Re: To save or not to save

["Daniel P. Berrange" <berrange redhat com>]

On Fri, Nov 16, 2012 at 12:57:27AM +0200, Zeeshan Ali (Khattak) wrote:
> On Thu, Nov 15, 2012 at 10:32 PM, Zeeshan Ali (Khattak)
> <zeeshanak gnome org> wrote:
> > On Thu, Nov 15, 2012 at 9:31 PM, Daniel P. Berrange <berrange redhat com> wrote:
> >> On Thu, Nov 15, 2012 at 07:49:46PM +0200, Zeeshan Ali (Khattak) wrote:
> >>> Hi everyone,
> >>>    As you all know currently we save running local VMs when user quits
> >>> Boxes UI. The reason for doing so were:
> >>>
> >>> 1.  We want to reclaim all the resources (especially RAM, cause CPU
> >>> could be saved by only pausing the VM)
> >>> 2. VM state should not be lost across reboots/shutdowns of host
> >>>
> >>> However, saving of VM seems to have been a very heavy operation,
> >>> especially on machines with not a lot of RAM (on my machine with only
> >>> 4GB RAM, host more or less freezes completely for a few seconds). The
> >>> reason AFAICT is that most of guest RAM being unused is swapped out so
> >>> saving then involves copying GBs of data per running VM from one part
> >>> of the drive to another.
> >>
> >> Have you tried making Boxes use the
> >>
> >>   VIR_DOMAIN_SAVE_BYPASS_CACHE
> >>
> >> flag yet ? oVirt saw serious problems with host performance when
> >> saving VMs to disk, unless they used VIR_DOMAIN_SAVE_BYPASS_CACHE
> >> to avoid trashing the I/O cache.
> >
> > Not yet. I somehow forgot about that actually. :( Now added to top of
> > my todo file so will do so soon enough. :)
> 
> Now I tried many times with and without this flag on my machine. *If*
> there is any difference, its that this flag makes the operation
> slower: on average I measured a difference of 2 seconds between
> hitting 'CTRL+Q" and boxes exiting.

It makes sense that it would be slower, because this is making sure that
we do direct I/O to the disk. Without this flag the I/O ends up in the
page cache, not actually flushed to disk until some arbitrary time after
the operation has finished.

The difference should be what impact it has on other programs - since
we're not filling the page cache with VM save data, other programs
doing I/O should not be so badly affected if they try todo I/O too
because they should be less likely to block on I/O while the kernel
flushes the cache to disk.

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 :|