Backup, disaster recovery, migration and architecture reviews

Why and what

RTO is the target time to restore service. RPO is the tolerated data-loss window. Availability handles failures while a service operates; disaster recovery handles larger recovery scenarios. Multi-AZ deployment, replicas and backups solve overlapping but different problems. A replica can faithfully replicate accidental deletion.

Recovery strategies

Backup-and-restore trades longer recovery for lower standby cost. Pilot light keeps essential components/data ready. Warm standby maintains a reduced running environment. Multi-site active architectures can reduce downtime but add data-consistency, routing and operational complexity. Choose based on requirements rather than a universal “best” pattern.

Practical recovery lab

1. Write an RTO/RPO goal for the reels capstone, for example a classroom target rather than a production promise.
2. Record infrastructure definitions, package versions, application code, secret retrieval and database schema.
3. Back up the database and retain required S3 object versions according to policy.
4. Restore into a separate environment. Point only a test hostname at it.
5. Verify a known post, its media object, database permissions, TLS and application health.
6. Measure restore time and the newest recoverable data timestamp. Compare with the stated targets.

Migration overview

Use discovery to inventory servers, dependencies, data volume and downtime limits. AWS Application Migration Service supports server migration patterns; Database Migration Service supports compatible database migration/replication scenarios; DataSync moves supported file/object datasets. Validate schema compatibility and application behaviour separately from byte transfer.

Architecture review

Review operational excellence, security, reliability, performance efficiency, cost optimization and sustainability. Ask concrete questions: what fails if one AZ disappears, who can delete backups, how are secrets rotated, and what is the most expensive request path? Keep a risk register with owners and evidence.

Interview and cleanup

Why is “snapshot completed” not a recovery test? It proves creation, not application usability. Remove the recovery environment after documenting results, while preserving the approved backups. Do not test failover by disrupting production resources without a change plan.

Official references

AWS Well-Architected AWS disaster recovery

Interview and revision check

How do RTO and RPO differ?

RTO limits recovery time; RPO limits tolerable data loss measured back from the incident. A design must satisfy both.