REPORT: the 2026-10-26 afternoon run. Docs only.
gates / gates (push) Successful in 1m48s

This commit is contained in:
2026-10-26 14:38:15 +01:00
parent b1bc63dd32
commit 0683cc2dab
+16 -21
View File
@@ -1,26 +1,21 @@
# REPORT — agent v0.166.0 released and vouched (2026-10-26, attended release run)
# REPORT — R-1285: the BYO SSH clean-up in the same update — committed, NOT released (2026-10-26 afternoon)
**Plain words first.** Agent 0.166.0 runs on Tester 1, demo-hp and demo-felhom, and the operator vouched it at 10:04:55
CET, so new installs get it (a fresh install today fetched it). A signed job's window is now at most 24 hours — seen
live. One part is not done: on a BYO host the SSH clean-up still comes one release late (R-1285, back to OPEN).
**Plain words first.** A BYO box coming from 0.164.0 kept our SSH file one release too long: in the normal update the
OLD agent installs the config bundle, so the new agent's clean-up pass never ran on it. Now the new agent runs that pass
once, right after its self-update commits. Not released today (operator: „not today"); its live proof needs a BYO box on
0.164.0, which rides the next agent release.
## Release
## The fix (`b1bc63d`, CI 2363 ✓)
- From a clean clone at `aa69114` (CI 2351 ✓): tag `v0.166.0`; binary sha256
`3f22c65b78914632fa476292f6b58e5af3ed9e9900f3a5a7a347a9c0bfea9c5c`; bundle
`093772c1bda8b974e86ed45472f8fadc8992f247563f5739876102a5156003a3`; verified by independent download.
- Delivered by signed `agent_update` + `agent_config_update`: each box COMPLETED, bundle INSTALLED, self-check 76/76,
update committed.
- D50 pin `day0_agent` → 0.166.0 (`b890b57`, CI 2356 ✓); the published gate green.
- `internal/selfupdate/commit.go`: `Manager.SetOnCommitted` — a hook run once after „update committed".
- `internal/osupdate/bundle.go`: `ConfigUpdateExecutor.PolicyPassAfterUpdate` (the wrapper's `bundle-policy` mode, host
layer, no envelope); `cmd/felhom-agent/main.go` `wireBundlePolicyAfterCommit` hooks it on.
- The smaller of the two fixes: agent-only — holding the bundle job until the commit would need new hub queue logic.
## Live checks
## Proof
- **R-1280 (closed):** a 25 h signed job on demo-hp → „REJECTED signed op — executor not called … op validity window too
long (25h0m0s)"; a 30 min job ran. `felhom-opsign -ttl 25h` refuses (rc 2); 24 h accepted.
- **R-1285 (back to OPEN):** in the normal two-job delivery the bundle job runs in the OLD 0.165.0 process (demo-hp: bundle
09:03:07–09:03:09 CET, restart into 0.166.0 at 09:03:10), so no policy pass ran on that bundle. Under 0.166.0 a later
bundle logged „BUNDLE POLICY DONE install_mode=appliance removed=0" — the pass is wired; its delivery order is not.
No BYO host is in the fleet. Fix options in the row.
- **R-1289 (hub):** the vouch logged „behind the fleet's 0.344.0" — a release version.
Evidence: felhom.eu `documentation/audits/release-monday-2026-10-26/agent-0166/`.
- `cmd/felhom-agent/r1285_policy_after_commit_test.go`, in the real order: a pending marker for the running version →
`commit`, then exactly one `bundle-policy` plan; no marker / a mismatched marker → nothing. Red-proved (the hook call
removed → commit only, no pass — the demo-hp symptom of 2026-10-26).
- **Not live:** only a box coming FROM 0.164.0 can show it (0.165.0's wrapper already applies D43 at plan time), which
needs a signed downgrade chain on a BYO box.