Restore Drill Report — May 2026
This is the report OffsiteDB generates and emails every month — one per account. It's the answer to the question that ends most backup conversations: “Can you prove the backups actually restore?” Every line below comes from a real restore drill: the snapshot was restored into a throwaway Postgres cluster and its rows were counted before it was marked proven. Forward it to an auditor, drop it into a security questionnaire, or hand it to your own leadership.
| Database | Policy | Runs | Drill evidence | Incidents | Verdict |
|---|---|---|---|---|---|
prod-api supabase | Hourly · keep 90d | 744 sealed / 2 failed (99.7%) 2026-05-01 → 2026-05-31 · 126.3 GB | 744 drills passed latest: 184 tables · 9.2M rows · 94s · worst-case 112s 4 pre-migration checkpoints sealed | 2 transient connection errors — both recovered on next run | DRILL-PROVEN |
billing-core neon | Every 6h · keep 365d | 124 sealed / 0 failed (100%) 2026-05-01 → 2026-05-31 · 26.1 GB | 124 drills passed latest: 61 tables · 2.1M rows · 38s · worst-case 44s 9 pre-migration checkpoints sealed | none | DRILL-PROVEN |
analytics-wh railway | Daily · keep 90d | 31 sealed / 0 failed (100%) 2026-05-01 → 2026-05-31 · 188.4 GB | 31 drills passed latest: 23 tables · 41.2M rows · 3m 28s · worst-case 3m 41s | none | DRILL-PROVEN |
legacy-crm amazon rds | Daily · keep 365d | 30 sealed / 0 failed (100%) 2026-05-01 → 2026-05-31 · 14.7 GB | 30 drills passed latest: 96 tables · 880K rows · 22s · worst-case 31s | 1 deadman alert on May 12 (credential rotation) — resolved same day, backups resumed | RESOLVED |
Generated automatically · every “proven” result is a real pg_restore into a scratch Postgres cluster with rows verified — not a checksum. Retention, RPO, and RTO are drawn from each database's actual schedule and measured drill times.
What makes this an audit answer, not a screenshot
- Tested, not assumed. Each row is a restore that actually ran — SOC 2's recovery-testing criterion (A1.3) asks for exactly this.
- Measured RTO. “Worst-case restore” is observed, not promised — real evidence of how fast you recover.
- Across the whole period. A Type II audit covers a window; the report spans the full month, every run.
- Honest about incidents. The deadman alert and its resolution are on the record. Auditors trust reports that show the blips and the fixes.
Get this for your own databases
Every plan includes restore drills and the monthly report — your data, in a bucket you own, proven to restore. Start a free trial, browse the backup guides for your provider, or run the SOC 2 checklist to see where you stand.