demo-hp local tier refused again: space freed; decision 285 + 56 amended (old infra/rc images); R-1366, R-1367 filed; R-548 note

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS

REPORT.md

@@ -1,24 +1,48 @@
-# REPORT — morning check, and hub 0.179.1 (2026-11-08, attended)
+# REPORT — demo-hp's full local tier, and what the whole-box copy holds (2026-11-09, attended)
 
-**Plain words first.** Last night's fixes worked. The N100 installed its Proxmox updates (no refusal) and a new kernel,
-and restarted itself. demo-hp made its first local whole-box copy in 10 days. The hub still said the N100 „needs a
-restart" after it had restarted; hub 0.179.1 fixes that. demo-hp does need a restart; the operator chose to wait for its
-kernel step.
+**Plain words first.** demo-hp did not restart on 11-08/09. Its local whole-box copy was refused for space again
+(12.5 GiB free, 13.3 GiB needed), so the night's OS run and its kernel step waited. With the operator's yes, CC freed
+space and filed the causes. The controller now deletes two kinds of old images by itself (v0.351.0, in code, not
+released).
 
-**Box facts (SSH).** demo-felhom: pve step `DONE upgraded=33` (05:32), kernel 7.0.14-28 one-shot boot 05:35, `KERNEL GOOD`
-(default now 7.0.14-28), 0 host processes on replaced libraries, guest 9201 running. demo-hp: local backup `completed`
-02:43 (`local:backup/vzdump-lxc-9201-2026_11_08-02_30_38`), pve step `upgraded=35`, no kernel step yet (7.0.14-28 pending),
-systemd/sshd/kvm/zed/rrdcached on replaced libraries.
+## What the whole-box copy holds (demo-hp guest 9201, `docker system df -v`, measured)
 
-**Hub 0.179.1** (`osupdates.layerLine`): the host's last boot (crash guard `last_boot_at`) cuts the host layer's
-reboot-needed run, less one minute of clock skew; a claim after the boot starts a new run. Feeds the System page and
-the 14-day alarm. Test `TestRebootNeeded_ClearedByAHostBootAfterTheClaim`, red on two mutants; `go test ./...` green.
-Deployed: image built from `ddbcabeb`, manifest `6b91a215`, ArgoCD sync of `Deployment/hub` only, rollout ok, pod
-logs `felhom-hub 0.179.1 starting`. **Live (endpoint method — `GET /system` with CC's token, the page the operator
-opens):** only demo-hp is listed. Evidence: `documentation/audits/hub-reboot-flag-2026-11-08/`.
+| Part | Size |
+|---|---|
+| Images used by containers | 13.2 GB |
+| The apps' undo images (decision 53, on purpose) | 9.8 GB |
+| Old controller images (25 of them test builds) | 2.2 GB |
+| Old infrastructure and helper images | 1.35 GB |
+| Docker volumes (app data) | ~1.5 GB |
+| `sys_drive` | 1.3 GB |
 
-**Remote restart buttons (the operator's question).** The host page has „Restart controller" and „Reboot the guest"
-per guest (agent actions, D65). There is no button that reboots the Proxmox host; the host reboots only by the kernel
-step. Not a defect — the agent's action list is closed by design.
+Packed: 9.8 GB, on a 39 GB system disk. The N100 writes its copy to its 1 TB backup drive (`felhom-backup`), 1.5 GB.
+Tester 1 and demo-hp write to the system disk (`local`). The same old-image pattern was on all three boxes (43–48% of
+image space unused; most of it is undo images).
 
-**Registers.** No row opened or closed (the hub fault fixed in the session, size rule). System poster: no fact changed.
+## What was done on demo-hp (operator's yes)
+
+- Inside guest 9201: 36 unused images deleted by name (25 old controller test builds; traefik v3.6.7/v3.7.13;
+  cloudflared 2026.6.0/2026.9.3; Samba 1.1.0; FileBrowser 1.3.3; alpine:3.20, curl, hello-world). The undo images,
+  the previous controller (0.349.0) and the previous pins were kept. Images 27.56 → 24.15 GB.
+- On the host: installer ISO 1.30.0 removed from `local` (`pvesm free`); its sha256 `590d7b9b…` matched the copy on
+  DooPlex and the file on iso.felhom.eu (HTTP 200). The old golden archive 0.236.0 stays (no other copy found).
+- `local`: 12.5 → 14.0 GiB free, above the 13.3 GiB the space check asks.
+- Evidence off-git: `/mnt/5_hdd/felhom.eu/evidence/demo-hp-images-2026-11-09/` (before, cleanup, journal noise).
+
+**Not yet observed:** tonight's local copy, OS run and kernel step on demo-hp. Positive observable: a new
+`vzdump-lxc-9201-2026_11_10-*.tar.zst` in `/var/lib/vz/dump/`, then the host's boot time after the kernel step.
+
+## Product changes
+
+- Controller v0.351.0 (felhom-controller CHANGELOG): decision 56 amended (old `-rc` controller images go) and decision
+  285 (pinned infrastructure images keep the pin and the previous pin). Six mutants red-proved.
+- `09-update-architecture.md` §3: decision 285 added, decision 56 amended.
+
+## Register
+
+Rows before 223, after 225; opened 2 (R-1366 the whole-box copy carries every image — operator decision; R-1367 the
+host journal's sudo session noise), closed 0. R-548 extended with the 11-09 refusal and the clean-up.
+
+Architecture read: `07-backup-architecture.md` (tiers), `09-update-architecture.md` §3 (decisions 53, 56).
+System poster: no fact changed (no backup tier, time or retention moved).