Practice Exams:

Microsoft AZ-104: Azure Backup and Recovery Design

Cloud & Architecture

A useful Azure recovery design starts with what the business must restore, how much data it can lose and who can authorize recovery. Selecting a vault is one important step, but it does not establish that a damaged application can return to service. Restoring a database copy is not the same as restoring a working business service. Document which dependencies return first, how access is re-established, what the recovery objective means, and how a failed recovery drill changes the plan.

On this page
  1. Define recovery outcomes before choosing backup settings
  2. Match the protected workload to the correct Azure Backup capability
  3. Separate backup retention from recoverability
  4. Resist destructive changes to recovery infrastructure
  5. Plan regional loss independently of routine restore
  6. Run a restore drill that verifies the application, not only the wizard
  7. Operate the protection estate as workloads change

Define recovery outcomes before choosing backup settings

An application owner might ask for “daily backup,” yet the real requirement could be to recover a deleted record within an hour, restore a virtual machine after a failed patch, or resume service after an entire region becomes unavailable. Those are different failure scenarios. Document the data that must survive, acceptable data loss and tolerable interruption for each service rather than using a single default policy for everything. Reliable backup planning in Microsoft AZ-104 requires a testable restore path and recovery evidence rather than relying on a successful backup-job status.

A recovery point objective (RPO) expresses the acceptable amount of data loss measured in time. A recovery time objective (RTO) expresses the target time to restore service. Both have to include application dependencies: identity, networking, encryption keys, DNS records, configuration and the people authorized to invoke recovery. A restored VM that cannot authenticate to its database has not met the recovery objective.

Classify recovery events separately: accidental deletion, data corruption, ransomware, zone failure, regional disruption and the loss of an administrator account. For each event, record the expected recovery mechanism and proof required. This prevents the common assumption that one scheduled backup automatically covers all types of outage.

Match the protected workload to the correct Azure Backup capability

Azure Backup protection varies by workload, vault type, region and supported policy. A VM backup policy is not interchangeable with a policy for Azure Files or a database running inside a VM. Before selecting a policy, check whether application-consistent recovery is supported, which disks or databases are included and which restore paths are available for that protected resource.

The current VM Enhanced backup policy offers capabilities that differ from Standard policy, including more frequent supported backup schedules and support for newer disk scenarios. The choice should be driven by the VM configuration and RPO requirements, not by assuming Enhanced is always appropriate for every resource. Verify current frequency, retention and restore limitations in official guidance.

Recovery Services vaults and Backup vaults support different sets of workloads and operational models. Include the approved vault placement, ownership, retention, cost center and the resources it protects in the infrastructure manifest. A diagram should not imply that vault creation itself turns on protection; record each protection instance, policy assignment and last successful recovery point.

Separate backup retention from recoverability

Retention is an obligation to keep certain recovery points; recoverability is the ability to restore usable data from those points. Long retention is valuable for audit or delayed corruption detection, but old data may depend on applications, keys or storage formats that no longer exist. A policy should specify operational snapshots, longer-term retention and legal or regulatory requirements distinctly.

For Azure VMs, inspect the relationship between instantaneous restore snapshots and longer-term vault recovery points. Confirm recovery-point age, location, storage redundancy and restore mode. Redundancy settings affect resilience and cost, but cannot be changed freely in every protected-resource state. Evaluate regional recovery needs before protection begins and test the exact restore workflow the service owner expects to use.

Backup health should distinguish scheduled success from data protection gaps. A job that has not failed may still protect the wrong machine, omit a newly added workload or retain fewer points than the owner requires. Periodic inventory reconciliation should compare the list of production resources against those with valid policies and verifiable recent recovery points.

Resist destructive changes to recovery infrastructure

An attacker with broad administrative privileges may try to delete backup data, weaken retention or disable protection before encrypting production systems. Protect vault management as a sensitive security function. Azure Backup includes security features such as soft delete, immutable vault settings and multi-user authorization for supported operations, but these have different scopes and eligibility conditions.

Multi-user authorization uses a separate Resource Guard approval boundary for critical backup operations where supported. Microsoft’s multi-user authorization guidance explains supported safeguards and configuration requirements. A properly separated Resource Guard owner can limit the effect of a compromised ordinary backup administrator; it cannot compensate for collusion or poorly governed privileged access.

Use least-privilege Azure RBAC, monitored administrative actions, controlled break-glass access and appropriate retention safeguards together. A delete lock is not a substitute for a workload recovery test, and a vault security setting does not guarantee your administrators can recover lost encryption material. Retain emergency procedure access without turning routine operations into permanent superuser privileges.

Plan regional loss independently of routine restore

A protected VM can fail within a healthy region, or the entire region can become unavailable. Routine single-resource restore may work for the first situation but not the second. Identify whether the workload needs cross-region restore, replication, alternate-region infrastructure or another business-continuity arrangement. Restore support depends on selected storage redundancy, vault configuration, region pairs and workload capabilities.

Recovery regions require more than saved VM disks. If the alternate environment lacks virtual networks, DNS delegation, usable identities, firewall allowances, certificates or dependent databases, data restoration cannot bring the service back. Define an application dependency graph and the order in which recovery actions must happen. A different disaster-recovery mechanism, such as the replication scenarios described in Azure Site Recovery, may be needed when target recovery times cannot be met by rebuilding from backup alone.

For a recovery design involving multiple subscriptions, specify who can create resources in the target subscription during a crisis. Test capacity and quotas rather than assuming a region will accept arbitrarily large emergency deployments. Failback is a separate design: decide how users return to the normal service, how divergent writes reconcile and which evidence proves the original environment is safe to resume.

Run a restore drill that verifies the application, not only the wizard

A real restore test begins with a selected recovery point and a written objective. Use an isolated test resource group or equivalent supported procedure where practical. Record when the restore began, the storage and network prerequisites, and when the restored workload passed an application-level test. Keep the test from overwriting or exposing live customer information.

A useful drill might restore a designated VM and confirm that it boots, its disks contain the expected data, the configured application service starts, and an authorized test user can complete a transaction. A database drill should include consistency and the required recent records. Compare actual data age to RPO and actual end-to-end recovery duration to RTO. An Azure job status of “completed” is only one piece of that evidence.

Make failures actionable. If the restore takes too long because network permissions are missing, assign a network runbook owner. If the archived backup cannot be decrypted, fix the key-handling plan. If the policy protects a discarded VM while a replacement VM remains unprotected, repair the onboarding control. Retry the drill after remediation rather than marking the original failure accepted.

Operate the protection estate as workloads change

New deployments, VM replacements, resource moves and application changes can break assumptions made when protection was first enabled. Automate workload discovery and baseline policy assignment where supported, then verify with daily exception reporting. Avoid simple metrics such as “100% of backup jobs succeeded” unless the denominator reflects every intended protected workload.

Alerting should cover failed jobs, stopped protection, older-than-required recovery points, suspicious policy changes and untested restore paths. Assign ownership to the service team that understands business recovery requirements rather than treating backup as a storage-only responsibility. Test at a frequency proportional to application impact and change rate; a static test performed before a major database migration no longer proves the new configuration.

The proof of a recovery design is a repeated restore test with an owner, a documented recovery point, a measured time to service, and explicit checks of application correctness. Include permissions, dependent identities, network resolution, and recovery credentials in the drill. A successful backup job is necessary evidence, but the actual business promise is that the workload can return to a usable state.

Continue learning

Related guides

Azure Backup vs. Site RecoveryDistinguish data restoration from workload replication and failover.Azure Storage Account Design DecisionsChoose storage redundancy, access rules, and lifecycle settings.Azure Availability Zones and Failure DomainsAssess zone failures, regional placement, and workload resilience.

Related Posts

• Microsoft AZ-104: Is Azure Administration the Right Career?

• Microsoft AZ-104 Exam Guide: Skills, Labs, and Preparation

• How Difficult Is the Microsoft AZ-104 Exam?

• Microsoft AZ-104 vs. AZ-103: What Changed?

• Cloud Strategy Unlocked: How AZ-305 Shapes Real Azure Architects

• Microsoft AZ-104: Troubleshoot VNets, Routes, and DNS

• Microsoft AZ-104: Entra Roles and Access

• Microsoft AZ-900: Cloud Economics: CapEx, OpEx, and the Cost of Commitment

• Microsoft AZ-104: Azure Well-Architected Operations

• Microsoft AZ-104: Zones and Failure Domains