UCCA — RTOpacks Cleanup Completion Report (Tier 4)¶
Headline. The audit found this is not a "stale binding" cleanup — it is the unfinished MIGRATE-02 Phase 6. The old UCCA-account RTOpacks DBs were retained (per MIGRATE-02, pending Phase-6 sign-off), and the workers around them are either live-and-actively-using them, or dormant config. Nothing was a dead binding to sweep. Nothing retired; everything escalated per the brief's gates. The real decision is a coherent Phase-6 call, not piecemeal deletes.
Item 0 — Tier-3 ruling grounded ✅¶
rtopacks.com.auis ABSENT from the UCCA account's zones (present:ucca.asia, ucca.au, ucca.college, ucca.com.au, ucca.live, ucca.online). Ruling confirmed by live state..tfstill declares the fullrtopacks.com.aurecord set (root/www/admin/media/mx1-5/spf…) — fossils.- Terraform state not reachable this session (backend not initialised in-shell) — the follow-on
must
terraform state listwith backend creds. - Applied: dated correction note to the recon's Tier-3 (yellow-tape).
infra/ucca-infra/untouched.
Per-item verdicts (4.1–4.7)¶
| Item | Binds | Deployed? | Call sites | Verdict | How we know |
|---|---|---|---|---|---|
4.1 ops-v2 (ucca-ops) |
rtopacks-db |
YES (2026-03-26) | 19 files (subscribers, rto-geocoding, jsa-enrichment, world scope/provenance/enrichment/orders, worlds/rtopacks/*) | ESCALATE — live, genuine dependency | grep call sites + wrangler deployments |
| 4.2 stats-cache | rto-nrt-db (DB) |
NO | uses env.DB in code |
RETIRE-candidate (config-only, RTO-domain) | not deployed; code uses DB |
| 4.3 qual-enrichment | rto-nrt-db (DB) |
NO | refreshEnrichment(env.DB) |
RETIRE-candidate | same |
| 4.4a cricos-sync | rto-nrt-db (DB) |
NO | ingester(env.DB, …) |
RETIRE-candidate | same |
4.4 backup (ucca-backup) |
rtopacks-db (backs it up) |
YES (2026-03-26) | backs up the 450 MB rtopacks-db | ESCALATE — live backup of the retained DB | wrangler name/deployment + config |
| 4.5 rtopacks-proxy | — | NO | — | RETIRE-candidate (no traffic — not deployed) | deployment check |
| 4.6 rtopacks-docs-proxy | — | NO | — | RETIRE-candidate (no traffic) | deployment check |
| 4.7 ops-v2 world code | via rtopacks_db |
(in ucca-ops) | lib/worlds/rtopacks/api/* (enrich/nrt/metrics/facets/quals/units/rtos) |
AUDIT ONLY — verdict deferred (brief §4.7; with/after Tier-2 severance) | no import-coupling to engine worlds/ (separate repos; talks via DB/API) |
Why nothing was retired (the brief's gates fired correctly)¶
- Live + active (ops-v2, ucca-backup): the brief's 4.1 gate — "genuine dependency that should have
died at untangle → escalate, do not repoint/retire silently." These actively operate on the
retained
rtopacks-db. Retiring them stops live function (ops-console world views; DB backups) and is a Phase-6 decision, not a binding sweep. - Dormant config (stats-cache, qual-enrichment, cricos-sync, both proxies): RETIRE-candidates (config-only, not deployed, RTO-domain, RTOpacks self-manages now) — but they are the same Phase-6 residue as the DBs and the live workers. Deleting them piecemeal while the live workers still depend on the retained DBs is incoherent. Held for the Phase-6 decision.
The escalation — the actual next step¶
This tier converts to a MIGRATE-02 Phase-6 decision, taken as a unit, confirmed with RTOpacks:
1. Have RTOpacks taken over the functions these workers did (cricos sync, qual enrichment, stats,
backups) on their side? (Almost certainly — they self-manage.)
2. If yes → retire the retained old UCCA-account DBs (rtopacks-db, rto-nrt-db, abs-db) and
all workers around them (the 5 dormant + ucca-backup) and ops-v2's rtopacks world surface, as
one reviewed unit. If any is still needed UCCA-side, that's the exception to name.
3. ops-v2 (4.1/4.7) is the sensitive one: L2 ops managing the rtopacks world may be by design — its
world views + the 19 call sites need Tim's keep-or-cut before ops-v2 loses rtopacks_db.
Follow-on noted, NOT started¶
- TERRAFORM-RECONCILE (Tier-3 state hygiene) — waits for the proof-job deploy, per the brief. Not started.
Full Cycle¶
- No applied change → no Infrastructure-State snapshot change. Recon Tier-3 got the dated correction.
- Divergences from the recon, recorded:
backupbindsrtopacks-dbfor backup (not a stale data binding as the recon grouped it); and 5 of the 6 "stale-binding" workers are not deployed — the recon presumed live bindings, live state shows config-only.
Completion report. Item 0 confirmed the ruling; every Tier-4 item hit a stop-and-report gate because none was a dead binding — it's the unfinished Phase 6. Nothing retired, nothing redeployed, no infra touched. The brief's Tier-4 does not close by deletion; it closes by escalating a Phase-6 decision.