Re: [Xen-devel] [RFC V5 2/5] Libxl Domain Snapshot API Design

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Ian Campbell writes ("Re: [RFC V5 2/5] Libxl Domain Snapshot API Design"):
> On Mon, 2014-07-07 at 15:46 +0800, Bamvor Jian Zhang wrote:
...
> >         ("memory",        string),       /* The path to save domain
> >                                           * memory image. 'empty' means
> >                                           * it is a disk-only snapshot.
> >                                           * note that "yes" or "no" is
> >                                           * not allowed, this is different
> >                                           * from xl.snapshot.pod.5

As I wrote on IRC:

I still don't really understand how the distinction between snapshots
with memory=yes and memory=no is supposed to operate.  Is it that if
you take a snapshot of a running domain with memory=no, and then later
resume it, it is as if the guest had crashed at the time of the
snapshot ?  (How does this interact with disk IO barriers?)

I think the semantics should be spelled out somewhere.

(Also I assume that a memory snapshot isn't available unless there's a
running domain, but that also needs to be specified.)

Thanks,
Ian.

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