LucD notes

In our PowerCLI book we presented a Delete-Harddisk function in Chapter 7.

One of our readers asked if that function could be used to remove orphaned VMDK files from one or more datastores. Now unfortunately that is not the case since the function we presented in chapter 7 uses the ReconfigVM_Task method to remove the harddisk.

In the PowerCLI Community there are some thread that provide scripts to report on orphaned VMDK files, but most of these are quite old.

So I decided to write a new script that would report on orphaned folders and VMDK files and that would have an option to remove these folders and files.

Update March 19th 2013

The following changes were made to the function

• Fixed the problem where a VMDK in the root of a datastore, which you shouldn’t have of course, would erase the complete datastore content (ouch). Thanks Olivier
• Added WhatIf functionality. You can now check what the Delete switch will do before actually removing files and folders.
• Added details for lone VMDK. See the properties of the returned objects.

Update May 22nd 2011

Two corrections in this new version of the script.

• NiTRo reported a problem with guests that had only a .vmdk, and not their .vmx, file located on the datastore(s) to be cleaned
• The COS became a virtual machine with ESX 4.x and thus has a .vmdk file as well. Although the COS files should be located on a local datastore, the script now ignores COS .vmdk files when it encounters them on the datastore(s) to be cleaned.

The Script

function Remove-OrphanedData {
<#
.SYNOPSIS   Remove orphaned folders and VMDK files
.DESCRIPTION   The function searches orphaned folders and VMDK files
   on one or more datastores and reports its findings.
   Optionally the function removes  the orphaned folders   and VMDK files
.NOTES   Author:  Luc Dekens
.PARAMETER Datastore
   One or more datastores.
   The default is to investigate all shared VMFS datastores
.PARAMETER Delete
   A switch that indicates if you want to remove the folders
   and VMDK files
.EXAMPLE
   PS> Remove-OrphanedData -Datastore ds1
.EXAMPLE
  PS> Get-Datastore ds* | Remove-OrphanedData
.EXAMPLE
  PS> Remove-OrphanedData -Datastore $ds -Delete
#>

  [CmdletBinding(SupportsShouldProcess=$true)]

  param(
  [parameter(Mandatory=$true,ValueFromPipeline=$true)]
  [PSObject[]]$Datastore,
  [switch]$Delete
  )

  begin{
    $fldList = @{}
    $hdList = @{}

    $fileMgr = Get-View FileManager
  }

  process{
    foreach($ds in $Datastore){
      if($ds.GetType().Name -eq "String"){
        $ds = Get-Datastore -Name $ds
      }
      if($ds.Type -eq "VMFS" -and $ds.ExtensionData.Summary.MultipleHostAccess){
        Get-VM -Datastore $ds | %{
          $_.Extensiondata.LayoutEx.File | where{"diskDescriptor","diskExtent" -contains $_.Type} | %{
            $fldList[$_.Name.Split('/')[0]] = $_.Name
            $hdList[$_.Name] = $_.Name
          }
        }
        Get-Template | where {$_.DatastoreIdList -contains $ds.Id} | %{
          $_.Extensiondata.LayoutEx.File | where{"diskDescriptor","diskExtent" -contains $_.Type} | %{
            $fldList[$_.Name.Split('/')[0]] = $_.Name
            $hdList[$_.Name] = $_.Name
          }
        }

        $dc = $ds.Datacenter.Extensiondata

        $flags = New-Object VMware.Vim.FileQueryFlags
        $flags.FileSize = $true
        $flags.FileType = $true

        $disk = New-Object VMware.Vim.VmDiskFileQuery
        $disk.details = New-Object VMware.Vim.VmDiskFileQueryFlags
        $disk.details.capacityKb = $true
        $disk.details.diskExtents = $true
        $disk.details.diskType = $true
        $disk.details.thin = $true

        $searchSpec = New-Object VMware.Vim.HostDatastoreBrowserSearchSpec
        $searchSpec.details = $flags
        $searchSpec.Query += $disk
        $searchSpec.sortFoldersFirst = $true

        $dsBrowser = Get-View $ds.ExtensionData.browser
        $rootPath = "[" + $ds.Name + "]"
        $searchResult = $dsBrowser.SearchDatastoreSubFolders($rootPath, $searchSpec)
        foreach($folder in $searchResult){
          if($fldList.ContainsKey($folder.FolderPath.TrimEnd('/'))){
            foreach ($file in $folder.File){
              if(!$hdList.ContainsKey($folder.FolderPath + $file.Path)){
                New-Object PSObject -Property @{
                  Folder = $folder.FolderPath
                  Name = $file.Path
                  Size = $file.FileSize
                  CapacityKB = $file.CapacityKb
                  Thin = $file.Thin
                  Extents = [string]::Join(',',($file.DiskExtents))
                }
                if($Delete){
                  If ($PSCmdlet.ShouldProcess(($folder.FolderPath + " " + $file.Path),"Remove VMDK")){
                    $dsBrowser.DeleteFile($folder.FolderPath + $file.Path)
                  }
                }
              }
            }
          }
          elseif($folder.File | where {"cos.vmdk","esxconsole.vmdk" -notcontains $_.Path}){
            $folder.File | %{
              New-Object PSObject -Property @{
                Folder = $folder.FolderPath
                Name = $_.Path
                Size = $_.FileSize
                CapacityKB = $_.CapacityKB
                Thin = $_.Thin
                Extents = [String]::Join(',',($_.DiskExtents))
              }
            }
            if($Delete){
              if($folder.FolderPath -eq $rootPath){
                $folder.File | %{
                  If ($PSCmdlet.ShouldProcess(($folder.FolderPath + " " + $_.Path),"Remove VMDK")){
                    $dsBrowser.DeleteFile($folder.FolderPath + $_.Path)
                  }
                }
              }
              else{
                If ($PSCmdlet.ShouldProcess($folder.FolderPath,"Remove Folder")){
                  $fileMgr.DeleteDatastoreFile($folder.FolderPath,$dc.MoRef)
                }
              }
            }
          }
        }
      }
    }
  }
}

Annotations

Line 22, 91, 110, 116: These handle the WhatIf switch

Line 26: The function accepts datastore by name or as a DatastoreImpl object. Hence the PSObject type. There has to be a Datastore argument !

Line 39-41: Cheap OBN implementation.

Line 42: The test makes sure only VMFS datastores that are shared are handled.

Line 43-48: The function creates a hash table with all the files that are used by the virtual machines on the datastore. The $fldList hash table contains the folders and the $hdList hash table contains all the disk related files (they have the diskDescriptor or the diskExtent type).

Line 49-54: The Get-Template cmdlet doesn’t have a Datastore parameter, that’s why this block uses a different method of determining if there are Templates on a specific datastore.

Line 58-75: The parameters for the SearchDatastoreSubFolders method are prepared and at the end the method is called.

Line 77-105: The returned results from the SearchDatastoreSubFolders method are analysed.

Line 87,103: The Extents property will contain all related filenames of the VMDK file. For example, a file called MyVM.vmdk will have an extent called MyVM-flat.vmdk.

Line 89-93: If the Delete switch is used, the script tests if the WhatIf switch was also specified. If not, the VMDK file is deleted.

Line 109: This condition handles the case when there is an orphaned VMDK file in the root of the datastore.

Line 110: For an orphaned VMDK in the root of the datastore, the DeleteFile method is used.

Line 118: For a folder with orphaned VMDK files, the DeleteDatastoreFile method is used

Sample runs

The function will only scan shared VMFS datastores for orphaned folders and files.

Remove-OrphanedData -Datastore DS1

This call will display all found folders and files and will not delete them.

When you add the Delete switch, the function will return the same information, but the folders and files will be deleted.

Remove-OrphanedData -Datastore DS1 -Delete

When you combine the Delete switch with the WhatIf switch, not files or folders will be deleted. But the function will show what actions will be taken without

You can specify more than one value with the Datastore parameter, the function will investigate all the datastores you specified.
You can pass the datastores by name

Remove-OrphanedData -Datastore "ds1","ds2"

or by object

$ds = Get-Datastore "ds1"
Remove-OrphanedData -Datastore $ds

And the function can of course also be used in a pipeline construction

Get-Datastore ds* | Remove-OrphanedData -Delete

Make absolutely sure that you know what you are doing before using the Delete switch.

Use the WhatIf switch before actually deleting anything !

Enjoy !