disaster-recovery – Other skill for AI agents

Disaster Recovery

Implement disaster recovery strategies including RTO/RPO planning, AWS cross-region failover patterns, DR testing procedures, and automated failover scripts.

RTO/RPO Planning

recovery_metrics:
  RTO:
    definition: "Recovery Time Objective - maximum acceptable downtime"
    measurement: "From the agreed service disruption point to verified recovery; record detection/declaration separately"
    factors:
      - Failover automation maturity
      - Data replication lag
      - DNS propagation time
      - Application warm-up time
      - Verification procedures

  RPO:
    definition: "Recovery Point Objective - maximum acceptable data loss"
    measurement: "Gap between latest recoverable consistent data and disruption; backups/replication may be stale or corrupted"
    factors:
      - Backup frequency
      - Replication method (sync vs. async)
      - Transaction log shipping interval
      - Cross-region replication lag

service_tier_targets:
  tier_1_critical:
    examples: "Authentication, payment processing, core API"
    rto: "< 15 minutes"
    rpo: "< 1 minute (near-zero)"
    strategy: "Multi-site active-active or warm standby"
    replication: "Synchronous or near-synchronous"
    testing: "Quarterly failover test"

  tier_2_essential:
    examples: "Customer dashboards, reporting, notifications"
    rto: "< 1 hour"
    rpo: "< 15 minutes"
    strategy: "Warm standby or pilot light"
    replication: "Asynchronous with short interval"
    testing: "Semi-annual failover test"

  tier_3_standard:
    examples: "Internal tools, analytics, batch processing"
    rto: "< 4 hours"
    rpo: "< 1 hour"
    strategy: "Pilot light or backup and restore"
    replication: "Periodic snapshots"
    testing: "Annual failover test"

  tier_4_non_essential:
    examples: "Development environments, documentation sites"
    rto: "< 24 hours"
    rpo: "< 24 hours"
    strategy: "Backup and restore"
    replication: "Daily backups"
    testing: "Annual backup restore verification"

DR Strategies Comparison

strategies:
  backup_and_restore:
    rto: "Hours"
    rpo: "Hours (depends on backup frequency)"
    cost: "$"
    description: "Regular backups stored in DR region. Restore from backup when needed."
    aws_services:
      - "S3 cross-region replication for backups"
      - "RDS automated snapshots copied to DR region"
      - "AMI copies in DR region"
      - "Terraform/CloudFormation for infrastructure rebuild"
    pros: "Lowest cost, simplest to maintain"
    cons: "Longest recovery time, highest data loss potential"

  pilot_light:
    rto: "Minutes to hours"
    rpo: "Minutes"
    cost: "$$"
    description: "Core infrastructure running in DR region (databases replicated). Scale up compute on failover."
    aws_services:
      - "RDS cross-region read replica (always running)"
      - "S3 cross-region replication"
      - "AMIs pre-built in DR region"
      - "Auto Scaling groups at zero/minimal capacity"
    pros: "Fast database recovery, moderate cost"
    cons: "Compute scale-up adds to recovery time"

  warm_standby:
    rto: "Minutes"
    rpo: "Seconds to minutes"
    cost: "$$$"
    description: "Scaled-down but functional environment in DR region. Scale up on failover."
    aws_services:
      - "RDS cross-region read replica"
      - "ECS/EKS running at reduced capacity"
      - "Route53 health checks for automated DNS failover"
      - "Global Accelerator for traffic management"
    pros: "Fast failover, reduced risk"
    cons: "Higher baseline cost for idle resources"

  multi_site_active:
    rto: "Near-zero"
    rpo: "Near-zero"
    cost: "$$$$"
    description: "Active-active across regions. Traffic served from both regions simultaneously."
    aws_services:
      - "DynamoDB Global Tables where appropriate; Aurora Global Database normally has one writer region, not multi-region active-active writes"
      - "Route53 latency/weighted routing"
      - "CloudFront with multi-origin"
      - "Global Accelerator"
      - "ECS/EKS in both regions"
    pros: "Minimal downtime and data loss"
    cons: "Highest cost, most complex to operate"

AWS Cross-Region DR Implementation

Fragments below provision or replace paid production configuration: confirm exact

account, engine/version support, target region, KMS/IAM grants, backups and approved

change window first. The listed tier timings/frequencies are planning examples,

not platform guarantees or universal compliance rules. Aurora secondary-cluster

creation also needs supported regional engine/version, subnet/security settings

and instances; a cluster object alone is not a working DR application.

S3 requires versioning on both buckets, replication-role/KMS permissions and

explicit handling of existing objects; put-bucket-replication replaces existing

rules, so merge a reviewed snapshot rather than overwriting it. Replication is not

an immutable backup and can replicate corruption/deletion. Use a private unique

temporary file instead of shared /tmp/replication-config.json when implementing.

AWS Cross-Region DR Implementation

# === Database Replication ===

# Create cross-region RDS read replica
aws rds create-db-instance-read-replica \
  --db-instance-identifier prod-db-dr-replica \
  --source-db-instance-identifier arn:aws:rds:us-east-1:123456789012:db:prod-db \
  --db-instance-class db.r6g.large \
  --region us-west-2 \
  --kms-key-id arn:aws:kms:us-west-2:123456789012:alias/rds-dr-key \
  --multi-az \
  --tags Key=Purpose,Value=DR Key=Environment,Value=production

# Create Aurora Global Database for near-zero RPO
aws rds create-global-cluster \
  --global-cluster-identifier prod-global-db \
  --source-db-cluster-identifier arn:aws:rds:us-east-1:123456789012:cluster:prod-aurora-cluster \
  --region us-east-1

# Add secondary region to Aurora Global Database
aws rds create-db-cluster \
  --db-cluster-identifier prod-aurora-dr \
  --global-cluster-identifier prod-global-db \
  --engine aurora-postgresql \
  --region us-west-2 \
  --kms-key-id arn:aws:kms:us-west-2:123456789012:alias/aurora-dr-key

# === Storage Replication ===

# S3 cross-region replication
cat > /tmp/replication-config.json << 'EOF'
{
  "Role": "arn:aws:iam::123456789012:role/s3-replication-role",
  "Rules": [
    {
      "ID": "ReplicateAll",
      "Status": "Enabled",
      "Filter": {"Prefix": ""},
      "Destination": {
        "Bucket": "arn:aws:s3:::prod-data-dr-usw2",
        "StorageClass": "STANDARD",
        "EncryptionConfiguration": {
          "ReplicaKmsKeyID": "arn:aws:kms:us-west-2:123456789012:alias/s3-dr-key"
        }
      },
      "DeleteMarkerReplication": {"Status": "Disabled"},
      "Priority": 1,
      "SourceSelectionCriteria": {"SseKmsEncryptedObjects": {"Status": "Enabled"}}
    }
  ]
}
EOF

aws s3api put-bucket-replication \
  --bucket prod-data-use1 \
  --replication-configuration file:///tmp/replication-config.json

# === DNS Failover ===

# Route53 health check for primary region
aws route53 create-health-check --caller-reference "prod-health-$(date +%s)" \
  --health-check-config '{
    "Type": "HTTPS",
    "FullyQualifiedDomainName": "primary-health.example.com",
    "Port": 443,
    "ResourcePath": "/health",
    "RequestInterval": 10,
    "FailureThreshold": 3,
    "EnableSNI": true
  }'

# Configure failover routing
aws route53 change-resource-record-sets --hosted-zone-id Z123456 \
  --change-batch '{
    "Changes": [
      {
        "Action": "UPSERT",
        "ResourceRecordSet": {
          "Name": "api.example.com",
          "Type": "A",
          "SetIdentifier": "primary",
          "Failover": "PRIMARY",
          "AliasTarget": {
            "HostedZoneId": "Z1234PRIMARY",
            "DNSName": "primary-alb.us-east-1.elb.amazonaws.com",
            "EvaluateTargetHealth": true
          },
          "HealthCheckId": "health-check-id-primary"
        }
      },
      {
        "Action": "UPSERT",
        "ResourceRecordSet": {
          "Name": "api.example.com",
          "Type": "A",
          "SetIdentifier": "secondary",
          "Failover": "SECONDARY",
          "AliasTarget": {
            "HostedZoneId": "Z5678SECONDARY",
            "DNSName": "dr-alb.us-west-2.elb.amazonaws.com",
            "EvaluateTargetHealth": true
          }
        }
      }
    ]
  }'

Failover runbook and worked drill

The former executable script promoted a replica and declared success after a

health URL without fencing the primary, updating app DB credentials or verifying

routing/data. Do not automate that sequence unchanged. Use an owner-approved,

service-specific runbook with explicit incident authority and rollback limits:

1. Record disruption/declaration times, intended account/regions and last verified

recoverable transaction/replication lag. Obtain the loss acceptance decision.

2. Fence the old writer and producers, including jobs/webhooks, and verify fencing;

a network partition can

… (full content is in the download package)