Terminate Azure Agent of Scale Set after a period of time

Question

I am currently trying to find a way to monitor agents in Azure that were scaled by Azure Scale-Sets. To avoid costs I want to kill every agent that has been running a job for longer than 1 hour. Is there a solution from Azure that would help me in this case?

I was thinking about using alerts and Azure functions but I dont know which alert signal to use since they all only consider metrics from the vmss and not a single agent.

Thank you

Answer 1

Terminate Azure Agent of Scale Set after a period of time

You can use an Azure Automation runbook to check and kill the Agent Service every hour. To create a runbook and schedule the job, you can follow the Stack link that I answered.

Here is thePowerShell script can monitor and terminate the Azure Monitor Agent Service if it runs longer than an hour.

    az login --identity --username "ed65fffjfnjff fhc01dcf5"
    
    # Define the script content
    $scriptContent = @"
    # Get the Azure Monitor Agent process
    \$process = Get-Process "AzureMonitorAgentService" -ErrorAction SilentlyContinue
    
    # Check if the process exists
    if (\$process) {
        # Get the process start time
        \$processStartTime = \$process.StartTime
        # Get the current time
        \$currentTime = Get-Date
    
        # Calculate the elapsed time since the process started
        \$elapsedTime = New-TimeSpan -Start \$processStartTime -End \$currentTime
    
        # Check if the elapsed time is greater than 1 hour
        if (\$elapsedTime.TotalHours -gt 1) {
            # Stop the process
            Stop-Process -Name AzureMonitorAgentService -Force
    
            Stop-Service -Name Azure Monitor Agent -Force -Confirm
    
            Write-Output "Azure Monitor Agent Service has been stopped."
        } else {
            Write-Output "Azure monitor Service has not been running for more than 1 hour."
            Write-Output "Process Start Time: " + \$process.StartTime
            Write-Output "Current Time: " + \$currentTime
            Write-Output "Elapsed Time: " + \$elapsedTime
        }
    } else {
        Write-Output "Azure Monitor Agent process is not running."
    }
    "@
    
    # Define the resource group and VMSS name
    $resourceGroup = "venkat-vmss_group"
    $vmssName = "venkat-vmss"
    
    # Get the list of instance IDs in the VMSS
    $instanceIds = az vmss list-instances --resource-group $resourceGroup --name $vmssName --query "[].instanceId" -o tsv | ConvertFrom-Json
    
    # Loop through each instance ID and execute the script
    foreach ($instanceId in $instanceIds) {
        Write-Host "Running script on VMSS instance $instanceId ..."
        az vmss run-command invoke `
            --resource-group $resourceGroup `
            --name $vmssName `
            --instance-id $instanceId `
            --command-id RunPowerShellScript `
            --scripts "$scriptContent"
    }
    
    Write-Host "Script execution completed for all VMSS instances."

To schedule the Job and check the Agent service status hourly, you can follow the steps shown in the snapshot below.

Output: