Re: reboot dom0 without vps downtime

[Wei Liu <wei.liu2@citrix.com>]

On Fri, Mar 17, 2017 at 09:56:44AM +0100, Juergen Gross wrote:
> On 17/03/17 09:49, Vasiliy Tolstov wrote:
> > 2017-03-17 11:47 GMT+03:00 Juergen Gross <jgross@suse.com>:
> >> (Dropping xen-users)
> >>
> >> I have plans for the future, but I'm not sure when I'll have enough
> >> time for it.
> >>
> >> Any help on this topic is welcome, of course!
> >>
> >> I think (at least) the following features are missing:
> >>
> >> - capability to specify a (new) dom0 kernel and initrd for reboot of
> >>   dom0: both must be transferred to the hypervisor in a memory buffer
> >>   which will be used for the reboot
> >> - dom0 reboot capability in hypervisor using the kernel and initrd
> >>   from the memory buffer, without destroying other domains
> >> - dom0 has to check for presence of xenstore-stubdom before trying to
> >>   create it. In case xenstore is already existing connect to it and
> >>   re-establish needed connections to other domains, don't auto-start
> >>   any other domains in this case
> >>
> >> I'm sure there are lots of other little pieces missing, but those can
> >> be addressed after above main features are available.
> > 
> > 
> > What about qemu processes (i'm interesting on hvm domains pvh or how
> > they naming now)
> 
> For pvh domains qemu processes aren't mandatory. All mandatory backends
> like disk and net must live in stubdoms for dom0 being rebootable, of
> course.
> 

stubdoms -> backend domains to be precise.  I think you can use any
normal OS for that purpose.

(I realise people might be using stubdom with slightly different
meanings, but to me it mostly means unikernel-like domains, because
that's how it is done at the moment in Xen. The definition of stubdom is
up for debate, of course).

Wei.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel