[one-users] Resource Overcommit / Provisioning Limitations

[Simon Boulet]

Hi,

On Mon, Apr 29, 2013 at 5:18 AM, Carlos Martín Sánchez <
cmartin at opennebula.org> wrote:

>
> On Sun, Apr 28, 2013 at 12:18 AM, Jon <three18ti at gmail.com> wrote:
>
>> Hello All,
>>
>> I'm only able to provision resources up to 750 / 800 CPU and 7GB / 7.8GB.
>> VMs provisioned after this limit has been reached causes the VMs to stay
>> in "pend" status until vms are turned off or destroyed.
>>
>
> The scheduler will reserve some memory for the hypervisor, this can be
> configured with the HYPERVISOR_MEM attribute in sched.conf [1].
>
>
>
>>   Even though 7GB has been provisioned, I'm hardly using 2G.
>>
>> >> HOST SHARES
>>
>> >> TOTAL MEM             : 7.8G
>> >> USED MEM (REAL)       : 1.9G
>> >> USED MEM (ALLOCATED)  : 7G
>> >> TOTAL CPU             : 800
>> >> USED CPU (REAL)       : 44
>> >> USED CPU (ALLOCATED)  : 750
>> >> RUNNING VMS           : 10
>>
>> How can I configure OpenNebula to allow overcommit of resources?
>>
>>
> Currently there is no way to do Memory over-commitment, but you can do cpu
>  over-commitment with the CPU & VCPU attributes
>


One trick is to set HYPERVISOR_MEM in sched.conf to a negative value, ex. :

HYPERVISOR_MEM = -0.4

This will allow for over provisioning memory by 40%. Beware that it will
not check actual memory usage on the host so setting this limit too high
will most likely crash the host or prevent VM from being deployed.

Simon
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20130429/6348f591/attachment-0002.htm>