Night 2026-10-06: morning note, STATUS, report

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
2026-10-07 05:46:57 +02:00
parent 877676a64f
commit 3aba200293
3 changed files with 119 additions and 2 deletions
@@ -0,0 +1,71 @@
+# Morning note — the second burn-down night, 2026-10-07
+
+**Day session: start with Parts B and D of `TASK-night-readback-offsite-gap-tester1-2026-10-07.md` (after 08:30), then
+the releases below.** Nothing was delivered to any box tonight, so the night read-back sees a clean night.
+
+## 1. Is anything broken right now?
+
+**No.** The hub answers (health 200, System page 200). demo-hp, demo-felhom and Tester 1 are online. Tester 2 is off,
+as usual. CI is green on the last commit of every repository. No CI job was lost tonight (R-887: 0).
+
+**You do not need to do anything this morning.**
+
+## 2. One decision I took myself (you may reverse it)
+
+The app catalog's storage check now only **notes** an empty folder; it does not fail the app for it. A household
+folder (photos, documents) is empty on every fresh install, so failing it would block four apps for ever.
+
+## 3. The four numbers
+
+Rows before: **138**. Rows after: **138**. Opened: **2**. Closed: **2**.
+
+## 4. What was fixed
+
+- **Backups:** after a restart, the host program now remembers the last good backup. An unreachable backup store no
+  longer looks like "never backed up", so no copy is asked for that cannot be made.
+- **Backups:** a reinstall no longer overwrites the old backup key in your safe store. The old copies stay readable.
+- **Hub:** two quick clicks on "Save" for off-site backup no longer create two storage accounts.
+- **Box:** the "format a drive" list no longer offers a drive the household already uses.
+- **Box:** 22 more messages follow the household's language, in the friendly form.
+- **Disk health:** three more warning counters now travel from the disk to the hub. Nothing alarms on them yet.
+- **Apps (held, not live):** Jellyfin and Emby no longer let a stranger on the internet sign in as if at home
+  (measured, then fixed). Also wishlist, homepage, gramps-web and two test tools.
+- **Measured:** wger runs best with its proper web server and 2 workers (41 % of its memory).
+
+**Waits for today's releases:**
+- agent: the restart memory and the disk counters, released together with the memory-kill check as v0.150.0;
+- controller: the drive list, the 22 messages and the disk counters;
+- hub: v0.141.0 is ready in the code, but not built (see 5);
+- catalog: six changes on the branch `night-held-2026-10-06`; merge them to the main line after the read-back.
+
+## 5. What failed, and why
+
+- **The Tester 1 box: still no entry.** Your key copy at 20:20 used the key of your own login, not this server's
+  key. Fix: run `ssh-copy-id -i ~/.ssh/id_ed25519.pub root@192.168.0.154` from this server as kisfenyo.
+- **The hub release did not run.** The safety check of my session refused the image build at 00:12. I stopped
+  there. The hub still runs 0.140.0, unchanged.
+- **I could not read the hub's list of apps per box.** The same safety check refused it. So every catalog change
+  went to the held branch.
+- **Two rule slips by helpers, both harmless:** one cleaned unused storage on the test machine by hand; one ran a
+  catalog test that starts small test containers on this server. Nothing was left running.
+
+## 6. The designs — each needs one answer from you
+
+1. **Empty disaster-recovery records:** two fields were never built and nothing reads them. Pick: remove them and
+   fix the documents.
+2. **Host updates:** Proxmox's own packages (78 behind) and the kernel never update. Pick: a new night lane for the
+   Proxmox packages, approved per set; the kernel stays a manual step for now.
+3. **Host program rights:** one left-over right is wider than the documents said: it can start any container image.
+   Pick: close it before the first paying customer; accept the other two.
+4. **RESET leaves old off-site data behind.** Pick: delete it through the customer's own login before the login is
+   removed.
+5. **Fake snapshots can steer the off-site clean-up.** Pick: accept and close; build the hub's own snapshot count
+   later (new item).
+6. **Old backups after a reinstall:** the key part is fixed. Pick: also tell you when old copies use another key.
+
+## 7. Two decisions for you
+
+1. **Release today in this order: hub 0.141.0, agent 0.150.0, controller, then merge the catalog branch.**
+   My pick: yes, after Parts B and D. If you do nothing: the fixes stay in the code and reach no box.
+2. **May my session build hub images at night?** The safety check blocked it. Pick: allow it for the brief's
+   window only. If you do nothing: night hub releases cannot happen.