# Pilot Light

Pilot light is a disaster recovery strategy that keeps essential components and replicated data ready in a recovery environment while leaving most application capacity inactive until needed.

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

> A pilot light environment maintains the minimum foundation required to rebuild a production workload after a disaster, typically including continuously replicated data and predefined infrastructure. It balances ongoing cost against recovery speed, usually recovering faster than backup and restore but requiring more activation work than warm standby.

Practice exams covering Pilot Light: [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), [SAA-C03 AWS Certified Solutions Architect - Associate](https://blowstack.com/practice-exams/aws-certified-solutions-architect-associate), [SAP-C03 AWS Certified Solutions Architect - Professional](https://blowstack.com/practice-exams/aws-certified-solutions-architect-professional).

## 6 key aspects of Pilot Light

Pilot light combines a minimal operational core, replicated data, and tested activation procedures to balance recovery objectives against ongoing infrastructure cost.

1. **Minimal operational recovery core**: Keep essential stateful components or a replication staging environment running in the recovery location. Leave most application capacity inactive until recovery is initiated.
2. **Replication aligned with RPO**: Choose replication mechanisms that support the workload's acceptable data-loss window, and monitor their lag and health. Asynchronous replication does not guarantee zero data loss.
3. **Activation aligned with RTO**: Measure the complete recovery workflow, including detection, authorization, provisioning, data recovery, application validation, and traffic cutover. Use those measurements to determine whether pilot light meets the business requirement.
4. **Repeatable infrastructure and recovery automation**: Maintain versioned infrastructure definitions, deployable artifacts, and tested recovery workflows. Prepare regional dependencies such as permissions, encryption keys, secrets, and quotas before an incident.
5. **Cost versus readiness trade-off**: Reduce ongoing compute costs by not running a complete application stack in the recovery location. Account for continuing replication, storage, database, and testing costs, as well as the extra work needed to activate service.
6. **Independent backups and tested failback**: Keep recoverable historical copies because replication can propagate damaging changes. Exercise both recovery and the eventual return to the preferred operating location, including write isolation and data synchronization.

## Cloud services for Pilot Light

Amazon RDS cross-Region replicas or AWS Elastic Disaster Recovery can maintain recoverable workload data while most application compute remains inactive. AWS CloudFormation defines the infrastructure needed to activate the recovery environment, and Amazon EC2 supplies compute capacity when recovery begins. After data recovery and application validation, Amazon Route 53 can direct users to healthy recovery endpoints. AWS Backup provides historical recovery points for failures that replication cannot safely address, such as propagated corruption. Engineers must connect these capabilities through tested orchestration, appropriate permissions, monitoring, and explicit failover and failback procedures.

- **[Amazon RDS](https://blowstack.com/tools/cloud-services-hub/amazon-rds)**: For supported database engines and configurations, cross-Region read replicas can maintain the data tier of a pilot light environment. Recovery includes promoting the replica, updating application connections, and validating data; asynchronous replication means recent transactions may be missing.
- **[AWS Elastic Disaster Recovery](https://blowstack.com/tools/cloud-services-hub/aws-elastic-disaster-recovery)**: Continuously replicates source-server block-level data into a staging area in the recovery Region. Recovery launches usable instances from the replicated data, supporting a pilot light-style approach without continuously running full recovery servers.
- **[AWS CloudFormation](https://blowstack.com/tools/cloud-services-hub/aws-cloudformation)**: Defines recovery infrastructure as version-controlled templates so missing application resources can be provisioned consistently. Templates must account for regional dependencies and be exercised regularly rather than assumed to work during an outage.
- **[Amazon EC2](https://blowstack.com/tools/cloud-services-hub/amazon-ec2)**: Provides application instances that can be launched or started during recovery using prepared images and launch templates. Recovery planning must address instance availability, quotas, configuration, and the time required to reach usable capacity.
- **[Amazon Route 53](https://blowstack.com/tools/cloud-services-hub/amazon-route-53)**: Can redirect DNS traffic to the recovery environment after its application endpoints are ready. Health checks and routing policies assist cutover, but they do not provision the missing application stack, and DNS caching affects transition time.
- **[AWS Backup](https://blowstack.com/tools/cloud-services-hub/aws-backup)**: Adds scheduled recovery points, retention policies, and supported cross-Region or cross-account backup copies alongside replication. It protects against scenarios requiring an earlier recovery point, but backup management alone does not establish a pilot light environment.

## Pilot Light exam questions

Real practice exam questions on Pilot Light, with the correct answer and its explanation.

### Question 1 (SOA-C03)

**Question:**

A company manages its recovery environment in a second AWS Region with a single AWS CloudFormation stack. The stack contains a continuously replicated database and Amazon EC2 application instances. To reduce costs, a scheduled workflow deletes the entire stack each night and recreates it each morning. Reestablishing database replication after recreation requires a full data synchronization.

The company now requires an RPO of 5 minutes at all times and an RTO of 30 minutes. It also wants to eliminate charges for idle application instances between recovery exercises. Monitoring confirms that replication lag remains below 1 minute when the recovery database is running. Recovery tests show that launching the application instances takes 12 minutes, and database promotion, application validation, and traffic redirection take another 8 minutes. Restoring and synchronizing a replacement database takes 70 minutes.

Which change should a CloudOps engineer make?

**Correct answer:** Separate the replicated database and its required dependencies into a persistent stack. Deploy the application instances from a separate stack only during recovery or exercises.

**Explanation:** Keeping the replicated data core operational preserves the low replication lag needed for the RPO. Provisioning application compute only when needed eliminates idle application-instance charges. The measured recovery time is 20 minutes, within the 30-minute RTO. This applies the pilot light strategy.

### Question 2 (SAP-C03)

**Question:**

A logistics company runs a shipment-tracking application on Amazon EC2 and an Aurora PostgreSQL global database. The primary application and database writer are in one AWS Region. A secondary DB cluster continuously receives replicated data in a recovery Region. Monitoring and recovery exercises demonstrate that replication meets the company's 1-minute RPO.

The company requires an end-to-end RTO of 25 minutes and wants to minimize ongoing recovery infrastructure costs. Application instances do not need to run in the recovery Region during normal operations. Regional networking, machine images, configuration, deployment artifacts, and sufficient service quotas are already available there.

Recovery exercises establish these maximum durations:
• Detect the outage and authorize recovery: 3 minutes.
• Promote the secondary database and make it ready for application connections: 4 minutes.
• Launch production-capacity application instances and complete initialization that does not require database connectivity: 14 minutes.
• Configure database connections and validate business transactions after both tiers are ready: 5 minutes.
• Redirect traffic and confirm service availability: 2 minutes.

Restoring a database from a recovery-Region backup instead of promoting the secondary takes at least 55 minutes. Application initialization can run independently of database promotion.

Which recovery design meets the requirements with the LOWEST ongoing infrastructure cost?

**Correct answer:** Retain the secondary database but keep application instances absent during normal operations. After recovery authorization, promote the database and launch production-capacity application instances concurrently. After both operations finish, configure connections, validate transactions, and redirect traffic.

**Explanation:** This pilot light design retains the continuously replicated database while avoiding ongoing application compute costs. Parallel database promotion and application initialization produce an end-to-end recovery time of 3 + max(4, 14) + 5 + 2 = 24 minutes, within the 25-minute RTO. The existing replication meets the RPO.

### Question 3 (SAA-C03)

**Question:**

A company runs a public order-processing application in one AWS Region. In a recovery Region, the company maintains a continuously replicated database, infrastructure templates, and application images. No application servers run in the recovery Region during normal operation. Database replication lag remains below 30 seconds.

The company requires an RPO of 1 minute and an RTO of 15 minutes for a primary Region outage. Tests show that an automated workflow can promote the recovery database, launch application servers, and validate order processing within 8 minutes. The same procedure takes at least 40 minutes when performed manually. Clients retry failed connections and honor the application's DNS record TTL of 60 seconds.

Which solution meets the recovery objectives while preserving the existing low-cost recovery strategy?

**Correct answer:** Automatically invoke the tested recovery workflow when a primary Region outage is detected. After the workflow promotes the database, launches the application servers, and validates order processing, have it update the application's Route 53 record to the recovery endpoint.

**Explanation:** This preserves the pilot light strategy by keeping application compute inactive until recovery is needed. The tested 8-minute activation workflow, followed by DNS redirection with a 60-second TTL, fits within the 15-minute RTO. Continuous replication with lag below 30 seconds supports the 1-minute RPO.

### Question 4 (DOP-C02)

**Question:**

A company runs an application on Amazon EC2 instances in one AWS Region. A database replica in a recovery Region maintains replication lag below 30 seconds. Networking, encryption keys, and instance quotas are already configured in the recovery Region, but no application instances normally run there.

The application release pipeline frequently changes both the application version and the database schema. Currently, application machine images and deployment configuration are available only in the primary Region. Recovery tests with locally available, schema-compatible artifacts complete database promotion, application deployment, transaction validation, and traffic redirection within 12 minutes. Restoring the database from backups takes at least 70 minutes.

The company requires an RTO of 20 minutes and an RPO of 60 seconds during a complete primary-Region outage. Recovery must not depend on accessing the primary Region. The company wants the LOWEST ongoing application compute cost.

Which change should a DevOps engineer implement?

**Correct answer:** Extend the release pipeline to copy application AMIs and versioned deployment configuration to the recovery Region. Require their availability and schema compatibility before activating each production release. Maintain a recovery-Region workflow that promotes the replica, launches application instances from those artifacts, validates transactions, and redirects traffic.

**Explanation:** This preserves the replicated data foundation while leaving application compute undeployed until recovery. Release gating ensures that recovery artifacts are available locally and compatible with the replicated database schema. The independent workflow can use the tested 12-minute recovery procedure, and replication lag supports the 60-second RPO without ongoing application instance costs.

## Practice Pilot Light in a full exam

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

- [SOA-C03](https://blowstack.com/practice-exams/aws-certified-cloudops-engineer-associate): 4 questions
- [SAA-C03](https://blowstack.com/practice-exams/aws-certified-solutions-architect-associate): 2 questions
- [DOP-C02](https://blowstack.com/practice-exams/aws-certified-devops-engineer-professional): 2 questions
- [SAP-C03](https://blowstack.com/practice-exams/aws-certified-solutions-architect-professional): 1 question

## What is Pilot Light

Pilot light is a disaster recovery strategy in which a small, essential part of a workload remains operational in a recovery location. The name refers to a small flame that can ignite a larger system when needed. In cloud architectures, that essential core commonly includes replicated databases or a replication staging environment, while application servers and other production components are stopped or deployed only during recovery. The recovery environment is not normally ready to serve the full application without additional actions.

Start by defining the recovery time objective, or RTO, and recovery point objective, or RPO. RTO is the target maximum time to restore service after disruption. RPO expresses the acceptable amount of data loss measured in time. Pilot light can support a low RPO through ongoing replication, but asynchronous replication can still lose recent writes. Its RTO includes incident detection, recovery authorization, data promotion or recovery, infrastructure provisioning, application startup, validation, and traffic redirection. Neither objective is guaranteed simply by selecting this strategy.

For example, an online retailer might operate its production application in one cloud region and maintain a replicated database in another. Network definitions, machine images, application artifacts, configuration, and infrastructure templates are prepared for the recovery region. Application compute remains absent or stopped. During a regional outage, an automated workflow promotes the recovery database, launches application instances, configures their database endpoints, verifies checkout transactions, and redirects customer traffic. Database promotion and compute provisioning should follow a tested dependency order.

The distinction from other disaster recovery strategies is important. Backup and restore primarily rebuilds from stored recovery copies rather than maintaining an operational replicated core. Warm standby maintains a scaled-down but functional copy of the complete workload that can already handle traffic at reduced capacity. Active-active operates the workload in multiple locations serving production traffic. Pilot light generally costs less than warm standby because less infrastructure runs continuously, but it usually requires more recovery work and a longer RTO. These are relative trade-offs, not fixed recovery-time guarantees.

Infrastructure as code and recovery automation make pilot light practical. Engineers should prepare identity permissions, encryption-key access, network connectivity, secrets, certificates, software artifacts, and service quotas in the recovery location. A database replica alone does not make an application recoverable. Available compute capacity must also be considered: a quota increase permits resource creation but does not reserve physical capacity.

Replication is not a substitute for backups. Accidental deletion, corruption, or malicious changes can propagate into the recovery environment. Retain independently recoverable backups, use appropriate retention and protection controls, and test point-in-time recovery where supported. On AWS, AWS Backup can complement the replicated core with protected recovery points, but scheduled backup copies alone do not constitute a pilot light architecture.

Run recovery exercises that measure application-level results rather than just successful resource creation. Check replication lag, restore or promotion duration, capacity availability, dependency connectivity, and the effect of DNS caching on traffic cutover. Prevent simultaneous conflicting writers by isolating the original primary before enabling recovery writes. Finally, document failback: recovering in a second region does not automatically synchronize changes back to the original environment or make the return transition safe.

## Learn Pilot Light in lessons

8 short lessons on Pilot Light in 4 certification courses.

- SOA-C03: Pilot Light: activating the recovery environment (Premium), https://app.blowstack.com/lessons/pilot-light-activating-the-recovery-environment-soa-c03
- SAP-C03: Pilot Light: designing the recovery footprint (Premium), https://app.blowstack.com/lessons/pilot-light-designing-the-recovery-footprint-sap-c03
- SAA-C03: Pilot Light vs Warm Standby (Premium), https://app.blowstack.com/lessons/pilot-light-vs-warm-standby-saa-c03
- DOP-C02: Pilot Light: automated recovery activation (Premium), https://app.blowstack.com/lessons/pilot-light-automated-recovery-activation-dop-c02
- SOA-C03: Pilot Light and Warm Standby: recovery readiness (Premium), https://app.blowstack.com/lessons/pilot-light-and-warm-standby-recovery-readiness-soa-c03
- SAP-C03: Pilot Light vs. Warm Standby (Premium), https://app.blowstack.com/lessons/pilot-light-vs-warm-standby-sap-c03
