 — RULED (D72 = yes, by the household's own choice; `09` §3 decision 292).** Not built. | — | Controller (D72): when the household pressed „delete backup data” for an app, its off-site copies end 30 days later, at the next clean-up after that day (the number of a restore copy, D62); when the household kept the backups, the off-site copies stay. The page shows the end date before anything is deleted — write that into `07-backup-architecture.md` before the build. Parts (2) and (3): a delete on the export page; a line on the kept-data page for the files a whole restore set aside. | CC |
| **R-1374** | Backup & restore | P3 | **Four kinds of copy are kept with no end named, and nobody is reminded of them.** Found 2026-11-10 (the endings list, `documentation/audits/endings-list-2026-11-10/TABLE.md` endings EC02, EC23, EC18, EC27; read in source; 9202 has no off-site store, so not measured). (1) **Off-site snapshots of a removed app**: the removal keeps them by ruling (closed R-474), and the weekly clean-up keeps the last snapshots of each app for ever (`internal/backup/offbox_window.go:80-82`), also for an app that is gone. No rule says when they end. (2) **A whole restore from the second drive** moves the old files beside the restored ones (`internal/backup/tier2_whole.go:52`, `:132`, `:345`); nothing removes them later and no page lists them. (3) **Exported app files** stay in the export folder of the drive; the pages can start, list and read an export, none deletes one, and nothing ages it (`internal/web/handler_export.go:30-87`). **Members (answers marked N):** EC02-4, EC23-4, EC18-4, EC27-4. | OPEN — **2026-11-14 (night build):** the design for part (1) is written, not built: `07-backup-architecture.md` §8.2f (the record, the date shown before the press, the take-back button, the deletion inside the weekly window after the hub agrees, 14 edge cases, the texts in Hungarian and English as a proposal). One open decision is in it: do the ended apps' snapshots count toward the window's weekly cap (CC's pick: no). Parts (2) and (3) are not touched. OPEN — filed 2026-11-10 (endings list) **2026-11-10 15:19 — RULED (D72 = yes, by the household's own choice; `09` §3 decision 292).** Not built. | — | Controller (D72): when the household pressed „delete backup data” for an app, its off-site copies end 30 days later, at the next clean-up after that day (the number of a restore copy, D62); when the household kept the backups, the off-site copies stay. The page shows the end date before anything is deleted — write that into `07-backup-architecture.md` before the build. Parts (2) and (3): a delete on the export page; a line on the kept-data page for the files a whole restore set aside. | CC |
| **R-1385** | Backup & restore | P3 | **Two timed jobs can meet with no lock and no order (12 answers).** Found 2026-11-11 (the Clock list, class W3, `documentation/audits/clock-list-2026-11-11/TABLE.md`). MEASURED on the night of 2026-11-11: on Tester 1 and the N100 the whole-guest backup started at about 05:30 local, the minute of the off-site proof restore (CC09-5); on demo-hp one whole-guest backup started at 02:30:38, with the dump leg; on the N100 the whole-guest backup starts about 5 minutes later each day. From source: the controller's auto-update default is the minute the whole-guest gate opens (CC11-5); the second-drive copy and the abandon sweep wait only for the clock (CC02-5, CC19-5, CC21-5); the PBS verify can run beside a backup (CA13-5, CA04-5); the timers on DooPlex and ep0 are not ordered against the hub's copies (CH05-8, CH24-5, CH25-3, CH26-5, CH26-6; from documents only, not measured). No harm was seen; the durations are from small test stores. | OPEN — filed 2026-11-11 (clock list) | — | Measure one night with a large store on Tier 0 (as a task), then decide per pair: a shared flag, a fixed offset, or „harmless, written down”. | CC |
## Storage & devices — 15 rows (P3 6, P4 9)
@@ -283,7 +283,7 @@ stopping line that lies.
| **R-1367** | Box system & updates | P3 | **The host journal fills a small system disk: 61% of its lines are `pam_unix(sudo:session)` open/close pairs from the agent's sudo calls.** MEASURED 2026-11-09, one hour, same on demo-hp and the N100: ~14,000 journal lines, ~8,500 of them sudo session lines; the top commands are `smartctl -a` (1,074 per hour — one every ~3 s), `lxc-info -n` (662), `mkdir -p` (540), `lvs` (537), `pct exec` (~520). The journal grows by journald's default cap (10% of the filesystem, max 4 GB): the N100 sits at 4 GB, demo-hp was 3.8 GB on 2026-11-07 and reached 0.7 GB again two days after a vacuum to 500 MB. On a 39 GB root that space is what the whole-guest local copy needs (R-548). | OPEN — filed 2026-11-09 **2026-11-10 (space list, `documentation/audits/space-list-2026-11-10/TABLE.md`):** place H7: the N100 (4.28 GB, 74 files) and Tester 1 (4.29 GB, 81 files) sit at the journal's own cap; Tester 1 writes about 150 MB a day. Lines written by `sudo` are 92 / 92 / 90 % of all lines in one hour — this counts three lines for each call (the command and the session pair), while this row's 61 % counts the session pair only: the same fact, counted two ways. No rank change is proposed: the journal stops at its own cap (about 4.3 GB), so it costs space and log history, not safety. | — | Agent, through the signed config bundle: `Defaults:felhom-agent !pam_session` (and `!syslog` only if the command log has no reader — check the audit trail's needs first) in the sudoers file; a journald cap drop-in (e.g. `SystemMaxUse=500M`) on the host; and read why `smartctl -a` runs every ~3 s (the SMART poll's interval or a loop). Red-proof each in the agent's tests; measure the line rate after on demo-hp. Evidence: `/mnt/5_hdd/felhom.eu/evidence/demo-hp-images-2026-11-09/journal-noise.txt` (off-git) | CC |
| **R-1368** | Box system & updates | P3 | **Five places on the host's system disk only grow: nothing deletes them and no rule says how many to keep.** MEASURED 2026-11-10 on demo-hp, demo-felhom and Tester 1 (the space list, `documentation/audits/space-list-2026-11-10/TABLE.md` places H5, H6, H9, H12, H13). (1) **Old kernels** (`/lib/modules`, `/boot`): 8 / 8 / 3 kernels = 8.15 / 8.15 / 3.05 GB. The ring-0 boxes got 7 kernels in 37 days and no kernel was ever removed: the update wrapper refuses a plan that removes a package (`11-os-updates.md` rule R4), and nothing runs `apt autoremove` (searched: architecture folder, both registers, runbooks, three source trees — 0 hits). The N100's system disk grew 4.6 GiB in 26 days, about a kernel a week. (2) **The agent's self-update folder** `/var/lib/felhom-agent/selfupdate`: 49 / 48 / 33 binaries = 740 / 724 / 509 MB, one for each agent version the box ever got; no document names the folder. (3) **Guest template archives** `vzdump-lxc-9100-*` in `/var/lib/vz/dump`: 1 / 4 / 1 = 0.65 / 2.56 / 0.65 GB. (4) **Old agent binary copies** `/usr/local/bin/felhom-agent.bak-*`: 1 / 11 / 0 files, up to 155 MB. (5) **Logs of failed whole-guest backup runs** in the dump folder (small; 9 on demo-hp). **Members (answers marked N):** H5-3, H5-4, H6-3, H9-3, H9-4, H9-8, H12-3, H12-4, H13-3, H13-4. | NARROWED — **2026-11-14 16:06 — RULED (D82, `09` §3 decision 307): a household's box may remove its old kernels, after two REAL night removals on each ring-0 box (the N100 and demo-hp); quiet nights do not count. Count so far: the N100 0, demo-hp 0 (the Tester 1 removal of 2026-11-14 was started by hand and is not a night). It needs a later agent release to switch it on.** Earlier: NARROWED — **2026-11-14 (release 2a): the old-kernel part is BUILT — agent 0.174.0, vouched 15:21 CET (`11` §5.11.1 `[FACT — built]`).** A ring-0 box removes its old kernels at night by the wrapper's own rule. On boxes: a real removal on Tester 1 (3 → 2 kernels, 1.12 GB, 8.5 s; the default, the pinned file, the generators and the state file unchanged; a second run removed nothing; one clean restart) — started by hand with a ring-0 mark set for the test on the operator's yes, because a night cannot be started by day (R-1394); dry runs on demo-hp (nothing to remove; refusals P1, P2, P3) and the N100 (6 of 8 will go). **LEFT for this part: the step by the agent at night — read the N100's and demo-hp's night of 2026-11-15; then a ruling for ring-1 boxes (D82).** Evidence `audits/release-2a-2026-11-14/`. The other four places of this row are unchanged. Earlier: OPEN — filed 2026-11-10 (space list) **2026-11-10 12:22 — RULED (D67 = yes, as a separate step; `09` §3 decision 286):** the box may remove old kernels by itself; rule R4 stays as it is for update plans; the clean-up is its own step with its own narrow rule (old kernel packages only, only after a good boot on the newest kernel). **2026-11-10:** the endings list leans on this row (`documentation/audits/endings-list-2026-11-10/TABLE.md` EA01-4, EA23-4, EA24-4, EA26-4). **2026-11-10 — the kernel part's fix shape is PROVEN by hand on demo-hp** (`documentation/audits/kernel-cleanup-spike-2026-11-10/`; design `11-os-updates.md` §5.11.1, not built): the rule removed 6 of 8 kernels and nothing else, 6.4 GB gained, boot settings unchanged, one orderly reboot back on the same kernel, a second run removed nothing. Not proven: a box whose running kernel is not its newest; a boot of the kept old kernel. **2026-11-10 — part (2) FIXED IN SOURCE, NOT RELEASED** (felhom-agent `776025c8`): after an update that ended well the agent removes every staged binary but the running version's; nothing is removed while an update is pending. `TestPrune_*` (6 tests, one runs the wrapper's real commit and rollback verbs), red seen three times. A box cleans its folder at the first good update after the release that carries it. Part (4), the `.bak-*` copies, is not reached (root-owned, made by a person). Parts (1), (3), (5) stay open. **2026-11-10 14:46 UTC — the kept old kernel boots:** on demo-hp the one-shot flag booted 7.0.14-26-pve to a healthy host in 313 s, and an orderly restart brought 7.0.14-28-pve back in 324 s; the pinned default, `grub.cfg` and the step's state file did not change, and the hub got no kernel report and no alarm (`documentation/audits/kernel-cleanup-spike-2026-11-10/evidence/fallback-boot/`). Still open for the kernel part: a second box; the step built as a program. **2026-11-10 (release run) — part (2) RELEASED, VOUCHED and PROVEN on three test boxes** (agent 0.173.0, vouched by the operator 15:29 UTC; `documentation/audits/release-space-fixes-2026-11-10/agent/`): the folder went 49 → 1, 48 → 1 and 33 → 1 files (about 700 / 690 / 485 MB back) at the commit of THIS update, not at the next one (the earlier sentence here and in the agent's CHANGELOG was wrong, and is corrected there). A rollback after the clean-up worked on Line truncated
| **R-1369** | Box system & updates | P3 | **When a disk is full there is no written way out: no runbook and no self-repair, for the host system disk, the thin pool, the guest disks or a data drive.** Found 2026-11-10 (the space list, `documentation/audits/space-list-2026-11-10/TABLE.md`, question 7). One missing runbook, seen from 21 places; the search of the runbooks folder for a disk-full procedure gave 0 hits (`evidence/searches.txt` S6). The alarms exist (hub: 90 / 95 % host disk, 85 / 90 % thin pool; controller: 85 / 95 % on the data volume and each drive) but they end in a mail, not in a step. For three small places nobody wrote what fails when they are full (EFI partition, host and guest temp folders). **Members (answers marked N):** H1-7, H2-7, H5-7, H9-7, H10-5, H10-7, H12-7, H13-7, H15-5, G1-7, G2-7, G7-7, G12-7, G13-7, G14-7, G17-5, D1-7, D2-7, D3-7, D6-7, D7-7, D8-7, D9-7, D10-7. | OPEN — filed 2026-11-10 (space list) **2026-11-10 (afternoon) — the runbook is written:** `documentation/runbooks/disk-full.md`, five parts, each hand step marked PROVEN (date and evidence; all on demo-hp) or NOT PROVEN. Still open: one rehearsal of the whole runbook on demo-hp; the NOT PROVEN steps (copies on a former target, template archives, the `.bak-*` binaries, the guest root disk); what fails when the EFI partition or a temp folder is full. | — | Docs: one runbook „a disk is full” with one section per disk (host system disk, thin pool, guest root, data volume, data drive): what to look at first, what is safe to delete (the places of R-1368 and R-1367), what is never deleted by hand (household data, the kept-data rulings). Rehearse it once on demo-hp. | CC |
| **R-1370** | Box system & updates | P3 | **The host's system disk has no space budget: no document says how much the host's own files may take, and the installer's sizing rule covers the guest only.** Found 2026-11-10 (the space list, `documentation/audits/space-list-2026-11-10/TABLE.md` place H1, answer H1-3; search S5 — 0 hits). Today: demo-hp 58 % of 41 GB, N100 38 % of 101 GB, Tester 1 34 % of 63 GB — and the parts that grow are the journal (R-1367), old kernels and agent binaries (R-1368), old backup copies (R-1087) and the local whole-guest copy (R-548). Second half, known but with no row until now: on a small disk the guest's own guards read the guest's 70 GiB while the thin pool behind it holds 54 GiB (demo-hp: 102 GiB promised on a 53.9 GiB pool, 70.67 % used), so only the hub's thin-pool alarm stands before a full pool. **Member:** H1-3. | OPEN — filed 2026-11-10 (space list) **2026-11-10 12:22 — RULED (D69 = yes; `09` §3 decision 288):** the section goes inside `16-storage-physical-identity.md`, no new document. **2026-11-10 (afternoon) — the section is written:** `16-storage-physical-identity.md` §14 (45 places, the list „no limit, by ruling”, the budget, the open design points); `PROMPT-TEMPLATE.md` §N.5 keeps it true. The budget from measured numbers: the root needs about 14 GB when the whole-box copies go to a backup drive, and 28–54 GB at keep 3 when they do not. **D75 (for the operator):** the smallest system disk Felhom accepts. CC's pick: 128 GB only with a backup drive, 200 GB without one. The other option, 128 GB always, lets a box with many apps and no backup drive stop its local copies. With no ruling the installer accepts any disk. Open after the ruling: the installer's refusal. **2026-11-10 15:19 — RULED (D75 = one number, 200 GB, with or without a backup drive; `09` §3 decision 295).** This replaces CC's two-step pick. **Open, not ruled:** what the installer does below 200 GB — refuse, or warn. demo-hp (128 GB) stays as a test box. `16` §14.3 carries the ruling. | — | Design first: what the installer does below 200 GB (refuse, or warn) — a later session. Then the installer. | CC |
| **R-1370** | Box system & updates | P3 | **The host's system disk has no space budget: no document says how much the host's own files may take, and the installer's sizing rule covers the guest only.** Found 2026-11-10 (the space list, `documentation/audits/space-list-2026-11-10/TABLE.md` place H1, answer H1-3; search S5 — 0 hits). Today: demo-hp 58 % of 41 GB, N100 38 % of 101 GB, Tester 1 34 % of 63 GB — and the parts that grow are the journal (R-1367), old kernels and agent binaries (R-1368), old backup copies (R-1087) and the local whole-guest copy (R-548). Second half, known but with no row until now: on a small disk the guest's own guards read the guest's 70 GiB while the thin pool behind it holds 54 GiB (demo-hp: 102 GiB promised on a 53.9 GiB pool, 70.67 % used), so only the hub's thin-pool alarm stands before a full pool. **Member:** H1-3. | OPEN — **2026-11-14 (night build):** the open point of D75 has a written design with a pick, not built: `16-storage-physical-identity.md` §14.5 — refuse below 200 GB on an appliance install, with one named flag for the operator; warn only on a BYO host; 8 edge cases. It waits for the operator's yes. OPEN — filed 2026-11-10 (space list) **2026-11-10 12:22 — RULED (D69 = yes; `09` §3 decision 288):** the section goes inside `16-storage-physical-identity.md`, no new document. **2026-11-10 (afternoon) — the section is written:** `16-storage-physical-identity.md` §14 (45 places, the list „no limit, by ruling”, the budget, the open design points); `PROMPT-TEMPLATE.md` §N.5 keeps it true. The budget from measured numbers: the root needs about 14 GB when the whole-box copies go to a backup drive, and 28–54 GB at keep 3 when they do not. **D75 (for the operator):** the smallest system disk Felhom accepts. CC's pick: 128 GB only with a backup drive, 200 GB without one. The other option, 128 GB always, lets a box with many apps and no backup drive stop its local copies. With no ruling the installer accepts any disk. Open after the ruling: the installer's refusal. **2026-11-10 15:19 — RULED (D75 = one number, 200 GB, with or without a backup drive; `09` §3 decision 295).** This replaces CC's two-step pick. **Open, not ruled:** what the installer does below 200 GB — refuse, or warn. demo-hp (128 GB) stays as a test box. `16` §14.3 carries the ruling. | — | Design first: what the installer does below 200 GB (refuse, or warn) — a later session. Then the installer. | CC |
| **R-984** | Box system & updates | P4 | **Host and agent hardening, low risk each: the backup-target route has no role check; the unsigned guest spec has no size bounds; signed ops have no maximum validity; the guarded mkfs does not refuse LVs, zvols or RAID devices; the local API is reachable from the LAN by a static route; a sudoers inspect pattern is wider than needed; install passes secrets on the command line; guest settings files are world-readable; the host offers rpcbind and DNS on the LAN.** Fix each in its own small change. Security review 2026-10-12 (`audits/security-review-2026-10-12/`); detail held off-git for the operator (`/mnt/5_hdd/felhom.eu/evidence/security-review-2026-10-12-PRIVATE/`). | OPEN | — | Batch into hardening work | CC |
| **R-979** | Box system & updates | P3 | **Controller updates are trusted by registry tag alone (no digest, no signature), and the Update button offers any newer tag, even an unvouched one; the built-in traefik, cloudflared and FileBrowser images are also pinned by tag only.** Whoever can push to the registry decides what runs. Fix: pin by digest from the hub's vouched list; offer only vouched versions. Security review 2026-10-12 (`audits/security-review-2026-10-12/`); detail held off-git for the operator (`/mnt/5_hdd/felhom.eu/evidence/security-review-2026-10-12-PRIVATE/`). | OPEN — design-proposal | — | Design note, then build | CC |
| **R-974** | Box system & updates | P3 | **A compromised host agent can likely become root on the host through the root wrapper that applies PBS settings: the wrapper accepts a folder path inside the agent's own writable tree and then acts on it as root.** Static (the symlink behaviour was reasoned, not run). Same class as R-861. Fix: a fixed root-owned folder, and refuse symlinks before any root action. Security review 2026-10-12 (`audits/security-review-2026-10-12/`); detail held off-git for the operator (`/mnt/5_hdd/felhom.eu/evidence/security-review-2026-10-12-PRIVATE/`). | OPEN | — | Fix the wrapper with a test on a throwaway box | CC |