TL;DR: WatchCluster.psm1 from Microsoft's DiskSpd repository provides a live, refreshing console view of cluster-wide CSV, S2D, SMB, and Hyper-V performance counters — useful for real-time troubleshooting of storage performance, IOPS, latency, and bandwidth across all Azure Local nodes from a single window.

Example output:

Recommended action:

  1. From a management workstation or one of the cluster nodes, open an elevated PowerShell window.

  2. Download the module. Either method works:

    Option A — direct download with Invoke-WebRequest:

    Invoke-WebRequest `
  -Uri "https://raw.githubusercontent.com/microsoft/diskspd/master/Frameworks/VMFleet/WatchCluster.psm1" `
  -OutFile "$env:TEMP\WatchCluster.psm1"

    Option B — download or clone from GitHub:

    Browse to https://github.com/microsoft/diskspd/blob/master/Frameworks/VMFleet/WatchCluster.psm1 and use the raw-file download, or run git clone https://github.com/microsoft/diskspd.git and pull the file from Frameworks/VMFleet/.

  3. Import the module:

    Import-Module "$env:TEMP\WatchCluster.psm1" -Force

  4. Launch the live CSV FS view (default):

    Watch-FleetCluster -Cluster <ClusterName> -SampleInterval 2

    Omit -Cluster (or pass .) to target the local cluster.

  5. To watch a different set of counters, use -Sets. Examples:

    # S2D bandwidth view (CSV, SBL, cache, disk tiers side-by-side)
Watch-FleetCluster -Cluster <ClusterName> -Sets "S2D BW"

# Hyper-V CPU utilization across all nodes
Watch-FleetCluster -Cluster <ClusterName> -Sets "Hyper-V LCPU"

# Multiple panels at once
Watch-FleetCluster -Cluster <ClusterName> -Sets "CSV FS","S2D BW","Hyper-V LCPU"

# Everything (as shown in the screenshot)
Watch-FleetCluster -Cluster <ClusterName> -Sets "*"

  6. Press Ctrl+C to stop.

Why:

The script fans out Get-Counter jobs to every cluster node via PowerShell remoting, aggregates the samples into a single refreshing console view, and self-heals when a node drops or a sampling job stalls. It is the same monitoring surface VM Fleet uses during DiskSpd benchmark runs, but does not require a VM Fleet workload to be active — it works against any live Azure Local / Storage Spaces Direct cluster.

Going forward:

Run from a dedicated PowerShell window — the tool calls Clear-Host on every refresh, so anything else in that window will be wiped. For long capture sessions, add -LogFile C:\Temp\watch.log to capture diagnostic messages about missing counters or downed nodes. After re-downloading the module following an upstream update, re-import with -Force to pick up changes.

Optional details:

Prerequisites on the machine running the tool:

  • PowerShell 5.1 or later (the module uses PowerShell classes).

  • Failover Clustering PowerShell module (Import-Module FailoverClusters — included with RSAT-Clustering-PowerShell, or installed by default on cluster nodes).

  • PowerShell Remoting (WinRM) enabled to all cluster nodes, with administrative rights.

Available -Sets values and what each panel shows:

SetShows
CSV FS (default)Cluster Shared Volume IOPS, bandwidth, latency, queue depth
S2D BWBandwidth at each tier: CSV → SBL → cache → physical disk
SBL, SBL Local, SBL Remote, SBL*Storage Bus Layer disk counters, total or split by local vs. remote traffic
SSB CacheStorage Spaces cache hit/miss, destage, page state (standby/dirty/L0/L1/L2)
SMB SRVSMB server-side share IOPS and bandwidth
SMB TransportSMB client-side request rate, including RDMA-transmitted requests
Hyper-V LCPULogical and root virtual processor utilization across all nodes
*All panels at once

Default parameter values: -SampleInterval 2 (seconds), -Cluster . (local cluster), -Sets "CSV FS".

The module is self-contained — it does not require the rest of the VM Fleet package, and dot-sourcing the .psm1 file (. .\WatchCluster.psm1) is a valid alternative to Import-Module for one-off use.

Source: https://github.com/microsoft/diskspd/blob/master/Frameworks/VMFleet/WatchCluster.psm1