When the monthly security update is released, the real question isn’t “should we install it”; it’s which server it goes on, in what order, with what rollback plan, and how to prove it was installed. Microsoft adding WSUS to its list of deprecated features reopened this question for many IT teams. Below we compare WSUS, Configuration Manager, Azure Update Manager and hotpatch by operational load in the field, and give a workable setup for maintenance windows, the ring plan, cluster servers and the compliance report.
Your options
For on-premises and hybrid Windows servers, there are in practice four paths:
- WSUS: The role that comes with Windows Server and approves and distributes updates locally. No license cost; weak reporting and scheduling.
- Configuration Manager (MECM): Manages deployment with collections, maintenance windows and detailed compliance reports. The software update point (SUP) still uses WSUS underneath.
- Azure Update Manager: Scans and schedules Azure VMs and servers on premises or in other clouds connected with Azure Arc from a single console. It manages Windows and Linux together.
- Hotpatch: Under the right version and licensing conditions, applies security updates to code in memory without a restart. It isn’t a deployment tool; it’s an update method.
These aren’t mutually exclusive. While Azure Update Manager manages when the update is installed, the server can still pull files from a local WSUS. Mixed setups are common. What causes trouble is not having in writing who manages what.
WSUS status and limits
Microsoft announced WSUS as deprecated in September 2024. The role isn’t going away immediately: it continues to be included in current Windows Server versions, and updates are published through the WSUS channel; but no new investment is being made in the role. Track changes in sub-features such as driver synchronization on the removed features page in Microsoft Learn.
Known weak points of WSUS in operation:
- Scheduling depends on the automatic update settings in Group Policy. There is no orchestration that says “update this server tonight at 02:00 and restart in this order.”
- If the database (WID or SQL) isn’t maintained, it bloats; the console times out and clients start giving scan errors.
- Reports answer the question “was the approved update installed,” but don’t give a unified view across multiple WSUS servers or for Linux.
A common WSUS failure is a database that hasn’t been cleaned up for years and is full of superseded updates. If you will continue with WSUS for a while, put the following cleanup on the calendar; it can take hours on large databases, so run it outside working hours:
Get-WsusServer | Invoke-WsusServerCleanup -CleanupObsoleteUpdates `
-CleanupUnneededContentFiles -DeclineExpiredUpdates -DeclineSupersededUpdates
Azure Update Manager and Azure Arc
Azure Update Manager replaced Azure Automation Update Management and uses the server’s own update mechanism instead of the Log Analytics agent. To manage an on-premises server, the server must first be connected with Azure Arc. The Arc agent (Connected Machine agent) represents the server as a resource in Azure; after that, the following can be used:
- Periodic assessment: Automatically scans for missing updates and lists them by classification.
- Maintenance configuration: The recurring window, the classifications to include, KB exclusions and restart behavior are defined here.
- Dynamic scope: Automatically attaches servers to the plan by subscription, resource group, location or tag filter. Giving a new server the right tag is enough to put it in the right ring.
- Pre and post events: Trigger scripts or automation before and after maintenance, such as stopping an application service or running a health check.
On cost, the service charges nothing extra for Azure VMs; for servers connected with Arc, a monthly per-server fee applies. Some licensing programs and Microsoft Defender for Servers plans may cover this fee. Prices and coverage can change; check the current pricing page before deciding. For details of how it works, the Azure Update Manager overview document is the right starting point.
Comparison
| Criterion | WSUS | Configuration Manager | Azure Update Manager |
|---|---|---|---|
| License/fee | Included with Windows Server | Separate license | Per-server fee for Arc servers |
| Scheduling | GPO-based, coarse | Maintenance windows | Maintenance configuration |
| Linux support | None | None (Linux client removed) | Yes |
| Reporting | Basic | Detailed | Resource Graph queries |
| Infrastructure load | Server + database | Site server, SQL, SUP | Arc agent |
| Suitable environment | Small, single site | Those with an existing MECM investment | Hybrid and multi-site |
A simple path to a decision: if Configuration Manager is already running, keep server patching there. If there is no MECM and the number of servers is in the tens, moving to Azure Update Manager usually brings less operational load than building a new on-premises management server. In isolated networks where internet access is closed by policy, WSUS or MECM remain realistic options. An isolated network can’t use a cloud tool.
When hotpatch makes sense
Hotpatch started in Azure with Windows Server 2022 Datacenter: Azure Edition; with Windows Server 2025 Standard and Datacenter it was opened to on-premises servers connected to Azure Arc as well. Microsoft’s current document states that hotpatch through Arc is offered at no extra charge; check the fee and prerequisites in the document before enabling it. The cycle works like this: in four baseline months a year, the cumulative update is installed and a restart is required; in the months in between, security updates are applied without a restart. The number of restarts per year drops from twelve to four.
It is most useful for critical standalone servers outside a cluster, where finding a downtime window is hard. The supported version, prerequisites such as virtualization-based security (VBS) and subscription conditions vary by version; verify the current requirements in the Microsoft documentation before implementing.
Maintenance window and ring plan
Patching all servers on the same night means a faulty update can take down every service at once. A ring plan spreads the risk over time. Microsoft releases updates on the second Tuesday of each month; the example plan below is built around this calendar:
| Ring | Scope | Timing (example) |
|---|---|---|
| 0 – Test | Test environment, IT management servers | Release + 1–2 days |
| 1 – Pilot | One member of each redundant service, non-critical servers | Release + 5–7 days |
| 2 – General | Application and file servers | Release + 10–14 days |
| 3 – Critical | Second group of DCs, databases, cluster nodes | Release + 14–21 days |
Points skipped in the field when applying this table:
- Put the members of redundant pairs (two DCs, two DHCP servers, two web servers) in different rings. If both go down on the same night, redundancy exists only on paper.
- Before each window, check that the backup finished successfully. On a virtual server, a snapshot is a short-term rollback tool, not a backup.
- Have an “emergency patch” procedure that compresses the rings for an actively exploited vulnerability written down in advance. The procedure isn’t debated on that night.
- Define a post-restart verification list: service states, application health checks, new errors in the event log.
For the general scope of scheduled maintenance, see the server maintenance checklist for SMBs article; for the risk of not receiving patches on versions whose support has ended, see the Windows Server 2016 end of support migration plan article.
Update tools carry out the deployment step. Define organization-wide approval, exception and record-keeping within change and patch management processes.
Cluster-Aware Updating for clusters
Updating Failover Cluster nodes with classic tools carries the risk of a node restarting without its roles being drained. In the field, the result is unplanned movement of virtual machines or a shaky quorum. Cluster-Aware Updating (CAU) runs the process node by node: it puts the node in maintenance mode, moves the roles to other nodes, installs the update, restarts, brings the node back into the cluster and moves on to the next.
CAU runs in two modes. In self-updating mode, a CAU role is added to the cluster and runs on a schedule; in remote-updating mode, it is triggered from a management machine outside the cluster. First test that the environment is ready, then do a one-time run:
# Prerequisite check
Test-CauSetup -ClusterName CL-HV01
# One-time update run
Invoke-CauRun -ClusterName CL-HV01 -CauPluginName Microsoft.WindowsUpdatePlugin `
-MaxFailedNodes 0 -MaxRetriesPerNode 2 -RequireAllNodesOnline -Force
# Scheduled self-updating mode (3rd Tuesday of each month)
Add-CauClusterRole -ClusterName CL-HV01 -DaysOfWeek Tuesday -WeeksOfMonth 3 `
-MaxFailedNodes 0 -RequireAllNodesOnline -EnableFirewallRules -Force
-RequireAllNodesOnline prevents updating from starting while a node is already failed. Don’t remove this parameter. If you use Azure Update Manager, don’t put cluster nodes in a single maintenance configuration that updates them at the same time; either continue with CAU or split the nodes into separate windows. For the cluster design itself, see the Hyper-V Failover Cluster planning article.
Reporting patch compliance
A patch you can’t prove was installed counts as not installed in an audit. The questions asked are always the same: how many critical updates are missing on which server, when was the last successful installation, which servers are exceptions and why?
Azure Update Manager writes assessment results to Azure Resource Graph. The query below lists the latest assessment status for Azure VMs and Arc servers:
patchassessmentresources
| where type in~ ("microsoft.compute/virtualmachines/patchassessmentresults",
"microsoft.hybridcompute/machines/patchassessmentresults")
| project id, status = properties.status,
lastAssessment = properties.lastModifiedDateTime,
missing = properties.availablePatchCountByClassification
In environments without Arc, at least collect the last installation date. Get-HotFix doesn’t show every update type, but it is enough to roughly see whether a server received the monthly cumulative update:
$servers = Get-Content .\servers.txt
Invoke-Command -ComputerName $servers -ScriptBlock {
Get-HotFix | Sort-Object InstalledOn -Descending | Select-Object -First 1
} | Select-Object PSComputerName, HotFixID, InstalledOn
To produce this report automatically with a scheduled task, use the approach in the server automation with PowerShell article. Show exceptions in a separate column: a server that can’t take a particular update because of vendor support should be recorded with its owner’s approval and an end date. An exception without a date is a permanent hole.
If you want to rebuild the update infrastructure in a hybrid server fleet or tie the current process to regular maintenance, you can see how we work on the server and cloud management and technical support and scheduled maintenance pages.
Frequently asked questions
If WSUS has been deprecated, will our existing server stop working?
No. Deprecated means the role won’t receive new features and may be removed in the future. Existing WSUS servers continue to receive and distribute updates; what changes is that you need to make your long-term infrastructure decision with this in mind.
Do servers need direct internet access to use Azure Update Manager?
The Azure Arc agent needs to reach Azure endpoints over HTTPS; this access can also be provided through a proxy or Private Link. The update files themselves are downloaded from the server’s configured source, i.e. Windows Update or WSUS.
With hotpatch, are servers never restarted?
They are. In the hotpatch cycle, four baseline updates a year require a restart; security updates in the months in between are applied without a restart. Some non-security updates and components such as .NET may still require a restart.
In which ring should domain controllers be updated?
DCs are updated a few days after the pilot ring, one after another rather than at the same time. Define two separate maintenance windows so that at least one DC in each domain stays online at all times.