4.3 KiB
2026.08.29.02 — deploy report
7 of 7 target hosts deployed. officer-dev skipped by design. Every host prints
plugin plugin_config — the two new command_hq_sessions columns exist on all eight databases.
All eight checkouts were at 58f50a52 (the 2026.08.29.01 tag) before this rollout, and all are at
78bf63c2 (deploy/2026.08.29.02) now. No host was ahead, none diverged, none was dirty.
This was the first release under a new watcher loop. The rollout itself behaved as written.
Per host
| Host | HEAD before → after | plugins:update | backup | db:push | verification query | officer | health |
|---|---|---|---|---|---|---|---|
officer-dev |
— SKIPPED | — | — | — | plugin plugin_config |
not restarted | — |
alpha |
58f50a52 → 78bf63c2 |
15/15, exit 0 | 12,309,974 B | exit 0 | plugin plugin_config |
3 → 4 restarts, online | 401 (9000) |
fofos |
58f50a52 → 78bf63c2 |
5/5, exit 0 | 2,792,097 B | exit 0 | plugin plugin_config |
26 → 27 restarts, online | 401 (9000) |
macbookpro |
58f50a52 → 78bf63c2 |
no plugins installed | 5,179,143 B | exit 0 | plugin plugin_config |
3 → 4 restarts, online | 401 (9010) |
macbruno |
58f50a52 → 78bf63c2 |
1/1, exit 0 | 86,480 B | exit 0 | plugin plugin_config |
6 → 7 restarts, online | 401 (9000) |
production-pertento-bridge |
58f50a52 → 78bf63c2 |
1/1, exit 0 | 187,488 B, second attempt | exit 0 | plugin plugin_config |
4 → 5 restarts, online | 401 (9010) |
edge-pertento-bridge |
58f50a52 → 78bf63c2 |
2/2, exit 0 | 5,916,050 B | exit 0 | plugin plugin_config |
5 → 6 restarts, online | 401 (9000) |
deadpool-rezio-bridge |
58f50a52 → 78bf63c2 |
2/2, exit 0 | 92,979 B | exit 0 | plugin plugin_config |
22 → 23 restarts, online | 401 (9000) |
401 is the healthy answer from /api/health behind the session gate, as the plan says.
What the release was for, confirmed on the host that matters
fofos carries av-club, and the commander template it now ships is on disk:
plugins/av-club/commanders/convert-on-arrival/COMMANDER.md, plugin checkout at 5b2fe03
("Convert on arrival — the first plugin-shipped commander"). The plugin arrived through the standard
bun run plugins:update step, as the plan predicted — nothing in this plan installed it.
What did not go to plan
production-pertento-bridge failed its backup on the first attempt and the host stopped there.
The plan's pg_dump is Ubuntu's 16.15; that host's Postgres is 18.3, and pg_dump refuses a newer
server outright:
pg_dump: error: aborting because of server version mismatch
pg_dump: detail: server version: 18.3; pg_dump version: 16.15
The host was left correct but half-done: code at the tag, plugins updated, no db:push and no
restart, so the running process was still serving the old code against the old schema. Nothing was
broken by stopping there.
Postgres on that host runs in a docker container (postgres:18.3-alpine), so a version-matched client
was already present. The backup was retaken with docker exec postgres pg_dump — same dump, correct
tool, read-only — and the plan then continued unchanged from step 2. Result above.
This is a per-host detail worth writing down, because it will recur on every release whose plan
backs the database up: the plan's pg_dump line is host-portable only where the client is at least as
new as the server. production-pertento-bridge is the one host where it is not.
fofos health read 000 immediately after the restart and 401 fifteen seconds later. The check
was racing a cold start, not finding a fault. The health check now retries for up to thirty seconds
before it reports a number.
fofos has one untracked file, scripts/chk.ts. Noted, not touched — a fast-forward could not have
disturbed it.
Skipped
officer-dev — the rollout ran on it. Its checkout was already at 78bf63c2 and its database already
carries both columns (verified by the same query: plugin plugin_config). Its officer process was
not restarted, so it is running the previous build. The owner restarts it deliberately:
pm2 restart officer
docs/servers.md lists nothing under "Deliberately skipped" today, and no host was unreachable —
macbookpro was awake.