VETERINARY HEALTHCARE / AWS RESILIENCE
Practical Disaster Recovery on AWS for a Write-Heavy Platform
Ficode configured active-passive recovery orchestration spanning health checks, standby validation, data-service promotion, application checks and AWS Global Accelerator switching. A live end-to-end drill has not yet been performed.
The project in context
A platform receiving continuous clinical updates cannot recover through server restarts alone. Ficode configured a guarded sequence to check the primary region, confirm a failure, examine standby readiness, promote relevant data services, validate the application, shift global traffic and record the action. The design remains unproven in a live full-system drill.
Data, applications and traffic had to recover in the right order.
Clinical case updates, files, notifications and sessions place dependencies across databases, caches and application services. A regional outage therefore requires more than a simple redirect; the standby must be prepared before users are sent to it.
The challenge Before
Regional outage
Primary failure needs confirmationState and application
Data and service readiness interdependentTraffic switch
Users must not reach unready standbyWith Ficode Connected Approach
Guarded detection
One-minute checks; eight failures; recheckReadiness gates
Standby, data promotion and applicationControlled routing
Global Accelerator dials; notify and logConfigure failover as an ordered sequence with validation gates
Monitor and confirm
Check health each minute; eight consecutive failed checks trigger re-confirmation of the outage.
Prepare data services
Validate APAC standby and promote DocumentDB, RDS for PostgreSQL and Redis / ElastiCache.
Validate then switch
Check endpoints and load-balancer target groups before changing Global Accelerator traffic dials.
Notify and record
Send notifications through SNS and store workflow state in DynamoDB; supporting artefacts use S3.
From confirmed regional failure to a configured traffic switch
Detect and confirm
Health checks, threshold and second checkPrepare standby
Check capacity; promote data; validate appSwitch and record
Global Accelerator dial; SNS; DynamoDBCONNECTED OUTCOME
The solution incorporated an eight-step automated recovery workflow designed to support a structured and consistent recovery process.
A defined recovery design, with explicit testing limits
One-minute monitoring
Primary-region health is checked at one-minute intervals.
Eight failed checks
Consecutive failures trigger the guarded recovery process.
Active-passive regions
European production and an APAC warm standby avoid a multi-active claim.
Recovery targets, not results
RTO up to 60 minutes and RPO up to 15 minutes are design objectives only.
Confirmed technology and implementation scope
Amazon CloudWatch; Amazon EventBridge; AWS Lambda; AWS Step Functions; Amazon DocumentDB; Amazon RDS for PostgreSQL; Amazon ElastiCache / Redis; AWS Global Accelerator; Amazon SNS; Amazon DynamoDB; Amazon S3.
