# Cleanup: delete test incidents and non-admin users Owner go-ahead required before step 4. Never run during business hours without telling HSE. 1. **Backup** (on the server, `ssh -p 9321 setia@175.136.239.121`): ```bash D=$(date +%Y%m%d-%H%M) pg_dump -p 5433 -Fc ims > /srv/apps/ims/backups/ims-pre-cleanup-$D.dump tar -czf /srv/apps/ims/backups/evidence-pre-cleanup-$D.tgz /var/lib/ims-evidence ls -lh /srv/apps/ims/backups/*pre-cleanup-$D* ``` Create `/srv/apps/ims/backups` first if missing. Both files must be non-empty. 2. **Upload** `db/ops/cleanup-test-data.sql` to `/srv/apps/ims/backups/` (not into `app/`). 3. **Dry run.** Use a DB role that owns the tables and bypasses RLS (superuser, or the owner role with BYPASSRLS). The tables have FORCE RLS and no DELETE policies, so as `app_user` or a plain owner the counts can read 0 and deletes silently remove nothing. Check first: `psql -p 5433 -d ims -Atc "select rolsuper or rolbypassrls from pg_roles where rolname = current_user"` must print `t`. Then: ```bash psql -p 5433 -d ims -v mode=dryrun -f /srv/apps/ims/backups/cleanup-test-data.sql ``` Send the admins list and the row counts to the owner. If incidents or users show 0 when the app shows rows, you are on the wrong role: stop. Mention that audit_log rows written by non-admin users about other tables (zones, trucks, settings) are also removed. Stop here until they confirm. 4. **Run** after explicit confirmation: ```bash psql -p 5433 -d ims -v ON_ERROR_STOP=1 -v mode=run -f /srv/apps/ims/backups/cleanup-test-data.sql ``` 5. **Verify**: `systemctl is-active ims` is `active`; log in as an admin at https://ims.setia.com.my/login; Incidents page is empty; Users page lists only admins. 6. **Rollback** (only if wrong): `pg_restore -p 5433 -d ims --clean --if-exists `; evidence files were never touched.