[Qemu-devel] Re: [PATCH 1/2] block: Add wr_highest_sector blockstat

[Daniel P. Berrange]

On Wed, Apr 28, 2010 at 12:47:49PM -0500, Anthony Liguori wrote:
> On 04/28/2010 12:04 PM, Luiz Capitulino wrote:
> >On Wed, 28 Apr 2010 11:01:12 -0500
> >Anthony Liguori<address@hidden>  wrote:
> >
> >   
> >>On 04/28/2010 10:56 AM, Kevin Wolf wrote:
> >>     
> >>>This adds the wr_highest_sector blockstat which implements what is 
> >>>generally
> >>>known as the high watermark. It is the highest offset of a sector 
> >>>written to
> >>>the respective BlockDriverState since it has been opened.
> >>>
> >>>The query-blockstat QMP command is extended to add this value to the 
> >>>result,
> >>>and also to add the statistics of the underlying protocol in a new 
> >>>"parent"
> >>>field. Note that to get the "high watermark" of a qcow2 image, you need 
> >>>to look
> >>>into the wr_highest_sector field of the parent (which can be a file, a
> >>>host_device, ...). The wr_highest_sector of the qcow2 BlockDriverState 
> >>>itself
> >>>is the highest offset on the _virtual_ disk that the guest has written 
> >>>to.
> >>>
> >>>Signed-off-by: Kevin Wolf<address@hidden>
> >>>
> >>>       
> >>I see, you did print out stats for each layer.
> >>
> >>I don't think we should take 2/2.  I don't mind QMP having more features
> >>than the user monitor.
> >>     
> >  I don't either, but Kevin has said to me that this information is also 
> >  good
> >for the user Monitor.
> >
> >  The real question here is whether or not we're going to stop supporting
> >stability for the user Monitor and if so, when we'll break it.
> >
> >  An arguable reasonable policy would be to try to maintain stability for
> >existing commands. In this specific case, 'info blockstats' is used by
> >libvirt afaik. So breaking it would mean that older libvirt versions won't
> >be able to talk to newer qemu (taking libvirt just as real known example).
> >   
> 
> I think we should try our best to maintain compatibility.  In this case, 
> this change would break any non-QMP version of libvirt so it would be 
> pretty painful for users.  That's why I'm inclined to not take.

I agree. There are alot of existing deployed libvirt's out there which all
still use the user monitor. libvirt will enable JSON by default once QEMU
brings out 0.13, but there'll still be people with older libvirt. We need
to avoid breaking the user monitor for at least one or two releaes to allow
time for apps to catch up with the switch to JSON. Mark the user monitor
deprecated in 0.13 by all means, but keep compatability until either the 
0.15 or 0.16 release if at all practical.

Regards,
Daniel
-- 
|: Red Hat, Engineering, London    -o-   http://people.redhat.com/berrange/ :|
|: http://libvirt.org -o- http://virt-manager.org -o- http://deltacloud.org :|
|: http://autobuild.org        -o-         http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505  -o-   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|