Determine incremental deployment methodology #28

Merged

tepichord merged 2 commits from feat/27-incremental-deployment-methodology into main

2026-08-22 04:37:17 -07:00

tepichord commented

2026-08-22 04:07:00 -07:00

Owner

Closes #27

Automated PR created by forgejo-done adapter.

Closes #27 Automated PR created by forgejo-done adapter.

tepichord added 2 commits

2026-08-22 04:07:00 -07:00

docs: define incremental deployment methodology 709c27c3b4

Document Milner's incremental deployment methodology in absolute terms so
agents adhere closely: (a) in-place Terraform delta is the default
everywhere past the VM boundary, (b) terraform apply -replace is used
only for a disposable VM at the base-infra layer, and (c) per-change
migration scripts are rejected. Adds the layer-boundary table, cleanup.sh
purpose (full cluster teardown only), best practices for (a)/(b), the
clean-install vs incremental-adopt single-graph framing, and the
per-component (a)/(b) mapping to ARCHITECTURE.md. Documentation only; no
source, Terraform config, cleanup.sh, or tests are modified.

Refs: #27

docs: reconcile cleanup.sh teardown description with its purpose dfdadac8af

The cleanup.sh section claimed the full-teardown tool 'must never
globally pkill QEMU' and 'must never silently delete state', which
contradicts both the script itself (pkill -9 -f qemu-system, removal
of ~/.talos cluster state) and the tool's own full-teardown purpose.
Reframe so destructive actions are reserved exclusively for full
teardown and never abused as a partial-recovery tool (recovery is
-replace). Docs only.