# Disaster recovery

Disaster recovery is the planned restoration of applications, infrastructure, and data after a disruptive event within defined downtime and data-loss limits.

Cloud concept. Updated 2026-10-29. Page: https://blowstack.com/tools/cloud-concepts-hub/disaster-recovery

> Disaster recovery combines recoverable data, replacement infrastructure, documented procedures, and testing to restore technology services after failures, attacks, or operational mistakes. In the cloud, it uses capabilities such as backups, replication, automated provisioning, and geographically separate deployments to balance recovery requirements against cost and complexity.

Practice exams covering Disaster recovery: [CLF-C02 AWS Certified Cloud Practitioner](https://blowstack.com/practice-exams/aws-certified-cloud-practitioner), [SOA-C03 AWS Certified CloudOps Engineer - Associate](https://blowstack.com/practice-exams/aws-certified-cloudops-engineer-associate), [DOP-C02 AWS Certified DevOps Engineer - Professional](https://blowstack.com/practice-exams/aws-certified-devops-engineer-professional), [SCS-C03 AWS Certified Security - Specialty](https://blowstack.com/practice-exams/aws-certified-security-specialty), [SAA-C03 AWS Certified Solutions Architect - Associate](https://blowstack.com/practice-exams/aws-certified-solutions-architect-associate).

## 6 key aspects of Disaster recovery

Effective disaster recovery combines measurable objectives, isolated and protected recovery resources, appropriate architecture, and repeatedly tested restoration procedures.

1. **Business-driven recovery objectives**: Define RTO and RPO for each workload based on the impact of downtime and lost data. Use these targets to select backup frequency, replication methods, infrastructure readiness, and testing schedules.
2. **Independent recovery locations**: Use separate failure domains appropriate to the threat, such as another Region for regional disruption or another account for administrative isolation. Confirm that recovery locations satisfy data residency requirements and have the necessary services, quotas, and capacity.
3. **Recoverable history alongside replication**: Use replication to improve data availability and retained backups to recover earlier states. Do not assume a current replica protects against logical corruption, malicious changes, or ransomware.
4. **Protected backup data and encryption keys**: Restrict backup deletion, apply immutable retention where appropriate, and separate backup administration from production administration. Ensure destination permissions and encryption-key policies support recovery without allowing unnecessary access.
5. **Cost-aware recovery strategies**: Choose between backup and restore, pilot light, warm standby, and active-active architectures according to workload requirements. Faster recovery generally requires more continuously provisioned resources and greater operational complexity.
6. **Tested restoration and failback**: Exercise recovery procedures with realistic dependencies and measure achieved RTO and RPO. Include application validation, traffic switching, rollback decisions, and controlled return to the primary environment.

## Cloud services for Disaster recovery

AWS Backup coordinates protection and retention across supported resources, while Amazon Data Lifecycle Manager handles EBS snapshot and AMI lifecycle needs. AWS Elastic Disaster Recovery maintains server replication for recovery workflows. AWS DataSync can move file and object datasets into a recovery location, and Amazon S3 can retain versioned or retention-protected copies. AWS KMS supplies encryption-key controls that must remain usable during restoration. Together, these services support a recovery architecture, but engineers must still configure geographic and account isolation, protect historical copies, rebuild dependencies, and test complete application recovery against RTO and RPO.

- **[AWS Backup](https://blowstack.com/tools/cloud-services-hub/aws-backup)**: AWS Backup centrally manages backup policies, schedules, retention, and restores for supported resources, with cross-account and cross-Region copying where supported. AWS Backup Vault Lock can enforce retention controls; compliance mode prevents changes or removal after its configured grace period, helping protect recovery points against malicious deletion.
- **[AWS Elastic Disaster Recovery](https://blowstack.com/tools/cloud-services-hub/aws-elastic-disaster-recovery)**: AWS Elastic Disaster Recovery replicates supported source servers into a staging area in AWS and launches recovery instances for drills or recovery. It supports lower-RPO server recovery designs, but application dependencies, validation, traffic redirection, and failback still require planning.
- **[AWS DataSync](https://blowstack.com/tools/cloud-services-hub/aws-datasync)**: AWS DataSync transfers file and object data between supported storage systems, including on-premises storage and AWS storage services. It can populate or refresh recovery datasets, but independent versions, retention policies, and destination protection are needed because a transfer task alone is not an immutable backup.
- **[Amazon S3](https://blowstack.com/tools/cloud-services-hub/amazon-s3)**: Amazon S3 provides storage for recovery data, with Versioning, replication, and Object Lock supporting different protection requirements. Object Lock protects specified object versions against deletion or overwrite according to its retention settings, while replication requires explicit configuration and monitoring.

## Disaster recovery exam questions

Real practice exam questions on Disaster recovery, with the correct answer and its explanation.

### Question 1 (CLF-C02)

**Question:**

A consulting company backs up its project-management database every night and uses infrastructure templates to recreate its application servers. During a recovery exercise, the company discovers that all database backups are stored in the same AWS Region as the production application. The company must be able to restore service if that entire Region becomes unavailable, and losing changes made since the previous night's backup is acceptable. Which improvement should the company make?

**Correct answer:** Maintain copies of the database backups in another Region and test restoring the application and its data there.

**Explanation:** Backups in another Region provide a recovery source outside the affected Region. Testing restoration verifies that the backups and infrastructure templates can be used together to restore service. Nightly backups fit the company's stated tolerance for data loss.

### Question 2 (SOA-C03)

**Question:**

A company uses Amazon Route 53 failover A records to route a public application to endpoints in two AWS Regions. Both records have a TTL of 3,600 seconds. The secondary endpoint is ready to serve production traffic.

A scheduled disaster recovery exercise will make the primary endpoint unhealthy and verify traffic redirection. Once Route 53 begins returning the secondary address, DNS resolvers must stop returning the cached primary address within 60 seconds. Assume that resolvers honor record TTLs, do not serve expired responses, and clients perform new DNS lookups rather than reuse existing connections. Health check detection time is measured separately.

Which preparation should a CloudOps engineer perform to meet this requirement?

**Correct answer:** Reduce both failover records' TTLs to 60 seconds at least 3,600 seconds before initiating the exercise.

**Explanation:** Waiting out the previous TTL allows answers cached with the original 3,600-second TTL to expire before the exercise. Subsequently cached answers have a TTL of 60 seconds, limiting how long compliant resolvers can continue returning the primary address after Route 53 changes its answer.

### Question 3 (SCS-C03)

**Question:**

A company is configuring a scheduled AWS DataSync task to copy files from an on-premises NFS server to an Amazon S3 bucket for recovery from accidental deletion or corruption. Production users cannot administer the destination bucket. The task will use TransferMode=CHANGED and map each source file to the same S3 key on every run.

The recovery design must meet these requirements:
• New and modified source files are copied during each scheduled run.
• Earlier copies of modified files remain recoverable after subsequent runs.
• If a file is deleted from the source, its most recently copied contents remain accessible through a standard S3 GET request without a version ID.

The bucket has no Lifecycle expiration rules, and no other process modifies destination objects. Which configuration meets these requirements?

**Correct answer:** Enable S3 Versioning before the first transfer. Set OverwriteMode=ALWAYS and PreserveDeletedFiles=PRESERVE.

**Explanation:** ALWAYS allows DataSync to copy changed source files over their existing destination keys. S3 Versioning retains the previous object versions for recovery. PRESERVE keeps destination objects when their corresponding source files are deleted, leaving their latest copied contents accessible without specifying a version ID.

### Question 4 (SAA-C03)

**Question:**

A company normally serves its public application from one AWS Region and maintains a fully provisioned disaster recovery deployment in a second Region. Each deployment has a fixed public IPv4 endpoint, can handle the full workload, and exposes a Region-specific HTTPS readiness endpoint. Data replication already meets the company's recovery requirements.

For a week-long disaster recovery exercise, the company wants approximately 80% of new authoritative DNS answers to select the primary deployment and 20% to select the recovery deployment while both are healthy. If either deployment becomes unhealthy, DNS answers must select only the healthy deployment. DNS caching delays and differences between DNS-answer percentages and actual request percentages are acceptable.

Which Amazon Route 53 configuration meets these requirements?

**Correct answer:** Create two weighted A records for the application hostname with weights of 80 and 20 for the primary and recovery endpoints, respectively. Associate each record with an HTTPS health check of its Region-specific readiness endpoint.

**Explanation:** Weighted routing distributes DNS answers approximately in proportion to the configured weights. With health checks associated with both records, Route 53 excludes an unhealthy endpoint while the other endpoint remains healthy. This supports both the controlled recovery exercise and routing away from a failed deployment.

### Question 5 (DOP-C02)

**Question:**

A company has failed over an application with a self-managed PostgreSQL database on Amazon EC2 to a recovery AWS Region. The recovery Region now processes all production writes. The original Region is available again, but its database is stale and remains fenced from application writes.

The company must automate the return to the original Region without losing any transaction acknowledged by the recovery environment. The application can be unavailable for no more than 5 minutes during this return. Rebuilding the database in the original Region from the active database takes 2 hours. The rebuilt database can then continuously receive transaction logs while production remains active in the recovery Region.

Tests show that blocking new writes, draining in-flight transactions, replaying the remaining logs, promoting the synchronized database, switching application connections, and validating service take a total of 3 minutes. Restoring a database snapshot takes 25 minutes.

Which automated procedure meets the requirements?

**Correct answer:** Rebuild the original Region's database from the active recovery database while production continues. Stream transaction logs to it. At cutover, block new writes and drain in-flight transactions in the recovery Region, confirm that the destination has replayed the final committed transaction, and keep the recovery database fenced while promoting the destination, switching connections, and validating service.

**Explanation:** The 2-hour rebuild occurs without interrupting production. Fencing writes and replaying through the final committed transaction preserve every acknowledged transaction and prevent divergent writers. The tested final cutover takes 3 minutes, meeting the 5-minute outage limit.

## Practice Disaster recovery in a full exam

Questions on Disaster recovery in each practice exam, with every answer option and an explanation for each:

- [CLF-C02](https://blowstack.com/practice-exams/aws-certified-cloud-practitioner): 9 questions
- [SAA-C03](https://blowstack.com/practice-exams/aws-certified-solutions-architect-associate): 8 questions
- [DOP-C02](https://blowstack.com/practice-exams/aws-certified-devops-engineer-professional): 8 questions
- [SOA-C03](https://blowstack.com/practice-exams/aws-certified-cloudops-engineer-associate): 7 questions
- [SCS-C03](https://blowstack.com/practice-exams/aws-certified-security-specialty): 3 questions

## What is Disaster recovery

Disaster recovery addresses how an organization restores technology services when their normal operating environment becomes unavailable or untrustworthy. Events include regional outages, accidental deletion, failed deployments, compromised credentials, and ransomware. It is part of business continuity, which also covers people, facilities, communications, and business processes. A disaster recovery plan must identify dependencies and recovery order: restoring an application server is insufficient if its database, identity provider, encryption keys, or network connectivity remain unavailable.

Begin with recovery time objective, or RTO, and recovery point objective, or RPO. RTO is the target maximum time to restore a service after disruption. RPO is the maximum acceptable data loss measured in time. For example, an order-processing application might require recovery within one hour and tolerate losing no more than five minutes of transactions. A daily backup alone cannot reliably meet that RPO; frequent backups, transaction-log capture, or replication would be needed. Actual recovery performance must be measured rather than inferred from a product feature.

Cloud recovery strategies range from backup and restore to pilot light, warm standby, and multi-site active-active deployments. Backup and restore generally has lower ongoing infrastructure costs but requires more rebuilding and data restoration. Pilot light keeps essential components available while other resources are deployed or scaled during recovery. Warm standby maintains a smaller functioning environment that can be expanded. Active-active deployments can reduce interruption but introduce costs and challenges around consistency, conflict resolution, routing, and operational coordination. None automatically guarantees zero downtime or zero data loss.

Availability Zones provide isolation within an AWS Region and help applications tolerate localized infrastructure failures. Multiple Zones do not provide protection against every Region-wide disruption. A secondary Region can provide geographic separation for recovery, but the application must have usable data, configuration, permissions, capacity, and recovery procedures there. Multiple Regions also support other goals, including lower latency for geographically distributed users and meeting jurisdictional requirements. These goals are distinct: copying data into another country for recovery can violate data-sovereignty restrictions unless the destination and processing arrangements are permitted.

Replication and backup solve different problems. Replication can make recent data available elsewhere, but it can also propagate corruption, unwanted changes, or encrypted files. Retained backups provide historical recovery points. For ransomware resilience, combine appropriately configured immutable retention with restricted deletion permissions, separate administrative boundaries, monitoring, and restore testing. Cross-account and cross-Region copies reduce different risks and are not interchangeable. Encrypt backups and replicated data, protect encryption keys from unauthorized use or deletion, and verify that recovery identities can decrypt the destination copies.

For example, an engineering team could centrally schedule database and volume backups, retain protected copies in a separate account, and copy supported backups to an approved recovery Region. It could also replicate critical servers to reduce recovery time. Infrastructure-as-code templates would recreate networks and application resources, while a runbook would specify database recovery, application validation, traffic redirection, and eventual failback. Recovery exercises should verify data integrity, application behavior, recovery duration, replication lag, quotas, and key access. A successful backup job is evidence that data was captured, not proof that the business service can be recovered.

## Learn Disaster recovery in lessons

10 short lessons on Disaster recovery in 5 certification courses.

- CLF-C02: Backup and Restore vs. Disaster Recovery (Premium), https://app.blowstack.com/lessons/backup-and-restore-vs-disaster-recovery-clf-c02
- SOA-C03: Backup and Restore vs Disaster Recovery (Premium), https://app.blowstack.com/lessons/backup-and-restore-vs-disaster-recovery-soa-c03
- SCS-C03: Disaster recovery: protecting recovery data (Premium), https://app.blowstack.com/lessons/disaster-recovery-protecting-recovery-data-scs-c03
- SAA-C03: Disaster recovery: recovery strategies (Premium), https://app.blowstack.com/lessons/disaster-recovery-recovery-strategies-saa-c03
- DOP-C02: Disaster recovery: recovery architecture trade-offs (Premium), https://app.blowstack.com/lessons/disaster-recovery-recovery-architecture-trade-offs-dop-c02
- CLF-C02: High availability and disaster recovery: continuity versus restoration (Premium), https://app.blowstack.com/lessons/high-availability-and-disaster-recovery-continuity-versus-restoration-clf-c02
