simpletest/who_need_help
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Record operator monitoring visibility

Browse Source
This commit is contained in:
SimpleTest 2026-08-21 02:04:15 +03:00
parent 01855f5d9a
commit a3d13bf95b
2 changed files with 28 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
docs/public-launch-checklist.md
View File

@ -207,7 +207,7 @@ as forward-only rather than receiving an invented database rollback.
with the run prefix. Anonymous contact verification still requires a with the run prefix. Anonymous contact verification still requires a
controlled production mail smoke before this combined item can be controlled production mail smoke before this combined item can be
checked. checked.
- [ ] Database, application, worker, email, push, backup, and edge monitoring - [x] Database, application, worker, email, push, backup, and edge monitoring
are visible to the responsible operator. are visible to the responsible operator.
Record observed timestamps, revision/image identities, and non-secret evidence Record observed timestamps, revision/image identities, and non-secret evidence

27
docs/verification.md
View File

@ -3372,3 +3372,30 @@ promoted.
Both public readiness endpoints returned the exact ready payload. The public Both public readiness endpoints returned the exact ready payload. The public
Git remote remained at `921e04b3608007675e22e7e26e0beb3975dbba58` and was Git remote remained at `921e04b3608007675e22e7e26e0beb3975dbba58` and was
not pushed. not pushed.
- The external monitor runs independently on `209.141.50.251`, while the
production application and edge resolve to `159.195.158.59`. Its user timer
was enabled and active on the minute schedule. At
`2026-08-20T23:02:22.000569Z` its persisted state reported aggregate,
readiness, authenticated metrics, and backup freshness as `up`; the metrics
node was `who_need_help@172.19.0.2` and the configured stale-backup threshold
was exactly 129600 seconds.
- A public read-only request to `https://whoneedhelp.com/healthz/ready` returned
HTTP/2 200, the exact `{"status":"ready"}` payload, and `Via: 1.1 Caddy`.
The readiness implementation checks the database with `SELECT 1` and, for
the observed combined application role, requires the supervised Oban worker
to be running. The authenticated metrics endpoint exposes bounded HTTP
exception, Oban failure, email delivery/exception, and push outcome counters
consumed by the external monitor; database timing and VM runtime metrics are
also exported for operator diagnosis.
- The production backup user timer was enabled and active on the workstation.
Its 2026-08-21 00:00 EEST invocation finished with result `success`; the next
invocation was scheduled for 2026-08-22 00:00 EEST. Backup monitoring uses
the restore-verified heartbeat rather than mere archive creation.
- In the already authenticated operator Gmail inbox, a read-only Playwright
inspection found the unread message `[Who Need Help] Operations monitoring
test`, delivered on 2026-08-20 at 20:04 EEST and labelled Inbox plus
Projects/WhoNeedHelp. The message was not opened or marked read. Together
with the live monitor state, this proves that the responsible operator can
receive the alert channel; the monitor's unit tests cover state-transition,
recovery, bounded metric parsing, and stale-backup alert behaviour without
inducing a production outage.