Skip to Content
Cloud Engineering Academy
12. High Availability & Disaster Recovery

Module 12 cheat sheet

One-page revision summary for High Availability & Disaster Recovery: services, Azure mapping, CLI commands, limits, decision rules, exam traps and cost traps.

SAA-C03SOA-C03DOP-C02SAP-C03

Assumes: AWS CLI v2 · AWS SDK for .NET v4 · Microsoft.Extensions.Http.Resilience · Region us-east-1 (DR us-west-2) · facts as stated in the Module 12 lessons; items marked VERIFY there are not repeated as certain here. Verified against official sources on 2026-11-18.

A printable summary of Module 12. Each row links back to the lesson with the detail.

Key services

ServiceWhat it is for
EC2 Auto Scaling / Application Auto ScalingKeep capacity between a minimum and maximum (EC2, ECS tasks, Lambda provisioned concurrency, DynamoDB, Aurora replicas)
Elastic Load Balancing (ALB/NLB)Health-checked traffic spreading across AZs, connection draining
RDS Multi-AZ / AuroraManaged in-Region failover; Aurora Global Database for cross-Region
DynamoDB global tablesMulti-Region replicas: MREC (async, last writer wins) or MRSC (sync, zero RPO)
Route 53 + health checksDNS-based failover, weighted and latency routing
Application Recovery ControllerZonal shift (AZ evacuation), routing controls (reliable manual Regional switch)
AWS BackupPlans, vaults, Vault Lock, cross-Region and cross-account copies, restore testing
CloudFront / ElastiCacheShield back ends from load and partial outages; never cache personal data carelessly
Fault Injection Service (FIS)Controlled chaos with CloudWatch-alarm stop conditions
Resilience Hub / Well-Architected ToolPolicy-based assessment against RTO/RPO; structured reviews

.NET / Azure → AWS

Microsoft / AzureAWS
Availability ZonesAvailability Zones (you spread EC2/ECS yourself)
VM Scale Sets / App Service scale-outEC2 ASG / ECS service auto scaling
Traffic Manager / Front DoorRoute 53 health checks + failover routing / CloudFront
Azure SQL geo-replicationAurora Global Database, RDS cross-Region replica
Cosmos DB multi-region writesDynamoDB global tables
Azure Backup vaultAWS Backup vault + Vault Lock
Azure Site RecoveryElastic Disaster Recovery, pilot light / warm standby
Azure Chaos StudioFIS experiment template
PollyMicrosoft.Extensions.Http.Resilience pipelines

Most useful CLI commands

TaskCommand
FIS: start / inspect / stopaws fis start-experiment --experiment-template-id ID · aws fis get-experiment --id ID · aws fis stop-experiment --id ID
Create a backup vaultaws backup create-backup-vault --backup-vault-name NAME
Lock a vaultaws backup put-backup-vault-lock-configuration --backup-vault-name NAME ...
Restore testing planaws backup create-restore-testing-plan --restore-testing-plan file://rtplan.json
Check restore jobsaws backup list-restore-jobs --by-status COMPLETED
Resilience Hub assessmentaws resiliencehub start-app-assessment ... · aws resiliencehub describe-app-assessment ...
Anything left over?aws resourcegroupstaggingapi get-resources --tag-filters Key=Project,Values=cea-lab

Key limits and defaults

ItemValueLesson
99.9% / 99.95% / 99.99% per 30-day month43.2 / 21.6 / 4.3 minutes allowed downtimeAvailability vocabulary
Series availabilityMultiply the links; the chain is worse than its weakest linkAvailability vocabulary
Capacity to survive losing one AZN x P / (N - 1): 3 AZs need 150% of peak, 2 AZs need 200%Auto Scaling everywhere
ALB health check defaults30 s interval, 5 s timeout, 2 failures to mark unhealthy, 5 successes to recoverELB for resilience
Deregistration delay default300 secondsELB for resilience
Cross-zone defaultsALB on; NLB off by defaultELB for resilience
Route 53 health check30 s standard or 10 s fast; unhealthy when 18% or fewer checkers report healthyFailover with Route 53
Worst-case DNS failoverCheck interval x threshold + TTL (30 s x 3 + 60 s = 150 s)Failover with Route 53
Aurora Global rds.global_db_rpoMinimum 20 seconds; blocks commits when lag exceeds itDatabase failover
DynamoDB MRSCExactly three Regions (or two plus witness); no transactions, TTL or LSIsDatabase failover
Vault Lock compliance modeGrace time minimum 3 days, then immutableAWS Backup in depth
Vault retention boundsMinimum at least 1 day; maximum up to 36,500 daysAWS Backup in depth
Restore testing quotas100 plans per account, 30 selections per planAWS Backup in depth

Decision rules

If…Then…
Surviving an AZ loss mattersThree AZs, minimum capacity already running (static stability); do not rely on scale-out during the event
Failure is transientTimeout, retry with backoff and jitter inside a retry budget, then a circuit breaker
RTO hours, RPO hoursBackup and restore
RTO tens of minutes, data replicated, compute offPilot light
RTO minutes, scaled-down copy runningWarm standby
RTO near zero, highest costMulti-site active-active
Ledger needing exactly-once debits and creditsSingle writer with a hot or warm second Region; MREC last-writer-wins is unsafe
Planned Region rotation or failback (Aurora Global)Switchover (RPO zero)
Real Regional outage (Aurora Global)Failover with data-loss acknowledgement
Human-decided Regional failoverRouting control or scripted runbook, not a probe that can flip back
Backups must resist deletionSeparate backup account, Vault Lock, cross-Region copy
First chaos experimentSandbox, one target, alarm stop condition, alarm currently OK

Common exam traps

  • Multi-AZ is not a backup and not DR: a bad delete replicates to the standby.
  • RDS Multi-AZ standby does not serve reads in the classic single-standby setup; read replicas scale reads. Aurora readers do both.
  • Cross-Region read replica promotion is one-way.
  • Route 53 failover time is a lower bound: resolvers and long-lived connections ignore TTLs.
  • Route 53 HTTPS health checks do not validate certificates.
  • Zonal shift moves traffic away from an AZ for a limited time and needs spare capacity; it is not a Region failover.
  • Reserved concurrency is both a cap and a guarantee for one Lambda function.
  • Vault Lock compliance mode becomes irreversible after the grace time.
  • A successful backup is not a proven restore: test restores.
  • ARC Readiness Check is closed to new customers; routing controls and zonal shift are separate capabilities.
  • Pilot light cannot take traffic without extra steps; warm standby can, at reduced capacity.

Cost traps

TrapAvoid it by
Idle DR capacity forgotten in the quarterly reviewPrice it with a cost model; right-size warm standby
Multi-Region data stores (Aurora Global, global tables)Design-only or minutes-long lab; delete after
NAT gateway per AZ and load balancers left runningTear down the same day
ARC clusters and fast health checksBilled per hour or per check; keep labs to minutes
Cross-zone and cross-Region data transferEstimate the traffic; keep chatty services AZ-aware
Over-high autoscaling maximumCap from the database connection budget; alarm when at maximum
Backups kept forever or copied everywhereLifecycle rules, tiering, retention from records policy
ElastiCache cluster left idleUse local Valkey in Docker for learning

Independent learning resource — not affiliated with, endorsed or sponsored by Amazon Web Services. AWS and related marks are trademarks of Amazon.com, Inc. or its affiliates. All banking data in examples is fictional. Sample architectures are for learning and are not automatically compliant with GDPR, PCI DSS or any banking regulation.

On this page