Change/reduce virtual CPU count after initialization


  • Does anyone know/has tried reducing the virtual CPU count on a VM (MS Server 2008 R2 Datacenter) originally configured for 2 CPU cores down to one CPU core?

    The reason for my question is that I'd rather not spend the time building another install if I can simply copy/clone an instance in which I've already invested the time.

    Monday, March 26, 2012 4:48 PM


All replies

  • you can change the number of vCPUs in your machine settings.

    not sure what you mean with copy/cloning, to use multiple instances of the same installation, you would have to sysprep the original installation ( after that you could use copies of your sysprepped vhd for multiple vm's (assign a copied vhd to a new vm)

    • Marked as answer by mlwest Monday, March 26, 2012 5:12 PM
    Monday, March 26, 2012 4:59 PM
  • Yes, I meant using the sysprepped vhd - just trying to cut down on the time spent patching the distribution from MS.

    That and I have the Datacenter licensed for both of my CPUs to run as many Datacenter OSes as my hardware will tolerate.

    Thanks for the update.

    • Marked as answer by mlwest Monday, March 26, 2012 5:12 PM
    • Unmarked as answer by mlwest Monday, March 26, 2012 5:12 PM
    Monday, March 26, 2012 5:12 PM
  • It's important that when you copy or clone VHDs that run run sysprep on them first.  Otherwise you could end up with duplicate SIDs and other bad things. 

    When talking about duplicating a file, copy and clone are nearly synonymous.  When I copy a file, under the covers I create a new file within NTFS and then do a bit for bit copy of the data.  When I clone a file, at least in the cmdlets I wrote, I create a new file in NTFS, then use underlying storage features to manipulate pointers instead of physically moving data.  In the Data ONTAP PowerShell Toolkit, the cmdlet to do this is Copy-NaHostFile. 

    Where a bit for bit copy of a 127GB file takes 15 or 20 minutes, cloning the same file only takes 30 seconds or so.  Fron the OS perspective, the result is indistinguishable.  From the storage perspective, it leverages the same pointer mechanisms used by snapshots and dedupe, just in a different way.  From the storage perspective, the data is both thin and deduped. 

    Either copy or clone of a sysprepped image can be integrated into a SCVMM or Orchestrator workflow to include the steps necessary to finish the deployed VM after initally creating it from the sysprepped image.  He's a video of one such integration that from an input ticket will rapidly deply completed VMs.

    Monday, March 26, 2012 6:25 PM