Recovery Testing separates service readiness, the data point restored, and business consistency. A healthy probe or a started database does not prove that users have recovered the right state.
Define the recovery objective from its source
Record the approved RTO and RPO definitions, the failure start, the business-recovery end, and the acceptable-loss boundary from project evidence. RTO ends when the agreed business path is ready, not automatically when a process or health check starts. RPO depends on an identified recovered data point; without it, do not calculate the value.
Join service, data, and business timelines
Connect the failure time, backups or logs, restore steps, dependency readiness, and business validation. Keep service restart and data recovery as separate events.
| Recovery evidence | What to reconcile |
|---|---|
| Restored data point | Source backup or log position and its timestamp |
| Business records | IDs, amounts, references, duplicates, and missing intervals |
| Dependencies | Queue replay, identity, configuration, and downstream readiness |
| Completion | Business validation, compensation, and operator handoff |
Compare restored or replayed records against source ledgers. Record missing intervals and compensation obligations instead of treating partial restoration as complete recovery.
Illustrative calculation: assume the project record puts the failure start at 10:00, the agreed business path becomes ready at 10:40, and the restored data point is 09:57, with all timestamps on the same time basis. RTO is 40 minutes. If the project defines its RPO comparison as the interval from failure start to the latest restored data point, the observed recovery gap is 3 minutes. Compare both results with approved objectives; missing objective or timestamp sources do not support a pass.
Make takeover and partial recovery visible
If automation stops or a validation step diverges, record who decides whether to continue, roll back, or escalate. A dry-run, process restart, or successful health probe is not evidence that the service has recovered the correct data or met its timing objective.
Distinguish recovery testing from disaster recovery and resilience
Recovery Testing focuses on the sourced restore point, recovery timeline, state validation, and Human takeover for a defined scenario. Disaster Recovery Testing broadens the scope to disaster boundaries, backup strategy, and cross-region dependencies. Resilience Testing evaluates preserved behavior during degradation as well as recovery. Keep those conclusions separate.
Record an RCV-## finding
Capture the affected object or rule, source, trigger or applicability, expected concern, evidence state, impact and priority, owner role, close condition, and validation method. A timeline should distinguish service readiness, recovered data, business consistency, and Human handoff.
Prepare a bounded recovery review
Compare supplied objectives, backups, runbooks, dependencies, validation queries, and prior evidence as static material. Any restore exercise is a separately authorized candidate with a defined stop condition. Do not access production or claim an exercise passed without its run evidence.
Related Skills
- Disaster Recovery Testing: https://inaodeng.com/en/qaskills/disaster-recovery-testing/ covers broader disaster scope and recovery dependencies.
- Resilience Testing: https://inaodeng.com/en/qaskills/resilience-testing/ checks degraded behavior and safe exit.
- Failover Testing: https://inaodeng.com/en/qaskills/failover-testing/ focuses on primary/standby traffic and write authority.
Install and invoke
Install the Skill:
npx skills add https://github.com/naodeng/awesome-qa-skills/tree/main/skills/en --skill recovery-testing
Invoke recovery-testing with sourced RTO/RPO definitions, failure and business-ready timestamps, backup or log evidence, dependencies, reconciliation records, and takeover rules:
Review this redacted recovery timeline with recovery-testing.
Objectives: [RTO/RPO definitions, failure start, business-ready end, accepted data loss]
Evidence: [backup/log data point, restore steps, dependency readiness, business reconciliation, and operator handoff]
Assess service readiness, the restored data point, and business consistency separately. Calculate only objective deltas supported by the timestamps, then list candidate RCV-## findings, missing evidence, and the smallest isolated validation action. Do not treat a passing health probe as business recovery.
FAQ
Does a passing health probe end the RTO clock? Only if the project defines it as the business-recovery event. Otherwise, use the agreed time when the business path is ready.
Can RPO be calculated without a recovered data point? No. Record the missing point and source; do not substitute a backup-file timestamp for the data actually restored.
Source
Read the Recovery Testing prompt: https://github.com/naodeng/awesome-qa-skills/tree/main/skills/en/testing-types/recovery-testing/prompts/recovery-testing.md. It defines the RCV-## finding contract and separates service, data, and business recovery.