The path to becoming a Microsoft MVP (Most Valuable Professional) is not as linear as some might think, others have a goal to receive the Microsoft MVP Award, and others have a passion for technology that shows in community activities such as speaking or user groups, helping others on forums, and helping to maintain documentation and helping others be up-to-speed with the ever changing ecosystem that is the Microsoft stack - it is not a one size fits all, just as there are multiple ways of learning, there are multiple ways to the MVP Award.

I join Christian Buckley for episode 167 of his #MVPbuzChat Podcast/Video chat, to tell my story, feel free to check it out (if you can ignore the bad camera angle!) and other MVPs talk about their journey to the Microsoft MVP Award.

• #MVPbuzzChat with Luke Murray

Solution architecture is concerned with the planning, design, implementation, and ongoing improvement of a technology system.

The architecture of a system must balance and align the business requirements with the technical capabilities that are needed to execute those requirements.

The finished architecture is a balance of risk, cost, and capability throughout the system and its components.

Running a solution in the cloud does not reduce the need for requirements to be clear. In fact, the flexibility and power provided by the cloud mean that it is even more important to have clear requirements from business stakeholders; otherwise, you could end up solving problems that don't exist, missing an important design decision, or going beyond the available budget by adding unnecessary resiliency.

Non-functional requirements (NFRs)​

Below is a short list of NFRs (not exhaustive) that may be provided by the business to help inform the design of a solution.

Reliability requirements​

• Service level agreement (SLA)
• Uptime objective
• Recovery time objective (RTO)
• Recovery point objective (RPO)
• Recoverability

Security requirements​

• Geographical location
• Compliance and legislation
• Identity and access management
• Privacy
• Data Integrity
• Public or private endpoints (or both)
• OWASP
• Hybrid connectivity
• DDOS

Performance requirements​

• Peak throughput, e.g., Requests per minute (RPM), active users
• Business plan for growth
• UX metrics (e.g., Page load time)
• Asynchronous vs Synchronous operations
• Workload profile (predictable, unpredictable, peak time of day)
• Scalability
• Data estate size and growth rate
• Time-to-live (TTL) of reports and views (real-time vs eventual consistency)

Operational requirements​

• Prod and non-prod environments (Dev/Test, QA, Pre-prod, Prod)
• Release frequency (hours / days / months)
• Time to onboard (new customer)
• Licensing
• Cost (Management)
• Manageability

Cost optimization​

• Cost per user
• Target hosting costs as a percentage of revenue
• Pricing model
• Tenancy model

Azure SLAs​

• Familiarize yourself with Azure service-level agreements
• An Azure Service-level Agreement (SLA) can also be read as a minimum service-level objective (SLO).
• An SLA is a financial guarantee, not an absolute guarantee
• Read the SLA details carefully, particularly the definition of "downtime" for each service, which gives important hints about failure modes

For example, in the SLA for Azure SQL Database, "downtime" is defined as:

"The total accumulated Deployment Minutes across all Databases in a given Microsoft Azure subscription during which the Database is unavailable. A minute is considered unavailable for a given Database if all continuous attempts by Customer to establish a connection to the Database within the minute fail."

The Azure SQL Database team expect almost all outages to be transient (brief and non-recurring). Therefore, the retry pattern should be used to continuously retry for up to a minute. This is typical in cloud services; retry has been the default behaviour in ADO.NET since .NET Framework 4.6.1.

External Resources​

Finally, resources such as the Azure Architecture Center, Cloud Adoption and Well-Architected Framework can help with thinking around the design and building blocks of your architecture

• Azure Architecture Center
• Microsoft Azure Well-Architected Framework
• Microsoft Cloud Adoption Framework for Azure

Did you know you can contribute to Microsoft documentation (ms docs)?

Suppose you see something not quite right, technically or even if the document's readability doesn't look right! Then, in true community style, you can contribute!

Tip: You can edit it straight from the Github webpage directly, or pressing "." in a Github repository will open up Visual Studio Code in Dev spaces with the markdown linter to help check against best practices from your browser.

See the image below for an example:

Once the pull request is made, it will be reviewed by designated technical document reviewers/product owners at Microsoft. Then your changes will be merged live if successful (and if not, the reviewers will let you know why and what changes could be made)!

• Microsoft Docs contributor guide overview

If you don't want to make the edit yourself, you can also raise an issue and give your feedback by linking to the document, and this will then be worked on by someone to review, contact the relevant product owners, and amended.

Try to be as concise as possible, as people reading it may not have the same experience as you!

Virtual Machines in Microsoft Azure have different states and, depending on what state the Virtual Machine is in, will determine whether you get billed or not (for the Compute, storage and network adapters are still billed).

Suppose a Virtual Machine is not being used. In that case, turning off a Virtual Machine from the Microsoft Azure Portal (or programmatically via PowerShell/Azure CLI) is recommended to ensure that the Virtual Machine is deallocated and its affinity on the host has been released.

However, you need to know this, and those new to Microsoft Azure, or users who don't have Virtual Machine Administrator rights to deallocate a Virtual Machine, may simply shut down the operating system, leaving the Virtual Machine in a 'Stopped' state, but still tied to an underlying Azure host and incurring cost.

Our solution can help; by triggering an Alert when a Virtual Machine becomes unavailable due to a user-initiated shutdown, we can then start an Azure Automation runbook to deallocate the Virtual Machine.

Overview​

Today, we are going to set up an Azure Automation runbook, triggered by a Resource Health alert that will go through the following steps:

1. User shutdowns Virtual Machine from within the Operating System
2. The Virtual Machine enters an unavailable state
3. A Resource Alert is triggered when the Virtual Machine becomes unavailable (after being available) by a user initiated event
4. The Alert triggers a Webhook to an Azure Automation runbook
5. Using permissions assigned to the Azure Automation account through a System Managed Identity connects to Microsoft Azure and checks the VM state; if the Virtual Machine state is still 'Stopped', then deallocate the virtual machine.
6. Then finally, resolve the triggered alert.

To do this, we need a few resources.

• Azure Automation Account
• Az.AlertsManagement module in the Azure Automation account
• Az.Accounts module (updated in the Azure Automation account)
• Azure Automation runbook (I will supply this below)
• Resource Health Alert
• Webhook (to trigger to the runbook and pass the JSON from the alert)

And, of course, 'Contributor' rights to the Microsoft Azure subscription to provide the resources and the alerts and resources and set up the system managed identity.

We will set up this from scratch using the Azure Portal and an already created PowerShell Azure Automation runbook.

Deploy Deallocate Solution​

Setup Azure Automation Account​

Create Azure Automation Account​

First, we need an Azure Automation resource.

1. Log into the Microsoft Azure Portal.
2. Click + Create a resource.
3. Type in automation
4. Select Create under Automation, and select Automation.
5. 
6. Select your subscription
7. Select your Resource Group or Create one if you don't already have one (I recommend placing your automation resources in an Azure Management or Automation resource group, this will also contain your Runbooks)
8. Select your region
9. 
10. Select Next
11. Make sure: System assigned is selected for Managed identities (this will be required for giving your automation account permissions to deallocate your Virtual Machine, but it can be enabled later if you already have an Azure Automation account).
12. Click Next
13. Leave Network connectivity as default (Public access)
14. Click Next
15. Enter in appropriate tags
16. 
17. Click Review + Create
18. After validation has passed, select Create

Configure System Identity​

Now that we have our Azure Automation account, its time to set up the System Managed Identity and grant it the following roles:

• Virtual Machine Contributor (to deallocate the Virtual Machine)
• Monitoring Contributor (to close the Azure Alert)

You can set up a custom role to be least privileged and use that instead. But in this article, we will stick to the built-in roles.

1. Log into the Microsoft Azure Portal.
2. Navigate to your Azure Automation account
3. Click on: Identity
4. Make sure that the System assigned toggle is: On and click Azure role assignments.
5. 
6. Click + Add role assignments
7. Select the Subscription (make sure this subscription matches the same subscription your Virtual Machines are in)
8. Select Role: Virtual Machine Contributor
9. Click Save
10. Now we repeat the same process for Monitoring Contributor
11. lick + Add role assignments
12. Select the Subscription (make sure this subscription matches the same subscription your Virtual Machines are in)
13. Select Role: Monitoring Contributor
14. Click Save
15. Click Refresh (it may take a few seconds to update the Portal, so if it is blank - give it 10 seconds and try again).
16. You have now set up the System Managed identity and granted it the roles necessary to execute the automation.

Import Modules​

We will use the Azure Runbook and use a few Azure PowerShell Modules; by default, Azure Automation has the base Azure PowerShell modules, but we will need to add Az.AlertsManagement, and update the Az.Accounts as required as a pre-requisite for Az.AlertsManagement.

1. Log into the Microsoft Azure Portal.
2. Navigate to your Azure Automation account
3. Click on Modules
4. Click on + Add a module
5. Click on Browse from Gallery
6. Click: Click here to browse from the gallery
7. Type in: Az.Accounts
8. Press Enter
9. Click on Az.Accounts
10. Click Select
11. 
12. Make sure that the Runtime version is: 5.1
13. Click Import
14. Now that the Az.Accounts have been updated, and it's time to import Az.AlertsManagement!
15. Click on Modules
16. Click on + Add a module
17. Click on Browse from Gallery
18. Click: Click here to browse from the gallery
19. Type in: Az.AlertsManagement (note its Alerts)
20. Click Az.AlertsManagement
21. 
22. Click Select
23. Make sure that the Runtime version is: 5.1
24. Click Import (if you get an error, make sure that Az.Accounts has been updated, through the Gallery import as above)
25. Now you have successfully added the dependent modules!

Import Runbook​

Now that the modules have been imported into your Azure Automation account, it is time to import the Azure Automation runbook.

1. Log into the Microsoft Azure Portal.
2. Navigate to your Azure Automation account
3. Click on Runbooks
4. Click + Create a runbook
5. Specify a name (i.e. Deallocate-AzureVirtualMachine)
6. Select Runbook type of: PowerShell
7. Select Runtime version of: 5.1
8. Type in a Description that explains the runbook (this isn't mandatory, but like Tags is recommended, this is an opportunity to indicate to others what it is for and who set it up)
9. 
10. Click Create
11. Now you will be greeted with a blank edit pane; paste in the Runbook from below:

#requires -Version 3.0 -Modules Az.Accounts, Az.AlertsManagement
<#
    .SYNOPSIS
    PowerShell Azure Automation Runbook for Stopping Virtual Machines, that have been Shutdown within the Windows Operating System (Stopped and not Deallocated). 
    .AUTHOR
    Luke Murray (https://github.com/lukemurraynz/)
#>

[OutputType('PSAzureOperationResponse')]
param ( 
    [Parameter(Mandatory = $true, HelpMessage = 'Data from the WebHook/Azure Alert')][Object]$WebhookData
)

Import-Module Az.AlertsManagement
$ErrorActionPreference = 'stop'

# Get the data object from WebhookData
$WebhookData = $WebhookData.RequestBody
Write-Output -InputObject $WebhookData 
$Schema = $WebhookData | ConvertFrom-Json

#Sets the Webhook data into object
$Essentials = [object] ($Schema.data).essentials
Write-Output -InputObject $Essentials 

# Get the first target only as this script doesn't handle multiple and and export variables for the resource.
$alertIdArray = (($Essentials.alertId)).Split('/')
$alertTargetIdArray = (($Essentials.alertTargetIds)[0]).Split('/')
$alertid =  ($alertIdArray)[6]
$SubId = ($alertTargetIdArray)[2]
$ResourceGroupName = ($alertTargetIdArray)[4]
$ResourceType = ($alertTargetIdArray)[6] + '/' + ($alertTargetIdArray)[7]
$ResourceName = ($alertTargetIdArray)[-1]
$status = $Essentials.monitorCondition
Write-Output -InputObject $alertTargetIdArray
Write-Output  -InputObject "status: $status" -Verbose

#Sets VM shutdown
if (($status -eq 'Activated') -or ($status -eq 'Fired')) {
    $status = $Essentials.monitorCondition
    Write-Output -InputObject "resourceType: $ResourceType" -Verbose
    Write-Output  -InputObject "resourceName: $ResourceName" -Verbose
    Write-Output  -InputObject "resourceGroupName: $ResourceGroupName" -Verbose
    Write-Output  -InputObject "subscriptionId: $SubId" -Verbose

    # Determine code path depending on the resourceType
    if ($ResourceType -eq 'Microsoft.Compute/virtualMachines') {
        # This is an Resource Manager VM
        Write-Output  -InputObject 'This is an Resource Manager VM.' -Verbose

        # Ensures you do not inherit an AzContext in your runbook
        Disable-AzContextAutosave -Scope Process

        # Connect to Azure with system-assigned managed identity
        $AzureContext = (Connect-AzAccount -Identity).context

        # set and store context
        $AzureContext = Set-AzContext -SubscriptionName $AzureContext.Subscription -DefaultProfile $AzureContext
        Write-Output   -InputObject $AzureContext 
        #Checks Azure VM status
        $VMStatus = Get-AzVM -ResourceGroupName $ResourceGroupName -Name $ResourceName -Status

        Write-Output  -InputObject $VMStatus
        If ($VMStatus.Statuses[1].Code -eq 'PowerState/stopped') {
            Write-Output  -InputObject "Stopping the VM, it was Shutdown without being Deallocated - $ResourceName - in resource group - $ResourceGroupName" -Verbose
            Stop-AzVM -Name $ResourceName -ResourceGroupName $ResourceGroupName -DefaultProfile $AzureContext -Force -Verbose
      
            #Check VM Status after deallocation
            $VMStatus = Get-AzVM -ResourceGroupName $ResourceGroupName -Name $ResourceName -Status -Verbose
            
            Write-Output  -InputObject $VMStatus

            If ($VMStatus.Statuses[1].Code -eq 'PowerState/deallocated') {
                #Closes Alert
                Write-Output  -InputObject $VMStatus.Statuses[1].Code
                Write-Output  -InputObject $alertid 
                Get-AzAlert -AlertId $alertid  -verbose -DefaultProfile $AzureContext
                Get-AzAlert -AlertId $alertid  -verbose -DefaultProfile $AzureContext | Update-AzAlertState -State 'Closed' -Verbose -DefaultProfile $AzureContext
            }
        }
             
        Elseif ($VMStatus.Statuses[1].Code -eq 'PowerState/deallocated') {
            Write-Output  -InputObject 'Already deallocated' -Verbose
        }

        Elseif ($VMStatus.Statuses[1].Code -eq 'PowerState/running') {
            Write-Output  -InputObject 'VM running. No further actions' -Verbose
        }

        # [OutputType(PSAzureOperationResponse")]
    }
}
else {
    # The alert status was not 'Activated' or 'Fired' so no action taken
    Write-Output  -InputObject ('No action taken. Alert status: ' + $status) -Verbose
}

13. Click Save
14. 
15. Click Publish (so the runbook is actually in production and can be used)
16. You can select View or Edit at any stage, but you have now imported the Azure Automation runbook!

Setup Webhook​

Now that the Azure runbook has been imported, we need to set up a Webhook for the Alert to trigger and start the runbook.

1. Log into the Microsoft Azure Portal.
2. Navigate to your Azure Automation account
3. Click on Runbooks
4. Click on the runbook you just imported (i.e. Deallocate-AzureVirtualMachine)
5. Click on Add webhook
6. Click Create a new webhook
7. Enter a name for the webhook
8. Make sure it is Enabled
9. You can edit the expiry date to match your security requirements; make sure you record the expiry date, as it will need to be renewed before it expires.
10. Copy the URL and paste it somewhere safe (you won't see this again! and you need it for the next steps)
11. 
12. Click Ok
13. Click on Configure parameters and run settings.
14. Because we will be taking in dynamic data from an Azure Alert, enter in: [EmptyString]
15. Click Ok
16. Click Create
17. You have now set up the webhook (make sure you have saved the URL from the earlier step as you will need it in the following steps)!

Setup Alert & Action Group​

Now that the Automation framework has been created with the Azure Automation account, runbook and webhook, we now need a way to detect if a Virtual Machine has been Stopped; this is where a Resource Health alert will come in.

1. Log into the Microsoft Azure Portal.
2. Navigate to: Monitor
3. Click on Service Health
4. Select Resource Health
5. Select + Add resource health alert
6. Select your subscription
7. Select Virtual machine for Resource Type
8. You can target specific Resource Groups for your alert (and, as such, your automation) or select all.
9. Check Include all future resource groups
10. Check include all future resources
11. Under the Alert conditions, make sure Event Status is: All selected
12. Set Current resource status to Unavailable
13. Set Previous resource status to All selected
14. For reason type, select: User initiated and unknown
15. 
16. Now that we have the Alert rule configured, we need to set up an Action group. That will get triggered when the alert gets fired.
17. Click Select Action groups.
18. Click + Create action group
19. Select your subscription and resource group (this is where the Action alert will go, I recommend your Azure Management/Monitoring resource group that may have a Log Analytics workspace as an example).
20. Give your Action Group a name, i.e. AzureAutomateActionGroup
21. The display name will be automatically generated, but feel free to adjust it to suit your naming convention
22. Click Next: Notifications
23. Under Notifications, you can trigger an email alert, which can be handy in determining how often the runbook runs. This can be modified and removed if it is running, especially during testing.
24. Click Next: Actions
25. Under Action Type, select Webhook
26. Paste in the URI created earlier when setting up the Webhook
27. Select Yes to enable the common alert schema (this is required as the JSON that the runbook is parsing is expecting it to the in the schema, if it isn't the runbook will fail)
28. 
29. Click Ok
30. Give the webhook a name.
31. Click Review + create
32. Click Create
33. Finally, enter in an Alert name and description, specify the resource group for the Alert to go into and click Save.

Test Deallocate Solution​

So now we have stood up our:

• Azure automation account
• Alert
• Action Group
• Azure automation runbook
• Webhook

It is time to test! I have a VM called: VM-D01, running Windows (theoretically, this runbook will also run against Linux workloads, as its relying on the Azure agent to send the correct status to the Azure Logs, but in my testing, it was against Windows workloads) in the same subscription that the alert has been deployed against.

As you can see below, I shut down the Virtual Machine. After a few minutes (be patient, Azure needs to wait for the status of the VM to be triggered), an Azure Alert was fired into Azure Monitor, which triggered the webhook and runbook, and the Virtual Machine was deallocated, and the Azure Alert was closed.

Tags in Microsoft Azure are pivotal to resource management, whether it's used for reporting or automation.

But sometimes, you need that extra bit of information to help discover what resources are for, or you may way to add information to a resource that isn't directly displayed in the portal, especially when complex tags are in use that might be used in automation initiatives.

This is where 'hidden' Azure Tags come in handy.

Tags starting with the prefix of 'hidden-' will not be displayed under Tags in the Azure Portal; however, they will be displayed in the resource metadata and utilised by PowerShell and Azure CLI for automation initiatives.

Examples are:

hidden-title​

As I mentioned above, every tag with 'hidden-' in front of it will be hidden in the Azure Portal. However, 'hidden-title' behaves differently.

You may have noticed that some resources in Azure, especially if the Azure ARM (Azure Resource Manager) creates them and the name is GUID based, has a '(name)' around them after the resource name; this is because of the hidden-title tag.

The hidden-title tag is especially useful for being able to pick resources that belong to a specific service or application.

An example is below:

In this case, I have used the hidden-title of 'Web Server', allowing me to quickly view what resources may be mapped to my Web Server.

You may notice that the Test-Virtual Machines title, is displayed in the Resource Groups search blade and not in the actual Resource Group, there are some areas of the Portal that will not display the hidden-title tag currently.

If I navigate to my Virtual Machine and click on the Tags blade, all I see is my CreatedBy tag.

However, if I navigate to the Overview page and click on JSON View, I can see the hidden tags in the resource metadata.

hidden tags​

Azure Portal​

You can use the Azure Portal directly to add the Tags to apply hidden tags.

You can remove the Tag by adding the hidden-tag again and keeping the value empty (ie blanking out the hidden-title will remove the title), but it will still be against the metadata (Resource Graph) as a Tag that exists (as seen in the screenshot below) - it is much cleaner to use PowerShell.

PowerShell​

Get-AzTag and Remove-AzTag, do not display the hidden tags, to add and remove the tags, you need to add them through 'Update-AzTag' and 'Replace' or 'Merge' to overwrite the tags, which requires the Resource targetted by Resource ID.

A handy snippet to use to add/remove the Tags on individual or multiple resources is:

$replacedTags = @{"hidden-title" = "Web Server"; "hidden-ShutdownAutomation" = "Yes"}
$resouceGroup = 'vm-dev-rg'
Get-AzResource -ResourceGroupName $resouceGroup | Select-Object ResourceId | Out-GridView -PassThru | Update-AzTag -Tag $replacedTags -Operation Merge

This will snippet will gather all the resources in your Resource Group, then select their Resource IDs; the script will then prompt with a GUI allowing you to select which resources or resources you want to update your tags on, then once you click Ok, it will update the Tags on the resources you selected.

You may be wondering if the Hidden tags are useful for automation, but if the 'Get-AzTag' cmdlet doesn't work, how can I retrieve the resources? It's a good question, and that is where 'Get-AzResource' comes to the rescue.

Examples are:

Get-AzResource -TagName hidden-ShutdownAutomation

Get-AzResource -TagValue Yes

$TagName = 'hidden-title'
$TagValue = 'Web Server'
Get-AzResource -TagName $TagName -TagValue $TagValue | Where-Object -FilterScript {
    $_.ResourceType -like 'Microsoft.Compute/virtualMachines' 
}

Azure Bicep​

You can also add the Tags, with Azure Bicep.

Example is:

param resourceTags object = {
   hidden-title: 'Web Server'
   hidden-ShutdownAutomation: 'Yes'
}

tags: resourceTags