A new box with an older hub sent `ended` in the window request; that hub reads
4 KiB of it, decided the window on a wrong count, and the weekly clean-up
stalled.
- The box sends `ended` only when the hub's last window answer said
accepts_ended (kept in settings, re-read from every answer, so a rolled-back
hub stops getting them after one answer). Otherwise the request is the one
every hub always got, and the ordinary clean-up runs.
- One request carries at most 5000 snapshot ids.
Tests seen red before the fix; 6 mutations seen red (felhom.eu
documentation/audits/build-release-4-2026-11-15/A/redproof-review-5.txt).
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
Built, not released; no box has run it. felhom.eu `07-backup-architecture.md` §8.2f.
It does nothing without a hub that knows it: an older hub agrees to no id.
- The record: settings `offsite_ends` {app, removed_at, ends_at}. One writer
(backup.StartOffsiteEnd), called by a removal only with „delete backup data”
and by „Törlés 30 nap múlva” on the restore page. A second press changes
nothing. No record, no deletion. Only on the hub-provisioned tier.
- The date is on the remove dialog before the press; the restore page shows
„Törlésre kijelölve …” with „Mégis megtartom”, and one history line per
carried-out end. An install of the same app deletes the record.
- The weekly window: after the guard has passed the policy's plan under the cap
as before, it adds the snapshots whose app tags are ALL due records, that the
hub agreed to and the box named in the same request; the result names them
apart (removed_for_ended_apps). `_shares`, the box's own stacks and an
installed app are never due.
- The report carries the pending records; the household gets one mail per
press (offsite_end_scheduled), in both languages.
MinAgent unchanged (0.158.1). 143 parity fixtures gained the dialog's one new
script line (patched, not regenerated); one new parity case.
Tests TestD72_*, TestD84_*; 56 mutations seen red
(felhom.eu documentation/audits/build-release-4-2026-11-15/A/redproof-ctrl-*.txt).
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
R-218's declaration half shipped in v0.201.0 and works. Its consume half never
existed. Reconcile ran exactly twice per process — at start-up and when the
recovery screen drives it — and BOTH fire before the hub has anything staged,
because the hub stages in RESPONSE to the declaration those runs precede.
Measured on the R-201 re-walk: unlock reconcile 11:43:07, hub staged 11:44:57
saying 'next cycle', a full report cycle ran 11:55:46, still unconsumed at
12:06. A guest command line applied it in 18 seconds — everything correct except
the trigger.
Bridge.RetryIfDeclared re-runs the SAME reconcile on a 5-minute tick, driven from
the box's own published declaration (OffboxReportStatus().State) — the very
statement the hub acts on, so the two cannot disagree.
Poll, not an ACK flag, decided on the promise: the no-target message says 'amint
megvannak' (no deadline) and the card says 'within a day'. Five minutes is inside
both by a wide margin and needs no hub change.
It stops by construction — a healthy box does no work and logs nothing — and the
settle gate is deliberately kept via ReconcileWhenSettled.
The marker was investigated and left alone: applied_marker lives in the guest's
DataDir, which a rebuild destroys, so it cannot suppress a legitimate re-run.
R-220's customer half: the refusal no longer tells the customer to choose from a
list that may be empty. It names the rebuild, points at the Meghajtók page, and
promises no outcome.
Red-proofs: remove the retry -> credential uncollected (the dead end reproduced);
drop the stop condition -> a healthy box hammers the hub; call Reconcile instead
of ReconcileWhenSettled -> settle gate bypassed; restore the old sentence -> the
impossible action returns.
28 packages ok, vet clean, all controller gates OK.
The day-0 race (DIAG-f10): a fresh box boots below the operator floor, the
apply-bridge consumes the single-use offsite password, then ~35s later the
managed auto-floor update replaces the container mid-install -> the new process
finds no installed key -> consume -> 404 -> offsite dead until an operator
Re-issue. Recurs on every onboarding whose ISO floor lags the managed floor.
Ordering-only fix (consume/install/persist internals + the 404-no-oracle
contract + the Consumer UNTOUCHED; R-71(b) rejected-by-design):
- New seam offsiteapply.SettleProvider.SettleState() + SettleFunc adapter over
the self-updater's own GetFloor()/IsUpdateRunning() (no second floor path).
- Bridge.AwaitSettle polls 10s BEFORE the 3-min Reconcile ctx: defers while an
update runs or the box is below the known floor; GOes at/above floor on the
first poll with zero added latency (B'). Bounds 90s floor sub-bound / 5min
overall, both GO+WARN (hub that can't serve a floor can't serve a consume ->
no burn risk; R-71c is the belt). ReconcileWhenSettled = gate then reconcile.
- main.go: bridge goroutine moved after the updater is built; wired only when an
updater exists (nil Settle = reconcile immediately, old behavior).
Finding: the floor is in-memory (report-ACK ~5-10s), NOT persisted -> unknown on
any restart until the first ACK; the 90s sub-bound is sized to that.
Tests (injectable clock, fake SettleState, recorded Consumer): A-E + nil-provider
+ cancelled-gate. Four red-proofs all observed FAIL then restored: gate removed /
updateRunning branch / floor sub-bound / overall bound. Deferral paths ship
unit-proven + red-proofed, NOT live-fired -- their precondition is now
structurally prevented by the v1.25.0 build gate. Layering: gate prevents, (a)
defers, (c) heals.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N7Drmtm2RzoqbkJZCNSFNQ
A hub-side quota raise now reaches the target: the marker hash changes and
the bridge re-applies via key-auth-first (no password consumed). Test:
quota-only change remaps the new quota with a panicking consumer.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PSK5g6qYLknKj8u3QAFEr6
QuotaGB rides the descriptor into OffboxTarget; RepoSizeBytes persisted
from restic stats. Pre-run gate: >=100% refuses NEW backups (Hungarian
notice + operator alert) but prune STILL runs (red-proofed) and restore is
never gated; >=80% warns. /backups usage bar (quota>0 only). The hub
report gains the non-secret offsite status object for the OffsiteChecker.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PSK5g6qYLknKj8u3QAFEr6
Key-auth-first: a KeyAuthProber seam lets the bridge skip consume+install
when the already-installed key still authenticates (pinned to the freshly
verified host key) — descriptor changes on provisioned guests no longer
loop on consume-404. Fingerprint verify still precedes everything.
Wipe-on-escrowed: confirm-escrow now calls the agent's new
DELETE /escrow/stage-secret (v0.78.0) best-effort, closing the hygiene gap
where a ceremony-less confirm left the staged password file behind.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PSK5g6qYLknKj8u3QAFEr6
Live finding: SFTP-mode ssh-copy-id mktemp's under ~/.ssh; the container
image has no /root/.ssh, so the install died locally AFTER the one-time
password was consumed (fail-safe held; password never transmitted).
Probe confirmed the pinned known_hosts + StrictHostKeyChecking=yes chain
is sound against the real box once ~/.ssh exists.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PSK5g6qYLknKj8u3QAFEr6
The SSHCopyIDInstaller used StrictHostKeyChecking=accept-new on the ssh-copy-id
and sftp-verify connections, so even though the bridge verifies the box host-key
fingerprint against the hub descriptor BEFORE installing, the actual install
connection was not pinned to that verified key — a MITM could substitute a
different key in the gap between the scan and the install (TOCTOU).
Now the bridge threads the scanner-verified known_hosts line into KeyInstaller,
which writes it to a temp known_hosts and connects with StrictHostKeyChecking=yes
+ UserKnownHostsFile — the install/verify sessions refuse any key but the one the
bridge already matched. Empty known_hosts now refuses to install.
Test asserts the installer receives the pinned known_hosts; red-proofed by passing
an empty line (the pre-fix TOFU shape) → test fails. Addresses the security-review
"host-key TOFU after verify" finding on internal/offsiteapply/seams.go.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PSK5g6qYLknKj8u3QAFEr6
On startup reconcile the hub-served offsite: descriptor into a key-only offbox
target. internal/offsiteapply.Bridge: verify-pin box host key vs host_fingerprint
(NO blind TOFU) → consume the one-time password (single-use, never logged) →
sshpass ssh-copy-id -s -f install + verify → configure offbox → EscrowState=pending
(fork-4 via Manager.ApplyOffsiteTarget) → persist a descriptor-hash marker LAST.
Idempotent + fail-safe. Seams faked in tests; both red-proofs run+reverted.
Dockerfile + sshpass. NOT yet live-applied (supervised end-to-end next runbook).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PSK5g6qYLknKj8u3QAFEr6