docs(P64): create phase plan — pre-mortem + teardown

---ci---
project: acdl
phase: P64
milestone: v1.11
status: plan
---/ci---
This commit is contained in:
Jon Chery
2026-07-28 20:39:09 +00:00
parent 06f4fc7705
commit 92bb03e808
+29 -47
View File
@@ -1,69 +1,51 @@
---
phase: P63
name: regression-registry-cost-docs
phase: P64
name: pre-mortem-teardown
milestone: v1.11
requirements: [REQ-119, REQ-121]
requirements: [REQ-120, REQ-122]
wave: 4
depends_on: [P62]
depends_on: [P63]
---
# P63Regression Registry + Cost Docs
# P64Pre-Mortem + Teardown
**Phase:** P63
**Phase:** P64
**Milestone:** v1.11 (RESTART)
**Requirements:** REQ-119 (COST.md), REQ-121 (CAP-017..022 in regression registry)
**Wave:** 4 (closing G-005/G-008)
**Requirements:** REQ-120 (PRE_MORTEM.md), REQ-122 (teardown)
**Wave:** 4 (closing G-005/G-008 + teardown)
**Branch:** `milestone/v1.11-restart`
## Goal
1. Add CAP-017..022 to `core/regression_verify.py` registry. Evidence =
"lifecycle pipeline green" (not boto3 probes, which are IAM-gated).
This closes G-005.
2. Author `.ciagent/COST.md` documenting the v1.0→v1.10 AWS spend window
(AWS Cost Explorer query). This closes G-008.
1. Author `.ciagent/PRE_MORTEM.md` — v1.10 decay root cause + forward
pre-mortem for the OSS reference + leadership pitch. Reuse the existing
pre-mortem content from commit e868a16 (branch phase/60-pre-mortem-doc).
2. `run_platform.sh --decommission` with CR CHG0680001 — tears down ALL
deployed stacks INCLUDING the 4 drifted VPCs from the failed first
attempt. HITL SRE gates (D-070 two-step). D-096 enforced (live resources
do not persist past v1.11).
## Tasks
### Task 1 — Add CAP-017..022 to regression registry
### Task 1 — Author .ciagent/PRE_MORTEM.md
Add 6 new capability checks to `core/regression_verify.py`:
- CAP-017: DynamoDB `acdl-contracts` table — evidence = L1 lifecycle
pipeline green (rds module apply/modify/destroy exit 0 proves the
terraform path works; DynamoDB table is created via the microservice
stack which is tested by the L2 lifecycle pipeline).
- CAP-018: Lambda contract-ingestor — evidence = local Lambda stub
(CAP-011) + lifecycle pipeline green.
- CAP-019: ECS cluster + service — evidence = L2 microservice lifecycle
pipeline green (apply/modify/destroy exit 0).
- CAP-020: CloudFront + WAF — evidence = L2 static-assets lifecycle
pipeline green (apply/modify/destroy exit 0).
- CAP-021: uptime-kuma — evidence = L1 uptime module lifecycle green.
- CAP-022: OIDC role — evidence = L1 iam-role module lifecycle green.
Cherry-pick commit e868a16 (branch phase/60-pre-mortem-doc) which contains
the pre-mortem document. Update to reflect v1.11 completion context.
Each check verifies that the corresponding lifecycle pipeline cell passed
(by checking the workflow run status or by running the lifecycle test
locally if AWS creds are available).
### Task 2 — Teardown (HITL escalation)
### Task 2 — Author .ciagent/COST.md
This task requires human approval (D-070 two-step + CR CHG0680001).
The teardown will be executed via:
- `run_platform.sh --decommission` with the CR
- Or via the Gitea CI workflow (if a teardown workflow exists)
- The CI VPC and any orphaned resources from failed lifecycle runs must
be cleaned up
Document the v1.0→v1.10 AWS spend window using AWS Cost Explorer data.
If Cost Explorer access is available (requires `ce:*` permissions), query
the 6-day window (2026-07-21 to 2026-07-27). If not, document the
expected cost ceiling based on the resources deployed.
### Task 3 — Update CAPABILITY_INVENTORY.md
Mark CAP-017..022 as "Verified live-aws via lifecycle pipeline" (no longer
"not auto-verified"). Remove the IAM-drift framing.
### Task 4 — Extend tests
Add tests for the new CAP-017..022 registry entries.
**Escalation:** This task pauses for human CR approval before executing
any destroy operations.
## Success Criteria (phase gate)
1. Regression registry includes CAP-017..022 with "lifecycle pipeline
green" evidence.
2. `COST.md` documents the v1.0→v1.10 spend window.
1. `PRE_MORTEM.md` documents the decay root cause + forward pre-mortem.
2. All deployed stacks torn down; zero live ACDL resources remain.
3. Full offline pytest suite green.