I' using rhcs, but I'm trying to setup a sort of actve active config. Where I have also clvmd and where I can have for example some LVs of a clustered VG active as disks of some VMs hosted on one server, while other LVs of this VG represent disks hosted on another host. Also, as I have no SAN in my test env I'm using drbd 8.3.2 in dual primary mode...
Before starting with the cluster arch I had been done tests with Qemu/Kvm and virt-manager using a standalone host. Then without restarting from zero and deleting things, I transformed my local VG into a drbd based one and I cloned the initial host
So the question arised about uuid values inside xml files... and for example if they have to be the same for storage config and network config or not...
I would like not to use GFS, so that I don't add another layer impacting performance and complexity.... But your saying I can use GFS for config does implicitly mean same config files, also in active-active config? Gianluca
On Sat, Jul 25, 2009 at 3:17 AM, Guido Güntheragx@sigxcpu.org wrote:
On Mon, Jul 20, 2009 at 10:45:56AM +0200, Gianluca Cecchi wrote:
Hello, is there any pointer about how to set up a cluster of Qemu/KVM hosts? I'm on F11 x86_64 for the hosts. I read requirements for KVM migration in general highlighted at http://www.linux-kvm.org/page/Migration but I keep having doubts, in particular regarding libvirt configuration in cluster environments... Some questions about libvirt configuration:
- /etc/libvirt/qemu/networks
do the files inside here be identical between hosts (default.xml and other custom configured network ones)? What are the uuid tags into the xml for? Do they have to be identical for clusters or do they have to be absolutely different for a sort of "identification" of host (as the term seems to suggest)? If I have guests that use virbr0 and use dhcp for their network config, and want two different qemu/kvm hosts able to migrate these guests between them, need to have same config or what?
- the same for storage configuration and other parameters.
You can use rhcs or pacemaker for this and share the vm configuration via e.g. gfs. Cheers, -- Guido