07 — Rollback / retire / cleanup plan
07 — ROLLBACK / RETIRE / CLEANUP PLAN
This entire lane is disposable. Nothing here is official runtime. Rollback is total and cheap.
On-disk copy: /opt/incomex/staging/c1/ROLLBACK.md.
A. Retire a single sandbox (normal cleanup, after a dry-run)
/opt/incomex/staging/c1/bin/dot-staging-sandbox-drop --sandbox-id c1_staging_<ts>
Governed retire: refuses any non-c1_staging_* DB and any DB without an active sbx_meta registry
row; captures a tombstone to /opt/incomex/staging/c1/_tombstones.jsonl; appends sandbox_drop to
the ledger; DROP DATABASE ... WITH (FORCE); readback-confirms gone; orphan sweep.
B. TTL
24h from each sandbox's sbx_meta.birth_certificate.created_at. After TTL, run A.
C. Emergency raw equivalent (audit reference — prefer A)
DROP DATABASE c1_staging_<ts> WITH (FORCE);
D. Full lane teardown (remove the staging deployment itself)
# 1) drop any live sandboxes (loop A over each c1_staging_* DB)
# 2) rm -rf /opt/incomex/staging/c1
Official runtime stays untouched (the lane never wrote to directus / dot_tools / dot/bin / any
official registry).
E. Nothing to revert in official runtime
- No official
dot_toolsrow created (registry is staging JSONL only). - No
/opt/incomex/dot/bindeploy. - No Directus / CAT-006 /
dot_agent_api_contractchange. - No APR / quorum / grant created; APR-0415 untouched. → before == after; rollback of official runtime is a no-op.
F. Current state to roll back from (this macro)
Only the staging deployment exists (/opt/incomex/staging/c1/, 18 files) and this KB package.
No sandbox DB exists yet (staging_DBs=0), so even the per-sandbox retire (A) has nothing to act on
until the dry-run runs. To undo this macro entirely: rm -rf /opt/incomex/staging/c1 (+ optionally
delete this KB package). That returns the system to the exact prior STAGED state.