Microsoft AZ-104: Understanding the Objective Groups

Microsoft’s AZ-104 objectives are organized into five domains, but an administrator rarely works on just one domain at a time. Creating a compute resource can involve subscription scope, permissions, virtual networking, storage and monitoring. A storage-access problem can turn out to be a private DNS issue; a healthy VM can host a failed application; an approved management change can break a security requirement. Understanding the relationship between objective groups matters because operational work crosses their boundaries.

This article follows a fictional city transit information service through one change lifecycle to show how the skills connect. It complements the current Microsoft AZ-104 Exam blueprint but does not recite every objective or prescribe a personal study schedule. Its purpose is to explain the combined administrative responsibility: plan an allowed change, deploy it, validate users’ actual experience, observe failures and restore an acceptable state.

Begin with the change and its business constraints

The transit agency runs a public bus-arrival application. Its backend reads vehicle telemetry from a message-processing component, stores service records and serves arrival predictions. A new change will add a staff-only delay-reason field for operations personnel without making personal employee data accessible to the public.

The requirement appears simple: update an API and store an additional field. Yet it immediately raises several administrative decisions. Which team may deploy it? Which identity reads and writes the data? Will public users receive any protected field? Which network paths should remain private? What signal will reveal that the new version is returning incorrect data or making requests slower?

Write those answers before configuring the first resource. The objective domains make more sense when the learner knows what normal behavior should be and which failure would matter. A clean resource deployment is not automatically evidence that the change was appropriate.

Identity and governance define the permitted work

Manage Azure identities and governance is the first published AZ-104 group. It includes managing Microsoft Entra identities and access, configuring role-based permissions, and implementing relevant subscription and resource governance. For the transit agency, the deployment identity should be able to change the intended application resources without receiving unlimited authority over unrelated systems.

Identify management permissions separately from the application identities used at runtime. An engineer’s Azure role may allow deployment, while a managed workload identity may need a narrow data-plane permission to read vehicle information. Authentication confirms a principal; authorization decides which operation it can perform. Azure Policy can constrain configuration even when the principal has management permission.

If a deployment fails, compare role scope, policy assignment and requested resource properties rather than simply granting more authority. The Azure Resource Manager hierarchy provides the context for that control-plane decision. A broadly privileged account is not a reliable fix for a violation of the agency’s security standards.

Storage connects data requirements to access and durability

Implement and manage storage includes storage configuration, security, access and relevant data services. The delay-reason field might live in a relational operational store, while archival events or exported reports could use a different storage model. A candidate should determine how the data is accessed, how long it is retained, what must be recoverable and which identities are permitted to read it.

Suppose the backend now needs to retrieve a protected blob containing a staff-only service bulletin. A correct data role may be insufficient if the application cannot reach the storage endpoint, and a correct network path does not create data permission. A test should include one authorized read and an explicitly prohibited read, using synthetic content rather than actual staff records.

Azure storage decisions also interact with cost and continuity. A redundant copy may improve certain failure tolerance without providing the historical state needed after accidental deletion. Retention and restoration requirements should follow the agency’s real records obligations rather than the most impressive service setting available.

Compute is where intended configuration meets runtime behavior

Deploy and manage Azure compute resources includes virtual machines, hosted application services, container-related options and deployment or operating tasks. The agency must choose a compute model supported by its team and ensure the runtime can authenticate, reach its dependencies and return the correct data.

A deployment can succeed while a process fails to start or an updated application no longer recognizes an expected configuration. The administrator should verify the version, environment settings, service state and one end-to-end request after release. A VM showing “running” is evidence about compute state, not proof that bus arrivals are accurate.

Consider the difference between managing a VM guest and deploying to a managed application platform. A virtual machine places additional operating-system and patching responsibility on the agency. A managed platform may remove parts of that work while changing how networking, configuration and scaling are handled. AZ-104 requires understanding the practical consequences rather than assuming one model is universally preferable.

Repeatable deployment can be documented in appropriate code or templates, with sensitive credentials managed separately. The model should describe resources and properties but not pretend to guarantee application correctness by itself.

Virtual networking controls reachability, not data ownership

Implement and manage virtual networking includes address spaces, network relationships, security controls and related connectivity. The bus-arrival API may be public, while its database and staff records should be accessible only through approved paths. The admin must understand source, destination, routing, DNS and network filtering without confusing those mechanisms with resource-specific permissions.

For a backend read, trace whether the application resolves the correct private endpoint, whether traffic can follow the intended route and whether filtering permits the required protocol. An NSG can deny traffic, but it cannot determine whether a signed-in transit employee may read another person’s personnel file. That entitlement remains a separate data or application authorization decision.

Use Azure network security groups for narrowly scoped allow/deny examples. Validate both the flow the service needs and one flow it must reject. If an endpoint is unreachable, testing from an administrator’s workstation does not prove the application subnet has the same DNS or routing behavior.

When a service spans multiple networks, a network specialist may own deeper design choices. AZ-104 still requires enough understanding to operate the configured path, recognize a boundary problem and collect evidence before escalating.

Monitoring should confirm the public service is actually useful

Monitor and maintain Azure resources includes resource health, monitoring, alerting and recovery operations. The agency might have low CPU usage and no obvious VM failures while customers see stale or incorrect arrival times. Resource metrics alone cannot prove that the full service transaction is correct.

Define a synthetic query for a known transit route and an expected response contract. Track API response failures, dependency errors and the age of underlying telemetry. If the application continues serving yesterday’s predictions, a green infrastructure status may be misleading even though the system is technically running.

Azure Monitor can provide metrics, logs and alerts, but a useful signal needs a response owner and an interpretation. A failed read might be caused by authentication, authorization, network access or application behavior. Correlation identifies where to investigate, not necessarily a complete causal proof.

Choose alert thresholds with real consequences in mind. A temporary processing delay may call for warning and investigation; a sustained loss of live arrivals may warrant escalation or a fallback service message. An alert with no defined responder adds noise without improving availability.

Maintenance and recovery revisit every earlier boundary

Suppose the change deploys and users begin seeing missing arrivals. The agency can revert to the previous application version, but a schema change or asynchronous data migration may also need attention. An effective rollback cannot consist solely of switching one compute image if stateful dependencies are no longer compatible.

Before recovery, identify the latest trustworthy data state, which configuration changed, the authentication and networking requirements for restored services and how the team will verify data correctness. Azure Backup is relevant when a recoverable point is required, while redundancy and scaling protect other failure models. Repeating a backup job does not prove the service can restore within the agency’s agreed downtime.

The practical test is a small, authorized rollback or restoration using synthetic state. Verify that public arrivals work again, staff-only records remain protected and monitoring reflects the restored condition. That exercise touches all five domains because the workload is a connected system.

Review the lifecycle as a single administrator decision

Map the change into stages: authorize the actor and configuration, select the data model and permissions, deploy the compute resources, verify the network path, monitor the business transaction and demonstrate recovery. For every stage, identify one failure that might occur while the other stages remain apparently healthy.

This produces a stronger understanding of AZ-104 than memorizing five unrelated headings. A candidate should be able to tell whether the problem belongs to role scope, policy, data-plane access, route, runtime, telemetry or restoration—and recognize when several need to be examined together.

The official blueprint defines the coverage boundary. The working service gives it meaning. Azure administration is the practice of maintaining the intended behavior across these connected controls, not merely completing a separate checklist for each domain.