Re: [Qemu-devel] RFC: async commands with QMP

[Daniel P. Berrange]

On Wed, Jul 29, 2015 at 05:32:59PM +0200, Marc-André Lureau wrote:
> Hi Eric
> 
> On Wed, Jul 29, 2015 at 5:10 PM, Eric Blake <address@hidden> wrote:
> > In fact, see commit 65207c59, where we ripped it out with prejudice.
> 
> Whoo, I should have looked at git history :)
> 
> That implementation looks quite bad indeed (suspending monitor?) And
> it wasn't properly documented it seems. No wonders it wasn't used.
> Btw, apparently it's the reason why "id" is there in the first place.
> I don't think there is a good reason to keep having it in sync
> commands.

Yep, the 'id' is pretty pointless now, but we keep accepting it
and sending it in responses in order to not break compatibility
with any existing apps which might be sending it / checking for
it in responses.

> >>  One of the benefits of this is that it means
> >> that libvirt can determine current status of ongoing background jobs
> >> when it restarts and reconnects to a previously launched QEMU, where
> >> as an async command approach is tied to the specific monitor connection
> >> that is open.
> >
> > And that is a real concern with any new proposal for async commands.
> 
> I don't think it's incompatible with having async commands, since in
> fact they are already async commands with dummy quick return. Some
> async should be cancelled when the client is gone (this depends on the
> async command, whether it is tight to a client or not). However, no
> old async return should be given to a new client.

Agreed, they're not mutually exclusive. I guess it is more a question
of whether there is a compelling enough reason to support both usage
approaches

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|