Skip to content

VMFleet ​

Tool: VMFleetStatus: Not proven on hardwareVMFleet 2.1.0.0

NOTE

Status: implemented and unit-tested, not proven on hardware. See the implementation status table. Command examples on this page show the intended workflow; the param block of each script is the authority for parameter names. Not proven on a live cluster.

VMFleet deploys a fleet of lightweight Server Core VMs across your Azure Local cluster nodes, each running DiskSpd workloads to stress-test Storage Spaces Direct (S2D). It measures IOPS, throughput, and latency under realistic storage load.


Run This Test ​

Prerequisites ​

  • Azure Local cluster with Storage Spaces Direct enabled
  • PowerShell 7.2+ on your management workstation
  • WinRM access to all cluster nodes
  • A Server Core base VHD on the cluster CSV
  • Framework prerequisites completed

The VMFleet orchestrator runs all phases end-to-end with checkpoint-based resume:

powershell
# From the repository root
.\tools\vmfleet\Invoke-VMFleetPipeline.ps1 `
    -ClusterConfigPath "config\clusters\my-cluster.yml" `
    -ProfilePath "tools\vmfleet\config\profiles\general.yml" `
    -FleetAdminCredential $fleetAdmin   # see Credential Management; for example Get-ManagedCredential -Name 'vmfleet_admin'

If the pipeline fails mid-run, resume from the last checkpoint:

powershell
.\tools\vmfleet\Invoke-VMFleetPipeline.ps1 `
    -ClusterConfigPath "config\clusters\my-cluster.yml" `
    -Resume

Option B — Run Each Step Manually ​

Run each phase independently. This is useful for debugging, re-running a single phase, or learning how the pipeline works.

1. Install VMFleet on the cluster

powershell
.\tools\vmfleet\scripts\Install-VMFleet.ps1 `
    -ClusterName "hci01.corp.infiniteimprobability.com" `
    -Nodes @("hci01-node1", "hci01-node2")

2. Deploy fleet VMs

powershell
.\tools\vmfleet\scripts\Deploy-VMFleet.ps1 `
    -ClusterName "hci01.corp.infiniteimprobability.com" `
    -Nodes @("hci01-node1", "hci01-node2") `
    -VMCountPerNode 4 `
    -VMMemoryGB 2

3. Start a workload test

powershell
.\tools\vmfleet\scripts\Start-VMFleetTest.ps1 `
    -ClusterName "hci01.corp.infiniteimprobability.com" `
    -Profile "General" `
    -DurationSeconds 600

4. Monitor in real-time (run in a separate terminal)

powershell
.\tools\vmfleet\scripts\Watch-VMFleetMonitor.ps1 `
    -ClusterName "hci01.corp.infiniteimprobability.com"

5. Stop the test (if needed before duration completes)

powershell
.\tools\vmfleet\scripts\Stop-VMFleetTest.ps1 `
    -ClusterName "hci01.corp.infiniteimprobability.com"

6. Collect and aggregate results

powershell
.\tools\vmfleet\scripts\Collect-VMFleetResults.ps1 `
    -ClusterName "hci01.corp.infiniteimprobability.com" `
    -Nodes @("hci01-node1", "hci01-node2") `
    -RunId "<run-id>"

7. Clean up fleet VMs (optional)

powershell
.\tools\vmfleet\scripts\Remove-VMFleet.ps1 `
    -ClusterName "hci01.corp.infiniteimprobability.com"

Available Workload Profiles ​

ProfileWrite %Random %Block SizeTarget IOPSUse Case
General30%70%8K~5,000Balanced mixed workload baseline
Peak IOPS0%100%4K~50,000Maximum random read throughput
VDI20%80%8K~3,000Virtual desktop simulation
SQL OLTP30%100%8K~10,000SQL Server transaction processing
Sequential Throughput100%0%512KN/A (MB/s)Sequential write ceiling

Profile YAML files are in tools/vmfleet/config/profiles/.


What You Get ​

After a test completes, you will have:

  • Raw DiskSpd XML per VM per node in tools/vmfleet/logs/
  • Aggregated JSON metrics with IOPS, throughput (MB/s), and latency (ms) per profile
  • Reports in PDF, DOCX, and/or XLSX format (if --GenerateReports was used)

File Locations ​

ComponentPath
Orchestratortools/vmfleet/Invoke-VMFleetPipeline.ps1
Scriptstools/vmfleet/scripts/
Workload Profilestools/vmfleet/config/profiles/
Monitoring Collectorstools/vmfleet/monitoring/collectors/
Alert Rulestools/vmfleet/monitoring/alerts/alert-rules.yml
Azure Monitor Workbooktools/vmfleet/monitoring/workbooks/vmfleet-workbook.json
Report Templatestools/vmfleet/reports/templates/
Logstools/vmfleet/logs/

Deep Dive ​

TopicPage
Cluster and VMFleet requirementsPrerequisites
VM deployment detailsDeployment
Profile parameters and tuningWorkload Profiles
Metric collection and dashboardsMonitoring
Report formats and templatesReporting
Common errors and fixesTroubleshooting

Automate This Test ​

To run VMFleet on a schedule or as part of a CI/CD workflow, see CI/CD Pipelines. The run-vmfleet.yml GitHub Actions workflow calls the same Invoke-VMFleetPipeline.ps1 script above with parameters from workflow dispatch inputs.


More VMFleet guides ​

Released under the MIT License.