On Fri, Jan 12, 2007 at 08:25:26AM +0100, Ulrich Windl wrote:
> On 11 Jan 2007 at 17:33, Ligesh wrote:
> > >
> >
> > How? The buffers are in the domU, the device is in the Dom0. The only
>
> Still the DomU has a device it maintains buffers for. Having flushed the
> buffers
> in DomU is rather independ of the data being written in Dom0 as well, but
> that's
> the LVM snapshot we are talking about.
I don't get you.
>
> > way is to send a sync to the DomU, and then pause it. Which is again the
> > sync-and-pause command.
>
> Things may change between the sync and the pause. YOU were talking about a
> 100%
> method, not about a best effort.
>
Yeah, that's why I said it has to done atomically, and as Mark pointed out,
the only way is to implement both sync and save inside the domU kernel itself.
Linux has a hibernate function right, so actually it should work. In fact, that
would turn out to be even better than I had initially thought.
xm sync-and-save domU file (The code is implemented in the domU, and so it is
atomic)
lvm snapshot
xm restore file
backup lvm + file
Then we won't even lose the applications buffers. We can restore it perfectly
to the exact state it was before. I think this feature is important enough for
real life scenarios. Uptime is the key. Installing backup software in each domU
completely nullifies any advantage of virtual hosting.
>>> mencken...
Mencken seems to have laughed himself into irrelevance. I have a seen a lot of
his quotes misattributed.
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home