Home / Azure

AzureChecklistIntermediate

Planning and running an Azure DR drill: RTO/RPO tiers, Front Door failover and evidence

Classify workloads into DR tiers, pick the matching Azure patterns, then plan, run and document a DR drill end to end, including traffic switching with Front Door and failback.

Deepak Bhardwaj3 min read

Applies toAzure workloads using Site Recovery, SQL failover groups, GZRS storage, Front Door / Traffic Manager

1. Agree DR tiers with the business

Tier RTO RPO Azure pattern
Tier 0 < 15 min ~0–5 min Active-active or hot standby (Front Door, failover groups, zone + region redundancy)
Tier 1 < 4 h < 15 min Warm standby (ASR for VMs, failover groups, GZRS, small standby app tier)
Tier 2 < 24 h < 24 h Backup and restore (Azure Backup with Cross Region Restore, IaC redeploy)
Tier 3 Best effort Last backup Backups only

Tag resources with drTier and keep a list of dependencies per application. Most failed drills fail on a forgotten dependency (DNS, a licence server, a certificate).

2. Before the drill

  • Scope: which applications, test or real failover, production or DR copy.
  • Runbook per application, with owners and timings.
  • Communication plan and change approval.
  • Rollback criteria ("if X isn't working by T+2h, abort and fail back").
  • Monitoring and alerting work in the DR region.
  • Access works in the DR region (Bastion, VPN, PIM roles).

3. Traffic switching with Front Door

Origins: primary region priority 1, DR region priority 2. To fail over on demand:

az afd origin update -g rg-edge --profile-name afd-web --origin-group-name og-web -n cin --enabled-state Disabled

Front Door then sends traffic to the priority-2 origin. Re-enable to fail back. With Traffic Manager, use priority routing and disable the primary endpoint the same way.

4. During the drill, record

Time Step Result Issue
10:00 Declare drill start
10:04 SQL failover group planned failover OK, 2 min
10:10 ASR recovery plan failover OK, 22 min App VM needed manual IP change
10:35 Front Door switch OK Cache purge required
10:40 Smoke tests Partial Batch job pointed to old storage endpoint

Measured RTO = time from start to "customers can use it". Measured RPO = last transaction available after failover.

5. Failback

Plan it as carefully as the failover: re-protect, planned failover back, DNS/Front Door switch, re-enable replication in the original direction.

6. After the drill

  • Report with measured RTO/RPO vs targets.
  • Issues turned into backlog items with owners.
  • Runbooks updated.
  • Next drill date set (at least yearly; Tier 0/1 twice a year).

Tip

Start with a drill of one Tier 2 application. Teams learn the process with low risk, and the customer sees progress early instead of a big-bang drill that fails.

Related fixes