Skip to main content
  1. Monthly (automated, verify): cron creates partitions N+2 for all designated tables; check job green + partitions exist (\d+ core.tracked_events).
  2. DEFAULT partition alert fires: rows landed outside ranges ⇒ a partition wasn’t created or bad occurred_at. Create missing partition, move rows, find producer bug, add test.
  3. Retention cycle (per tenant plan): export partition → Storage exports/{tenant}/{table}/{yyyymm}.ndjson.gz → checksum → verify readable → DROP partition → record in audit_log (system actor).
  4. pg_partman spike result: [pending] if available on Supabase, migrate creation/retention to partman config and simplify steps 1–3.
  5. Restore request: locate export in Storage → load into restore_{table} staging → hand to tenant-restore runbook.