1 · DR Fundamentals: RTO vs RPO
RTO (Recovery Time Objective) — max tolerable downtime. RPO (Recovery Point Objective) — max acceptable data loss (time since last good backup/replication).
Cost increases as RTO/RPO decrease. The architect's job: match tier to business criticality.
2 · Azure Site Recovery (ASR)
ASR provides continuous replication, automated failover, and failback for:
- Azure-to-Azure — replicate VMs between regions (no additional infra).
- On-premises-to-Azure — Hyper-V, VMware, or physical servers via process/config servers.
- On-premises-to-on-premises — secondary datacenter (VMware/Hyper-V).
ASR replicates at the block level with crash-consistent and app-consistent recovery points every 5 min / 1–4 hr respectively.
3 · ASR Architecture & Workflow
Key Concepts
- Recovery Vault — stores replication metadata & recovery points.
- Replication Policy — RPO threshold, retention, app-consistent frequency.
- Recovery Plan — groups VMs into ordered failover steps with pre/post scripts.
- Test Failover — validates DR in isolated VNet without impacting production.
- Failback — reverse-replicate from target back to source after issue resolved.
4 · Multi-Region DR Patterns
- Paired Regions — Azure pairs regions 300+ miles apart; platform updates staggered. Prefer paired regions for ASR targets.
- Traffic Manager / Front Door — priority or weighted routing with health probes auto-redirects traffic on failure.
- Cross-region Load Balancer — global tier distributes across regional backends with instant failover.
- Availability Zones first — for HA within region; DR adds cross-region protection for regional outages.
5 · DR for PaaS Services
| Service | DR Mechanism | RPO |
|---|---|---|
| Azure SQL | Active Geo-Replication / Auto-failover groups | < 5 sec |
| Cosmos DB | Multi-region writes, automatic failover | ~0 (strong consistency) to seconds |
| Storage (GRS/GZRS) | Async replication to paired region; customer-initiated failover | ~15 min |
| App Service | Deploy to secondary region + Traffic Manager | Near-zero with slot swap |
| AKS | Multi-cluster with Azure Front Door; GitOps redeploy | Minutes (depends on state) |
6 · DR Testing & Drills
- Test Failover (ASR) — spins up replicas in isolated VNet; no production impact. Validate app connectivity, DNS, and data integrity.
- Planned Failover — zero data-loss failover for planned maintenance; requires source still running.
- DR Drill Schedule — quarterly minimum; document results, fix gaps, update runbooks.
- Chaos Engineering — Azure Chaos Studio can inject region-down faults to validate end-to-end DR.
7 · DR Automation & Recovery Plans
Recovery Plans orchestrate multi-VM failover with:
- Groups — order VMs (DB tier → app tier → web tier).
- Pre-actions — Azure Automation runbooks (e.g., start dependencies, update DNS).
- Post-actions — health checks, notification via Logic Apps or webhooks.
- Manual steps — pause for human approval if needed.
Combine with Azure Policy to enforce ASR enablement on all production VMs ("deployIfNotExists").
8 · Real-World: Tiered Enterprise DR Strategy
🏢 Enterprise Scenario
A financial services company classifies 200+ workloads into three tiers. Tier-1 (trading platform) runs hot with multi-region Cosmos DB and SQL failover groups. Tier-2 (internal apps) uses ASR with 5-minute RPO. Tier-3 (dev/test) relies on daily backups and IaC redeployment. This saves 60% vs. running everything as hot standby while meeting each SLA.
📝 AZ-305 Exam Tip
Know when to recommend ASR vs. PaaS-native DR. ASR is for IaaS VMs; for SQL Database use auto-failover groups, for Cosmos DB use multi-region writes. The exam tests whether you pick the right DR mechanism per service type and can calculate RTO/RPO alignment with business requirements.
✅ Knowledge Check
1. A company needs RPO < 5 seconds for their SQL Database. Which DR feature should they use?
2. What is the primary benefit of ASR test failover?
3. Which DR tier offers the lowest RTO but highest cost?
4. In an ASR Recovery Plan, what ensures the database VMs start before application VMs?