PowerCLI to get overly provisioned ESXi hosts in Data Center ?

[AlbertWT]

Hi All,

Can anyone here please assist me with the pwoershell script that can shows which ESXi hosts and its VM is overly provisioned in terms of vCPU which affecting the VM performance or with CPU ready % greater than 5 ?

Thanks

[LucD]

What would you take as a criteria to determine if a VM is vCPU overprovisioned ?

Would cpu.usage.average be a candidate ?

Which thresholds would be taken and over which interval ?

[deang1609]

I recently created a script to retrieve a statistic and report the peak, 99th and 95th percentile (nearest rank) over a sample period and extract to a CSV file.

By default the script will retrieve data over a sample period of 30 days, however you can specify the sample period as a mandatory parameter ($Days).

One thing to consider when specifying the metric, is the statistical level, interval duration and save for period. For Example, if you decide to retrieve the 'cpu.usage.metric' which is a level 1 statistic this will retrieve date at two hour intervals by default.

Param ([Parameter(Mandatory=$true)] [string] $vCenter, [Parameter(Mandatory=$true)][string] $Output,[string] $Days = "30") 

Connect-VIServer $vCenter 

$VMs = Get-VM 

$Results = ForEach ($VM in $VMs)    
      {    

  $Metric = (Get-Stat -Entity $VM -Stat 'cpu.usage.average' -Start (Date).AddDays(-$Days) -Finish (Date) ).Value | Sort -Descending 

 "" | Select @{N="Name";E={$VM}},    
 @{N="CPU Usage (Peak)";E={ $MetricRow = ($Metric.Count) - ([Math]::Round($Metric.Count * 1))     
 [Math]::Round(($Metric[$MetricRow]),2)}},     
 @{N="CPU Usage (p99)"; E={$MetricRow = ($Metric.Count) - ([Math]::Round($Metric.Count * 0.99))     
 [Math]::Round(($Metric[$MetricRow]),2)}},     
 @{N="CPU Usage (p95)";E={$MetricRow = ($Metric.Count) - ([Math]::Round($Metric.Count * 0.95))     
 [Math]::Round(($Metric[$MetricRow]),2)}}         

     } 

$Results | Export-Csv -Path $Output -NoTypeInformation 

[AlbertWT]

Hi Luc,

What about the count total of VMs with CPU ready greater than 10% ?

would that be make more sense ?

I got this script from you long time ago:

$vms = Get-VM

$start = (Get-Date).AddDays(-30)

$metric = "cpu.ready.summation"

Get-Stat -Entity $vms -Stat $metric -Start $start | Group-Object -Property {$_.Entity.Name} | %{

  New-Object PSObject -Property @{

    VM = $_.Values[0]

    ReadyAvg = &{

      $interval = $_.Group[0].IntervalSecs * 1000

      $value = $_.Group | Measure-Object -Property Value -Average | Select -ExpandProperty Average

      "{0:p}" -f ($value/$interval)

    }

  }

}

but it doesn't list the ESXi servers where the VMs is residing.

[AlbertWT]

Dean,

Thanks for the reply with the script. Somehow I cannot see the CPU Usage (Peak), CPU Usage (p99), CPU Usage (p95) column values, it's all blank ?

I can only see the first column which is the VM name itself.

Do I need to set / enable certain logging level to get the previous 30 days value ?

[LucD]

You should be able to get the ESXi server that hosts a VM like this

$vms = Get-VM

$start = (Get-Date).AddDays(-1)

$metric = "cpu.ready.summation"

Get-Stat -Entity $vms -Stat $metric -Start $start | Group-Object -Property {$_.Entity.Name} | %{

  New-Object PSObject -Property @{

     VM = $_.Values[0]

     ESX = $_.Group[0].Entity.VMHost.Name

     ReadyAvg = &{

         $interval = $_.Group[0].IntervalSecs * 1000

         $value = $_.Group | Measure-Object -Property Value -Average | Select -ExpandProperty Average

         "{0:p}" -f ($value/$interval)

     }

  }

}

[AlbertWT]

Luc,

thanks for the assistance, how can I export the result into CSV file ?

[LucD]

Use the trick with the call operator

$vms = Get-VM

$start = (Get-Date).AddDays(-1)

$metric = "cpu.ready.summation"

&{Get-Stat -Entity $vms -Stat $metric -Start $start | Group-Object -Property {$_.Entity.Name} | %{

  New-Object PSObject -Property @{

     VM = $_.Values[0]

     ESX = $_.Group[0].Entity.VMHost.Name

     ReadyAvg = &{

         $interval = $_.Group[0].IntervalSecs * 1000

         $value = $_.Group | Measure-Object -Property Value -Average | Select -ExpandProperty Average

         "{0:p}" -f ($value/$interval)

     }

  }

}} | Export-Csv report.csv -NoTypeInformation -UseCulture

[deang1609]

I would expect values to be retrieved as this is a level 1 metric, if you was to invoke the below agaisnt one of the VMs (replacing $VM) do receive output?

Get-Stat -Entity $VM -Stat 'cpu.usage.average' -Start (Date).AddDays(-30) -Finish (Date)

You should receive output similar to the below sample.

             Value Timestamp          MetricId           Unit               Description        Entity             EntityId                 IntervalSecs Instance      

             ----- ---------          --------           ----               -----------        ------             --------                 ------------ -------- 

              5.16 24/10/2014 15:0... cpu.usage.average  %                  CPU usage as a ... vm1            VirtualMachine-...               7200                   

              5.22 24/10/2014 13:0... cpu.usage.average  %                  CPU usage as a ... vm1            VirtualMachine-...               7200                  

              4.29 24/10/2014 11:0... cpu.usage.average  %                  CPU usage as a ... vm1            VirtualMachine-...               7200                  

              2.08 24/10/2014 09:0... cpu.usage.average  %                  CPU usage as a ... vm1            VirtualMachine-...               7200