You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Phase 11: Old account cleanup and deprecation #332
Clean up the personal Cloudflare account after migration is complete. Keep redirects in place for backward compatibility.
Tasks
Verify all traffic has shifted to SCCN account (check Workers analytics on both accounts)
Keep redirect worker api.osc.earth/nemar -> api.nemar.orgNOT done as designed -> root cause of Data integrity: 53 published datasets (with DOIs) have missing GitHub repos #883. The old nemar-api/nemar-api-dev were never reduced to a redirect; they were left as the FULL live backend with cron 0 3 * * * + GITHUB_ADMIN_PAT, which deleted 53 prod repos. Now deleted outright (2026-07-01); api.osc.earth/nemar/* -> 404. Re-add a tiny redirect-only worker if old-CLI backward-compat is still wanted.
Remove old D1 database on personal account -> deleted nemar-db, nemar-db-dev
Remove old Vectorize index on personal account -> deleted nemar-dataset-index, nemar-dataset-index-dev
Document the migration in project docs (what changed, new URLs)
Update CLAUDE.md with new infrastructure details
Update README and any references to old URLs
Depends On
All previous phases complete
Confidence that SCCN deployment is stable
Notes
Left intact on the personal account: osa-worker(+dev) serving api.osc.earth/osa/, hed-bot/hedit workers, wg-mesh(+db), printlab-db, and 3 non-nemar R2 buckets.
nemar-cli@latest (0.8.0+) on npm so users have a real upgrade path.
Old-URL traffic measured: monitor api.osc.earth/nemar request count over 30-90 days. Decommission once it falls below an acceptable threshold (e.g., <1% of total or a hard cap on absolute count).
Cleanup checklist (when ready)
Personal Cloudflare worker: switch from MAINTENANCE_MODE=read-only to a 301 redirect to api.nemar.org (or full removal). Decision point at start of phase.
Personal D1 (nemar-db, 0a168b1a-...) — archive, then delete. Keep a backup at backups/<date>/personal-prod.sql.
Personal Vectorize index — delete.
Personal Workers secrets — delete (security cleanup).
Update CLAUDE.md to remove the "Legacy URL" line.
Update infra_sccn_migration.md memory file with retirement date.
Decision needed at start of phase
Permanent 301 vs hard removal of the personal worker? 301 protects very-old CLIs forever but costs continued worker invocations. Suggest 301 for ~6 months post-removal of the personal D1 (read-only logic disappears anyway), then hard-remove.
Executed the personal-account teardown (crons removed, then all 4 nemar-* workers + both nemar D1s + both Vectorize indexes deleted). Verified api.osc.earth/nemar/* -> 404, /osa/ -> 200. This phase's incomplete teardown (full backend left live instead of a redirect) was the root cause of the #883 repo-deletion incident - see #883 and #314 for the full write-up. Remaining Phase 11 items: docs/README/CLAUDE.md URL updates, and an optional redirect-only worker for old CLIs.
Decision (2026-07-01): leave api.osc.earth/nemar/* at 404 — no redirect worker. Sunset is complete; old CLIs already cut over to api.nemar.org. The redirect-worker task is intentionally a won't-do. Remaining Phase 11 items are docs-only (docs/README/CLAUDE.md URL cleanup) plus optional prod secret rotation (#885).
Activity
neuromechanist commentedon Apr 26, 2026
Pickup notes
Prerequisites before starting
nemar-cli@latest(0.8.0+) on npm so users have a real upgrade path.api.osc.earth/nemarrequest count over 30-90 days. Decommission once it falls below an acceptable threshold (e.g., <1% of total or a hard cap on absolute count).Cleanup checklist (when ready)
api.nemar.org(or full removal). Decision point at start of phase.nemar-db,0a168b1a-...) — archive, then delete. Keep a backup atbackups/<date>/personal-prod.sql.infra_sccn_migration.mdmemory file with retirement date.Decision needed at start of phase
Permanent 301 vs hard removal of the personal worker? 301 protects very-old CLIs forever but costs continued worker invocations. Suggest 301 for ~6 months post-removal of the personal D1 (read-only logic disappears anyway), then hard-remove.
neuromechanist commentedon Jul 1, 2026
Executed the personal-account teardown (crons removed, then all 4
nemar-*workers + both nemar D1s + both Vectorize indexes deleted). Verifiedapi.osc.earth/nemar/*-> 404,/osa/-> 200. This phase's incomplete teardown (full backend left live instead of a redirect) was the root cause of the #883 repo-deletion incident - see #883 and #314 for the full write-up. Remaining Phase 11 items: docs/README/CLAUDE.md URL updates, and an optional redirect-only worker for old CLIs.neuromechanist commentedon Jul 1, 2026
Decision (2026-07-01): leave
api.osc.earth/nemar/*at 404 — no redirect worker. Sunset is complete; old CLIs already cut over toapi.nemar.org. The redirect-worker task is intentionally a won't-do. Remaining Phase 11 items are docs-only (docs/README/CLAUDE.md URL cleanup) plus optional prod secret rotation (#885).neuromechanist commentedon Jul 1, 2026
Personal-account NEMAR teardown complete: all 4
nemar-*workers + both nemar D1s + both Vectorize indexes + per-worker secrets removed;api.osc.earth/nemar/*→ 404 by decision (no redirect).osa-worker/hed/hedit/wg-mesh/printlab left intact. Residual doc-only cleanup (README/CLAUDE.md URLs) is non-blocking. Closing.