Azure Resource Manager (ARM) Virtual Machine Fleet, aka: ARM-VMFleet is a PowerShell performance / load testing tool that is designed for use with Azure Stack Hub.
ARM-VMFleet uses the following components to automate load / performance testing:
- The Azure "Az" PowerShell module to automate the deployment of virtual machines (VMs) with multiple data disks, parameters exists for VM size, number of VMs and number of data disks.
- PowerShell Desired State Configuration (DSC) VM Extension that is a VM extension that needs to be syndicated (downloaded) to the target Azure Stack Hub Marketplace. DCS is used to automate all of the items below inside each VM:
- Windows Storage Spaces, the feature is enabled, and used to create a single Striped Volume made up using all available data disks added to each VM in the fleet. The stripe width is equal to the number of disks, DiskPart is used to assign the drive letter F:\. This single volume is used aggregate the IOPs of all data disks assigned to the VM, the IOPs per disk is determined based on the 'VMSize' parameter.
- Windows Performance Monitor (PerfMon) which is enabled inside each VMs Guest OS to capture a BLG file of the OS performance.
- DiskSpd is executed as part of the DSC automation, using the input parameters for the length and type of test. DiskSpd generates the Input / Output (IO) load on the F:\ drive of each VM in the Fleet.
- PowerShell Jobs enables parallelism for the VM creation process, with individual log files written to "C:\ARM-VMFleet-Logs" by default, this folder can be changed using the $logfilefolder parameter.
For testing Windows Server Hyper-Converged environments that do not have a local Azure Resource Manager (ARM) control plane available, please see the VM Fleet in the DiskSpd repository.
The Releases page shows each released version, please use the "Latest" version available to enable you to download the "Source code" (zip) file which contains all required scripts. Releases will be updated periodically, however there is no guarantee or SLA on time scales for updates. Please raise a new issue if you encounter an error or issue, or feel free to submit a PR if you would like to contribute to this project.
- Requires access to a physical Azure Stack Hub scale unit, do not run this tool on a virtual ASDK instance. Requires access to a user account with permissions to update Quotas using the Administrator Portal or Admin ARM Endpoint, as per item 2 below:
- Update the "Services and quotas" in the Plan that is linked to the Offer used to create the User Subscription. Specifically update the Compute quotas, such as "Maximum number of VMs", "Maximum number of VM cores" and the "Capacity(GB) of premium managed disk" to be greater than the total number of VMs, and Cores per VM (based on size) multiplied by number of VMs, and the number and size of Data Disks you plan to create.
- The "Windows Server 2019-Datacenter" virtual machine image must be syndicated from Azure to the Azure Stack Hub marketplace.
- The "PowerShell Desired State Configuration", Version = "2.83.1.0", Type = "Virtual Machine Extension" must be syndicated from Azure to the Azure Stack Hub marketplace.
- It is recommended to create a new / empty User Subscription, using the Offer that has the updated with the required Compute Quotas, as outlined in item 2 above.
- The workstation or device used to run the scripts, must have the Azure Az and Azure Stack PowerShell modules installed, as these are used to automate the VM creation and configuring the DSC extension.
- After completing the prerequisites above, first connect to your target Azure Stack Hub scale unit using the Az PowerShell module, example commands to execute can be found in the "_pre-req_Example_Connect_ARM.ps1" script. Note: Replace the parameters in <> brackets to match your environment prior to executing.
- Run the commands in the "_pre-req_Initialise_DSC.ps1" script in install the DSC resources locally. This script installs the required DSC Resources on your workstation or device, this is required for the scripts to publish the custom DSC package (zip file) to a storage account, which is automated using Publish-AzVMDscConfiguration in the main ARM-VMFleet.ps1.
- Open the "Run.ps1" script in VSCode or similar editor, update the parameters to match your environment, such as "location" to match the Hub scale unit region name, and the "region" and "fqdn" in the -storageUrlDomain parameter. Update the -totalVmCount and -vmsize parameters to match your requirements and save the script file.
- Execute the "Run.ps1" script, that will call "ARM_VMFleet.ps1" on line 84, passing the parameters saved from step 3 above.
- Deployment logs are saved to "C:\ARM-VMFleet-Logs" on the device executing the scripts. It is expected to take 1 to 2 hours to complete, if using 30 to 50 VMs, each with 30 to 50 x Data Disks each.
Do NOT use ARM-VMFleet on a production Azure Stack Hub scale unit, as this could cause significant additional load on the scale unit, impacting other workload for the duration of the test(s).
This sample is not supported under any Microsoft standard support program or service. The script is provided "AS IS" without warranty of any kind. Microsoft further disclaims all implied warranties including, without limitation, any implied warranties of merchantability or of fitness for a particular purpose. The entire risk arising out of the use or performance of the sample and documentation remains with you. In no event shall Microsoft, its authors, or anyone else involved in the creation, production, or delivery of the script be liable for any damages whatsoever (including, without limitation, damages for loss of business profits, business interruption, loss of business information, or other pecuniary loss) arising out of the use of or inability to use the sample or documentation, even if Microsoft has been advised of the possibility of such damages, rising out of the use of or inability to use the sample script, even if Microsoft has been advised of the possibility of such damages.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.