docs(index): OtOpcUa QA canary, #890 re-apply, live #794 check and #780 runaway drill; #782/#801/#794/#890 closed

Claude-Session: https://claude.ai/code/session_01GT67ACmpMR7b9LGRPLijMi
This commit is contained in:
Joseph Doherty
2026-11-12 19:11:11 -05:00
parent cde1297c4a
commit 610c40afa5
2 changed files with 30 additions and 2 deletions
@@ -556,3 +556,31 @@ above (#879–#887, #880, #884/#885, #876) is now deployed on vd03 and on QA.**
- the ServiceLevel order is correct again (250 Primary);
- the #887 Severity order flip applies to Galaxy alarms;
- script-backed values return `BadResourceUnavailable` while the runner is unavailable (ScadaBridge#112 handles it).
### 2026-11-12 (evening) — QA: #890 re-apply, script-runner canary, live #794 check, #780 runaway drill
Owner request: "add the canary tag and finish QA". WinRM only for on-box deploy steps; no service restart; pair stayed
vq03 **250** / vq04 **240**. Record: OtOpcUa `docs/deploy/qa-upgrade-b8b4c18f-to-fea7de0a.md`, last section (`a17ffb72`).
- **#890 re-applied** on both nodes (installer at `f75ee7d0`). `-DenyPaths` = install root, `pki`, `data` and
**`C:\ProgramData\ZB\kek`** (the `Secrets:MasterKey:FilePath` KEK lives outside the install root). Second apply 12/12
unchanged. Installer nit: `HostServiceEnv` reports `changed` when the env list was only reordered.
- **Canary on QA** (kept as a runner monitor), authored in the AdminUI and deployed (`bcaba714`, rev `82dd0598…`):
Calculation driver `script-runner-canary` → `Device1` → `canary-calc-seconds` (`ns=2;s=script-runner-canary/Device1/canary-calc-seconds`)
and VT `canary-vt-second` on equipment `EQ-ec3979550d90` (`ns=3;s=EQ-ec3979550d90/canary-vt-second`), 1 s timers.
Good and advancing on both nodes. QA had no ConfigDb before this.
- **QA AdminUI access gotcha:** the auth cookie is `Secure`, so a browser over `http://wonder-app-vq0x:9000` cannot keep
a session. It works through a local forwarder (`http://localhost:…`). QA also has **no LDAP group → role grant**, so the
box GLAuth logins authenticate with zero roles (AccessDenied). For this run a temporary `LdapGroupRoleMapping` row was
added and then deleted.
- **#794 step 5, live:** the worker token is restricted (only SeChangeNotify; restricting SIDs identity/Everyone/Users/RESTRICTED).
It is denied reads of the pfx, `secrets.db`, LocalDb, `appsettings.json`, logs and the KEKs, and denied writes in
`pki\trusted\certs`/`data`/`app`. No sockets. The PE cache fills and hits after a worker restart. **#891:** it can still read
other apps' `appsettings.json` and the OtOpcUaHost `.reg` export under `E:\psrem-stage` (Users-readable).
- **#780 QA runaway drill PASS:** a `while(true)` VT was killed at the 4000 ms hard deadline twice and quarantined on both
nodes. Sibling VT/Calc stayed all Good, `/health/ready` 200, ServiceLevel 250/240 unchanged. Removed afterwards and
redeployed back to rev `82dd0598…`. P4 recorded (InProcess = default + rollback). The ScriptedAlarm seam is still off on QA.
- **Issues:** closed #890, #794, #801, **#782 (SDK 2 upgrade)**. #780 open (ScriptedAlarm seam + other tiers). New #891.
- **ScadaBridge consumer impact: none today.** ScadaBridge QA has no OPC UA link to OtOpcUa. If it binds the canary, a
sick runner shows as `BadResourceUnavailable` on those two nodes (ScadaBridge#112). ScadaBridge#125: the 2.0 deferral
condition (#799 + #801 closed) is now met.
+2 -2
View File
File diff suppressed because one or more lines are too long