Test your Patches! A Staged Patching Solution with Azure Update Manager

This post has been republished via RSS; it originally appeared at: Core Infrastructure and Security Blog articles.

A common challenge faced by most enterprise organizations who, hopefully, automate their operating systems patching cycles is to ensure that only the Windows and Linux packages updates that were tested in dev/test pre-production environments reach production machines. This article is for those readers who have been implementing automated, scheduled patching with Azure Update Manager and now want to put into practice a staged patching solution following good patching reliability practices. Keep tuned and read the solution my colleague @Wiszanyel Cruz and myself have developed and helped put it in place for some customers.


Azure Update Manager Staged Patching with Azure Automation


With a staged patching solution, OS updates are first deployed in a test environment and are later deployed in pre-production and production environments, ensuring the latter environments only get the specific updates initially deployed in the test environment. With this approach, you significantly decrease the chances of having an OS update breaking a production system.


The typical setup of staging patching can be as follows:


  1. Dev/Test machines of a specific OS type/version are recurrently patched (e.g., every few weeks) for all update classifications (stage 0).
  2. Each Dev/Test patch cycle ends with a specific set of updates (e.g., specific Windows KB IDs or Linux package versions) that were successfully installed across all Dev/Test machines.
  3. Pre-production machines are patched a few days later (stage 1) with the specific updates that were deployed in Dev/Test (stage 0).
  4. Production machines are patched one or two weeks later (stage 2) with the same updates that were deployed and tested in Dev/Test (stage 0) and Pre-Production (stage 1).


This staged patching approach can be implemented with the help of the Create-StagedMaintenanceConfiguration.ps1 PowerShell script, which runs after stage 0 and automates stages 1 and 2. This script can be, for example, deployed as an Azure Automation Runbook scheduled to run after the Dev/Test recurrent update cycle (see diagram below). It works for both Windows and Linux scenarios (Azure VMs and Azure Arc-enabled servers).


Staged Patching ArchitectureStaged Patching Architecture



Recommended staged patching strategy


The value of a staged patching solution is to ensure that patches deployed in a production environment are previously tested in non-production environments. The more consistent and repeatable the patching workflow is, the more confidence you have in the patches that reach production. For this reason, it is recommended to define maintenance configurations specifically for each OS version and ensure further stages are applied only to machines of the same OS version. For example, if your environment has a mix of Windows 2016, Windows 2019, Ubuntu 20.4, and Ubuntu 22.4 servers, you should define four different staged patching workflows, one for each OS version. With this approach, for example, Windows 2019 production machines will only get patches that were tested in similar Windows 2019 non-production machines and, similarly, Ubuntu 20.4 production servers will only get package updates that were tested on Ubuntu 20.4 non-production servers.


Tagging is your best friend in this strategy. By tagging your servers according to their OS version and patching stage, it will be easier to dynamically define the scope of a specific patching stage. Continuing with the example above, your servers can be tagged as follows:


  • An aum-stage tag for each of the patching stages (e.g., aum-stage=dev, aum-stage=preprod, aum-stage=prod, aum-stage=prod-ha-instance1, aum-stage=prod-ha-instance2, etc.).
  • An os-name tag for each of the OS versions of your environment (e.g., os-name=windows2016, os-name=windows2019, os-name=ubuntu20, os-name=ubuntu22, etc.)


You can choose whatever tagging strategy that meets your staged patching requirements, provided you end up with a predictable patching workflow. Once you have tagged all your servers according to their stage in the different patching workflows, you must define a stage 0 recurring Maintenance Configuration (e.g., for servers tagged aum-stage=dev), for each of the OS versions. For the next stages, you have two options to leverage this solution:


  1. Schedule the Create-StagedMaintenanceConfiguration runbook to run after stage 0 (e.g., the next day) and configure it to create/update all the next stages (e.g., stages 1, 2, etc.) based on the results of stage 0.
  2. Chain all the stages to each other, by scheduling the Create-StagedMaintenanceConfiguration runbook for each of the stages before production:
    • After stage 0, create/update stage 1 based on the results of stage 0
    • After stage 1, create/update stage 2 based on the results of stage 1
    • etc.


IMPORTANT: you must ensure the last stage (production) is not scheduled before the next iteration of the phase 0 stage, otherwise you will end up having the production Maintenance Configuration overwritten with a new schedule/patch selection before it is deployed. Also, bear in mind that, as Azure Resource Graph keeps the patching results history for up to 30 days, your updates cycle must not exceed this interval.


Validating the quality of the patching stages before production is one important perspective not addressed by this solution. It is not sufficient to patch dev/test servers - we must ensure the patches pass minimum quality tests before reaching production. At this moment, you must run a parallel process that performs this validation (e.g., automated tests running in the patched servers). With the support for post-maintenance tasks in Azure Update Manager, you can integrate quality assurance in this solution.




  • The machines in the scope of this solution must be supported by Azure Update Manager and fulfill its pre-requisites.
  • The machines in the scope of this solution must have the Customer Managed Schedules patch orchestration mode, a pre-requisite for scheduled patching.
  • At least one recurrent scheduled patching Maintenance Configuration covering a part of the machines in scope. As this maintenance configuration will serve as the reference for the following patching stages, it should be assigned to non-production machines and, ideally, recur every few weeks. See the above recommendations for an effective patching strategy.
  • An Azure Automation Account with an associated Managed Identity (can be a system or user-assigned identity) and the following modules installed: Az.Accounts, Az.Resources, and Az.ResourceGraph. This solution is based on an Automation Account, but you can use other approaches, such as Azure Functions.
  • The Automation Account Managed Identity must have the following minimum permissions (as a custom role) on the subscription where the reference maintenance configuration was created:
    • */read
    • Microsoft.Maintenance/maintenanceConfigurations/write
    • Microsoft.Maintenance/configurationAssignments/write
    • Microsoft.Resources/deployments/*


Setup instructions


  1. Ensure your Azure VMs and Azure Arc-enabled servers are tagged according to the staged patching strategy you defined (see above). Use tags to group your servers according to their patching phase and OS version.
  2. Create a recurrent scheduled patching Maintenance Configuration for Phase 0 of each OS version in your environment. See instructions here. Assign this Maintenance Configuration to the servers that will serve as the initial testbed for your patches. You can assign servers either directly to the Maintenance Configuration or dynamically, with a Dynamic Scope.
  3. Create or reuse an Azure Automation Account.
  4. Install all required Automation Account modules (Az.Accounts and Az.Resources are usually built-in, but Az.ResourceGraph is not). See here how to install additional modules.
  5. Assign a Managed Identity to the Automation Account and grant it the required privileges (see pre-requisites above).
  6. Import the Create-StagedMaintenanceConfiguration.ps1 Runbook into the Automation Account, by following these steps. Download the runbook first to your local machine. The runbook must be configured as PowerShell 5.1. Do not forget to publish the runbook (draft runbooks cannot be scheduled).
  7. Create an Azure Automation schedule for each of the Phase 0 Maintenance Configurations (one per OS version). The schedule must have the same frequency as the Maintenance Configuration it refers to, with at least an 8-hour offset. For example, if the Maintenance Configuration is scheduled on Mondays, every 2 weeks at 8:00 p.m., then the respective Azure Automation schedule should be scheduled on Tuesdays, every 2 weeks at least at 4:00 a.m.
  8. Link the Create-StagedMaintenanceConfiguration runbook to each of the schedules and specify its parameters according to the instructions below. To obtain the Maintenance Configuration ID, check the "Properties" blade of the Maintenance Configuration.
  9. (Optional) If you prefer to adopt a more conservative chained staged approach, you need to create additional schedules (for further stages before production) and link them to the same runbook. In this case, you will have to anticipate the Maintenance Configuration IDs that will result from the previous stages' executions, which will be in the form /subscriptions/<phase 0 maintenance configuration subscription ID>/resourceGroups/<phase 0 maintenance configuration resource group>/providers/Microsoft.Maintenance/maintenanceConfigurations/<previous stage name>.


Pre- and post-maintenance tasks


Pre- and post-maintenance tasks in Azure Update Manager follow an event-based architecture, in which you subscribe to events coming from a system topic associated to the Configuration Maintenance, and use, for example, an Azure Automation runbook or Azure Function as the destination of the event. You can learn more about how to configure pre- and post-maintenance tasks in the Azure Update Manager pre and post events overview, and on the how-to guide and tutorials for Azure Automation- and Azure Functions-based tasks. You can find in our Azure Update Manager Tools code repository a code sample for one of the pre- and post-maintenance scenarios: turn machines on with Start-StagedMaintenanceVMs.ps1 and turn them off with Deallocate-StagedMaintenanceVMs.ps1.


NOTE: the staged patching solution here described does not propagate pre- and post-maintenance tasks coming from the reference maintenance configuration to the following stages, but you can manually configure them once the subsequent stages have been created for the first time.


Create-StagedMaintenanceConfiguration script parameters


The Create-StagedMaintenanceConfiguration.ps1 PowerShell script receives the following parameters:


  • MaintenanceConfigurationId: Azure Resource Manager ID of the Maintenance Configuration to be used as a reference to create maintenance configurations for further stages
  • NextStagePropertiesJson: JSON-formatted parameter that will define the scope of the next maintenance configurations, with the schema described in detail and followed by a JSON sample you can check in our code repository.


Once you have followed the setup instructions, you should see new maintenance configurations showing up in your environment for each of the stages you specified as parameters. Happy and well-tested patching! :smiling_face_with_smiling_eyes:




This solution is not supported under any Microsoft standard support program or service. The scripts are provided AS IS without warranty of any kind. The entire risk arising out of the use or performance of the scripts and documentation remains with you.

Leave a Reply

Your email address will not be published. Required fields are marked *


This site uses Akismet to reduce spam. Learn how your comment data is processed.