I just configure the monitor driver to report 2x the cpu than it normally does. Basically when the monitor picks up the amount of cpu, lets say 1600 for a 16 core cpu, I configure the monitor to report 3200 or some other multiplier. I find scheduling with cgroups a bit hairy, I do use cpusets to jail vms to specific numa nodes and avoid cross memory domain traffic but that's about it.
On Fri, Oct 11, 2013 at 10:20 AM, Gerry O'Brien <[email protected]> wrote: > Hi, > > Is overcommitment in OpenNebula 4.2 using KVM possible? If so, how is > it done? > > Am I correct in believing that by default cgroups are not enabled and > that the only use of the CPU parameter is to guide the scheduler? If > cgroups are not enabled and the CPU paramter is set very low, say 0.1, for > a VM then can the cpu share actually used by the VM can be much larger than > 0.1? > > What is the correct interpretation of the "used CPU" value on the > Sunstone Virtual Machines view? > > Can balloon drivers be used with KVM in OpenNebula? If so, is there > any way around the fact that the MEMORY parameter seems to stop the > allocation of VMs to a host once the total sum of all MEMORY parameters > exceeds the physical memory available on the host? > > Is there any good reason why the scheduler only uses 90% of available > memory by default? > > I've tried searching the documentation and threads but cannot find any > answers to these questions so any help would be much appreciated. > > Regards, > Gerry > > -- > Gerry O'Brien > > Systems Manager > School of Computer Science and Statistics > Trinity College Dublin > Dublin 2 > IRELAND > > 00 353 1 896 1341 > > ______________________________**_________________ > Users mailing list > [email protected] > http://lists.opennebula.org/**listinfo.cgi/users-opennebula.**org<http://lists.opennebula.org/listinfo.cgi/users-opennebula.org> >
_______________________________________________ Users mailing list [email protected] http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
