Home / server / Questions / 384184
Accepted
Lord Loh.
Asked: 2012-04-29 21:45:58 +0800 CST

Adding RAM to a virtual machine without restart

  • 772

Could anyone suggest me a hypervisor that permits adding RAM to a running virtual machine without restarting the guest OS. Is this even possible? Can this be done on xen? I have never found any documentation for such a thing.

virtualization hypervisor

5 Answers

  1. You can hot-add memory with vSphere. However, there is another question you should be asking: "What OS supports the hot addition of RAM?" The guest OS will have to have the ability to see RAM that has been added without a reboot. For example, Windows Server 2003 Enterprise (not standard) supports this, and all Server 2008 versions do as well.

    • 5
  2. Best Answer

    Xen,KVM,Hyper-V,Vmware all support increasing RAM with out restarting.

    Its done through balloon drivers.

    check for its support on Guest

    cat /boot/config-`uname -r` |grep -i balloon

    In my VM it shows 

    CONFIG_VMWARE_BALLOON=m

CONFIG_XEN_BALLOON=y

CONFIG_HYPERV_BALLOON=m

CONFIG_VIRTIO_BALLOON=m

    But for ballooning RAM up,you need to set max RAM on guests,with out this value you can only decrease RAM but can not increase RAM.(at least in Xen and KVM it is so not sure how to do it in Vmware and Hyper-V)

    • 5

  3. As nice as the hot-add option is, I wouldn't base a hypervisor choice off of that feature. Are you unable to acquire the downtime needed to make a quick configuration change to size your VM's RAM appropriately?

    • 4

  4. See my question and the answers/comments. Yes. XEN can do this. But it depends - both ends have to support it.

    • 2

  5. When the VM was created, you must have enabled the "Virtual CPU hot add" and "Memory hot add" features.

    The VM guest OS must support hot-add for CPU and RAM.

    The VM guest OS must have VM Tools installed.

    If your VM does not have the 'Virtual CPU hot add" and "Memory hot add" features enabled, you must power off the VM to enable them.

    • 0