Re: [CentOS-virt] Live migration haswell, broadwell

2017-11-14 Thread Sarah Newman
On 11/14/2017 03:13 AM, T.Weyergraf wrote:
> Hi
> 
> I wonder, if live migration (back and forth) is possible on mixed Haswell 
> (Xeon V3) and Broadwell (Xeon V4) installations. The only notable difference
> between the two is apparently a working TSX implementation on V4, which got 
> disabled on V3 due to bugs. The rest (VMCS-shadowing, posted interrupts)
> should not apply to our environment, as we do not run nested-vmx nor 
> device-passthrough on our Xen servers
> 
> Now, I found no sane way to disable TSX on a given system but I cannot rule 
> out, that some (Linux-)software, such as Postgres will use them
> eventually. Also, I have a hard time trying to assess, if TSX can be disabled 
> on V4 to enable seamless migration.
> 
> Any hint would be greately appreciated.

Have you looked at CPUID masking? I haven't tried this but someone wrote a 
handy calculator for it. https://github.com/twizted/xen_maskcalc
___
CentOS-virt mailing list
CentOS-virt@centos.org
https://lists.centos.org/mailman/listinfo/centos-virt


[CentOS-virt] Live migration haswell, broadwell

2017-11-14 Thread T.Weyergraf

Hi

I wonder, if live migration (back and forth) is possible on mixed 
Haswell (Xeon V3) and Broadwell (Xeon V4) installations. The only 
notable difference between the two is apparently a working TSX 
implementation on V4, which got disabled on V3 due to bugs. The rest 
(VMCS-shadowing, posted interrupts) should not apply to our environment, 
as we do not run nested-vmx nor device-passthrough on our Xen servers.


Now, I found no sane way to disable TSX on a given system but I cannot 
rule out, that some (Linux-)software, such as Postgres will use them 
eventually. Also, I have a hard time trying to assess, if TSX can be 
disabled on V4 to enable seamless migration.


Any hint would be greately appreciated.

___
CentOS-virt mailing list
CentOS-virt@centos.org
https://lists.centos.org/mailman/listinfo/centos-virt