RE: [Xen-devel] xl list vs xm list: backwards compatibility vs forward progress?

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Wed, 28 Jul 2010, Dan Magenheimer wrote:
> > From: Ian Jackson [mailto:Ian.Jackson@xxxxxxxxxxxxx]
> > Sent: Wednesday, July 28, 2010 7:56 AM
> > To: Dan Magenheimer
> > Cc: Xen-Devel (xen-devel@xxxxxxxxxxxxxxxxxxx)
> > Subject: Re: [Xen-devel] xl list vs xm list: backwards compatibility vs
> > forward progress?
> > 
> > Dan Magenheimer writes ("[Xen-devel] xl list vs xm list: backwards
> > compatibility vs forward progress?"):
> > > I was pleased to find out that "xl list" shows the
> > > current amount of physical memory in use by a domain
> > > rather than the start amount.  This is very useful
> > > for monitoring self-ballooning!  I like it!
> > >
> > > But... it is not backwards compatible with "xm list".
> > 
> > Isn't it ?
> > 
> > I'm pretty sure I've written a program which does "xm mem-set" and
> > then polls the output from "xm list" to wait for the target domain to
> > balloon down.
> 
> Hmmm... perhaps xm keeps track of xm mem-set commands... but a
> domain can do ballooning without involvement of the toolchain.
> Self-ballooning (see xen.hg/tools/xenballoon, in tree for about
> two years now) does that in-guest-userland and I have kernel rpms
> for EL5u5 and RHEL6b2 that do it in-guest-kernel.  (The latter
> will be the preferred guest deployment method for Xen systems
> running tmem to optimize memory utilization.)
> 
> With self-ballooning in a guest, xl list and xm list
> very definitely show different values for memory.
> 
> Dan
> 
> P.S. Frankly, I think the xm list behavior is a bug, but
> backwards compatibility -- plus my poor parseltongue --
> stopped me from trying to fix it.
> 

I agree, xm list should be the one to be fixed.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel