On 28/10/16 17:46 +0100, david caughey wrote:
Hi,
I'm building a homeserver to run ovirt and wanted to get opinions on the best approach. The server will be used as a test/studybed for ovirt/kvm/vcloud/openstack/ceph. The server will be based around a Xeon E5 10 core with 128GB ram. Option 1: Build server with CentOS 7.2 and deploy ovirt directly on top.
Perfect if you have/expect multiple machines. How do you expect to handle storage?
Option 2: Build server with CentOS 7.2 and deploy multiple ovirt instances on top of KVM.
Perfect for single machine setup.
Which will be the most stable versatile method? If a GPU is used as a passthrough device can it be used on several vm's or is it restricted to 1 vm?
Single card for single VM meaning 1:1 mapping only for now.
If 2 GPU's are used can 1 be used as a dedicated passthrough to 1 vm and the other shared between the remaining vm's?
As stated above, no. Most typical home lab setup uses 2 GPUs, one for hypervisor and other one to be assigned to a VM.
Is CentOS/RH the best platform for ovirt?
For playing around and general usage, yes (in my opinion :) ).
Is it okay/advisable to load the latest kernel, (4.8 ish), on to CentOS before installing ovirt?
I wouldn't really do that, you'd have to bump up version of KVM, possibly QEMU and libvirt -- in the end you'd end up running fedora. mpolednik
Any and all comments/advice welcome,
David
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users