Microsoft AZ-104: Azure Arc and Hybrid Windows Server Management
A Windows Server does not have to run in Azure to appear in an Azure management inventory. Azure Arc-enabled servers register supported on-premises or other-cloud machines as Azure resources while the operating system and its applications remain where they are. The operational challenge is deciding what Azure should manage, what stays under local server administration and how to keep the two control planes from contradicting one another.
On this page
- Decide what Arc changes—and what it does not
- Verify prerequisites before deploying the Connected Machine agent
- Separate Azure permissions from Windows permissions
- Make patch governance a complete workflow
- Deploy extensions with an owner and a recovery plan
- Use Windows Admin Center with its actual support boundary
- Apply configuration and monitoring policy to the right machines
- Build a hybrid pilot around one recoverable change
- Keep hybrid management ownership explicit
Decide what Arc changes—and what it does not
The Azure Arc-enabled servers service uses the Connected Machine agent to register a physical or virtual machine in Azure. The Azure resource represents the machine for management, policy, extensions and inventory workflows. Onboarding does not migrate the server into an Azure virtual machine, replace Active Directory, change its network route or automatically make every Azure management feature available. This distinction matters when a team wants one inventory and update process across on-premises Windows Server, hosted virtual machines and Azure VMs. Hybrid server management in Microsoft AZ-104 starts by locating the management plane, identity path and recovery boundary before deploying an agent.
Arc can provide a consistent Azure view of supported server assets, but the local operating system still depends on its own patch, identity, network and application infrastructure. The team should decide which system owns an operation before enabling it in both places.
Start with a clear outcome: inventory the machines, verify patch compliance, deploy a monitoring extension, enforce a supported configuration policy or provide a controlled administrative access path. “Put servers in Azure Arc” is an onboarding activity, not a finished management objective.
Verify prerequisites before deploying the Connected Machine agent
A successful onboarding plan needs an Azure subscription, required resource providers, permission to register machines and local administrative access to install and configure the agent. Microsoft distinguishes the Azure Connected Machine Onboarding role used for enrollment from the resource-administration roles used later to manage registered machines. Use separate identities when different teams perform those responsibilities.
The server must meet the Connected Machine agent prerequisites, including a supported operating system and the Azure endpoints required for its network design. Microsoft documents direct connectivity, approved proxies, Azure Arc Gateway and supported private connectivity options. An Azure resource entry alone cannot make an isolated server manageable if the agent has no permitted connection path.
Plan registration names, Azure region, resource groups and tags before onboarding hundreds of machines. Those decisions affect inventory searches, governance scope and incident ownership. A server name copied from a temporary deployment convention can become a confusing identifier when another machine with the same short name is later registered.
Separate Azure permissions from Windows permissions
Arc creates an Azure resource that can be governed using Azure RBAC. A person may have permission to see an Arc-enabled machine in the portal without having Windows local administrator rights on that server. Conversely, a Windows domain administrator does not automatically receive Azure permissions to install extensions or change Azure-side policy assignments.
Design three distinct boundaries: permission to onboard the machine, permission to manage its Azure Arc resource, and permission to perform operating-system actions on the host. Some management features use extensions or remote-access components that introduce another layer of authorization. Inspect those requirements individually rather than assuming that a broad Azure role is equivalent to a local Windows role.
Use resource-group scope to delegate inventory and operations where appropriate. Separate production and test assets so an extension rollout intended for a lab cannot target business-critical servers through an overly broad policy assignment. Audit Azure-side administrative changes and keep corresponding local server logs for the actions that occur inside Windows.
Make patch governance a complete workflow
Azure Update Manager can assess update status and schedule patch deployments for supported Arc-enabled servers. Its value is the operating process around those jobs: which server groups receive which updates, how maintenance windows are chosen, what happens to machines that are offline and how a failed deployment is investigated.
Start with a test ring. Check update assessment data, select a maintenance configuration appropriate for the workload and confirm how the server handles reboots. Then move to a broader ring only after the test group produces acceptable results. Domain controllers, application servers and clustered systems may need different windows and recovery precautions even when they appear in one inventory.
Update Manager entitlements depend on the machine and licensing arrangement. Microsoft describes Windows Server Management benefits for eligible licensed Arc-enabled Windows Server machines, subject to the required activation or enrollment. Confirm the actual fleet’s eligibility, subscription settings and billing; other services such as log ingestion or storage can still cost extra. The presence of an Arc resource alone does not prove that every update-management operation is free.
Deploy extensions with an owner and a recovery plan
Arc-enabled server extensions can help connect supported machines to monitoring, security and other Azure services. A centrally deployed agent still changes the server. It can consume resources, require network access, fail during upgrade or conflict with a previously installed local component.
Maintain an approved extension catalog with versions, supported operating systems, configuration ownership and rollback procedures. Test installation and removal on representative Windows Server versions. Capture the Azure extension status alongside local evidence from the host, because a portal state alone does not prove that every required workload signal has arrived in its destination.
Extensions should follow the same change-management practices as other server software. A team that can push extensions to all production servers has consequential authority; restrict that capability and avoid broad assignments merely to simplify deployment.
Use Windows Admin Center with its actual support boundary
Windows Admin Center provides Windows Server administration tools for files, services, events, certificates and other operating-system tasks. Microsoft also documents a Windows Admin Center experience in the Azure portal for Arc-enabled servers. That Azure-portal integration is documented as preview in current Windows Server guidance; this does not mean the standalone Windows Admin Center application is itself a preview feature.
A pilot for portal-based administration should check the supported Windows Server version, required agent or extension components, networking, authentication path and any licensing conditions. Test the exact tools your operators need rather than assuming a general portal description guarantees each feature in every environment.
Avoid duplicate administration paths. If the operations team already uses Windows Admin Center locally, PowerShell remoting or a managed configuration service, document which path should be used for routine maintenance and which is the controlled fallback. Remote convenience must not hide which identity performed an action or which server received it.
Apply configuration and monitoring policy to the right machines
Azure Policy and machine configuration can assess or enforce supported guest settings on Arc-enabled servers. Build assignments from known baselines: approved operating-system configuration, required monitoring, allowed extensions and asset ownership. The Azure resource group is a convenient deployment boundary, but tags or other resource metadata may be needed to express environment, application and maintenance window.
For monitoring, decide whether Azure Monitor Agent and a data collection rule are appropriate. Select only the logs and performance counters needed for the team’s investigations and service objectives. A large inventory is not useful if administrators cannot trace an alert back to the correct server, owner and application, or if log volume grows without a retention plan.
Check disconnections explicitly. A machine that has not contacted Azure may show stale assessment information. Build a workflow for identifying offline agents, distinguishing planned maintenance from network failure and confirming that missed configuration or update work resumes safely when the connection returns.
Build a hybrid pilot around one recoverable change
Choose three nonproduction Windows Server machines with different responsibilities: for example a domain member, a file server and an application server. Record their operating-system versions, ownership, current update process, monitoring requirements and licensing status. Confirm Arc prerequisites and onboard them through a supported method.
Then select one bounded task: assess update compliance, deploy one approved monitoring extension or evaluate one supported configuration rule. Record the intended scope, Azure identity, server-level permissions, expected data and rollback path. Deliberately disconnect one test machine from the management service and document which signals become stale. Reconnect it and verify how the management state catches up.
The pilot is successful when the team can answer four questions: which machine was affected, which system authorized the change, what happened inside Windows, and how the action would be reversed if it failed. Those answers are stronger evidence than the number of servers appearing as Connected in a portal dashboard.
Keep hybrid management ownership explicit
After the three-server pilot, assign accountable owners to each operation rather than declaring success because all machines appear as Connected. The following is an illustrative split; organizations may combine or redistribute roles, but each change still needs a clear authority boundary and evidence of the result.
| Operation | Accountable control | Evidence after change |
|---|---|---|
| Onboarding and resource placement | Azure platform owner with approved registration rights | Resource ID, subscription, tags and healthy agent connection |
| Server updates and reboot windows | Windows workload owner with approved Update Manager access | Ring, assessment, maintenance window and post-reboot health |
| Extensions and monitoring | Platform operator, coordinated with the application owner | Extension version, local status and expected telemetry |
| Offboarding and license eligibility | Asset owner with platform and procurement review | Retirement record, removed Arc relationships and entitlement decision |
Keep the owner map alongside extension permissions, agent versions, Update Manager schedules, Windows access and emergency rollback procedures. When an agent disconnects or a patch fails, the record should identify who can diagnose the Azure side, who controls Windows, and who is allowed to reverse the change.
A hybrid-management pilot has earned expansion only when operators can explain which system authorized a change, what actually happened on the server and how recovery would work. That is a stricter test than counting connected machines.