When using OpenStack Compute with Citrix XenServer or XCP hypervisor, OpenStack Compute should be installed in a virtual machine running on your hypervisor, rather than installed directly on the hypervisor, as you would do when using the Libvirt driver. For more information see: XenAPI Install.
Given how you should deploy OpenStack with XenServer, the first step when setting up the compute nodes in your OpenStack cloud is to install XenServer and install the required XenServer plugins. You can install XCP by installing Debian or Ubuntu, but generally rather than installing the operating system of your choice on your compute nodes, you should first install XenServer. For more information see: XenAPI Deployment Architecture.
Once you have installed XenServer and the XenAPI plugins on all your compute nodes, you next need to create a virtual machine on each of those compute nodes. This must be a Linux virtual machine running in para-virtualized mode. It is inside each of these VMs that you will run the OpenStack components. You can follow the previous distribution specific instructions to get the OpenStack code running in your Virtual Machine. Once installed, you will need to configure OpenStack Compute to talk to your XenServer or XCP installation. For more information see: Introduction to Xen.
