Re: Review Request 12679: CLOUDSTACK-904: Changed multiple vcpus to one vcpu with multiple sockets

[Animesh Chaturvedi <An...@citrix.com>]

> On July 17, 2013, 4:43 p.m., Wido den Hollander wrote:
> > Hmm, I'm not sure about this one.
> > 
> > In the UI we indeed mention the amount of cores, however, the API parameter is "cpunumber".
> > 
> > So I get the reasoning, but we would be changing a lot here. I don't know if this changes anything by the way Qemu spreads out threads over the hypervisor's cores?
> 
> Rene Diepstraten wrote:
>     Perhaps it's best if it's decided that a complete review of the cpu cores model is needed.
>     The API should in my opinion accept both cpu sockets and cpu cores as parameters. It should for example be possible to create a vm with 2 sockets and 16 cores.
>     Qemu/KVM does support this for quite some time, as do for example Red Hat Enterprise Virtualization and Proxmox VE, which both use KVM .
>

May be this should be brought to mailing list as a DISCUSS topic?


- Animesh


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/12679/#review23290
-----------------------------------------------------------


On July 18, 2013, 4:06 p.m., Rene Diepstraten wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/12679/
> -----------------------------------------------------------
> 
> (Updated July 18, 2013, 4:06 p.m.)
> 
> 
> Review request for cloudstack and Wido den Hollander.
> 
> 
> Bugs: CLOUDSTACK-904
>     https://issues.apache.org/jira/browse/CLOUDSTACK-904
> 
> 
> Repository: cloudstack-git
> 
> 
> Description
> -------
> 
> Cloudstack defines a vm with multiple CPUs instead of one CPU with multiple cores. 
> For Windows as well as RHEL guests, the license is based on the amount of CPU sockets.
> The definition of VCPUs should therefore translate to the amount of cores, not CPUs.
> 
> 
> Diffs
> -----
> 
>   plugins/hypervisors/kvm/src/com/cloud/hypervisor/kvm/resource/LibvirtVMDef.java 5120870 
> 
> Diff: https://reviews.apache.org/r/12679/diff/
> 
> 
> Testing
> -------
> 
> Only tested the output of the append rule.
> 
> 
> Thanks,
> 
> Rene Diepstraten
> 
>