Daniel J Gubber wrote:
> Liam,
> 
> Thanks for the responses first of all, I do appreciate it.
> 
> I still have a concern though as far as having to reboot the control
> domain. If in fact, the control
> domain is actually in control of the virtual device layer, and is merely
> acting as a "virtual presentor"
> of these devices to the guest domain layer, if you disrupt the virtual
> presentation of devices,
> you disrupt the guest domain.
> 
> The only way this would not be the case, is if the control domain DOES NOT
> act as the
> "virtual presentor" and the control domain is simply used as a
> configuration manager.
> The saved configuration "define virtual device paths" at the firmware layer
> for use by the
> guest domains, thus making it possible to reboot the control domain without
> disrupting
> guest domain services.
> 
> Is this what's really going on there ? Please let me know
> and thanks again!
> 

I think my last mail (which crossed your mail) covers it but it's the
first scenario above.

The guest sends I/O operations to the control/service domain which
in turn passes them the the physical device. If the control/service
domain is rebooted then the operations are queued in the guest until
the control/service domain comes back or the redundant path is used
(which requires extra configuration)

-- Liam


------------------------------------------------------------------------
Liam Merwick    | SPARC Platform S/W - Common S/W Eng (Ireland)
Ext. x(70)19038 | Tel. +353-21-4778600   |  Mobile. +353-86-8366459


Reply via email to