Re: [Qemu-devel] Tunneled Migration with Non-Shared Storage

[Dr. David Alan Gilbert]

* Gary R Hook (address@hidden) wrote:
> What I really need to figure out is why, when performing a migration
> using non-shared storage, an entire VM is copied into memory before it???s
> sent across the wire using this method. A copy-on-read operation
> is performed first, then the disk is sent, then the dirty pages, then
> the RAM.
> 
> The odd thing is that non-tunneled migrations use a completely different
> code path that does _not_ execute a copy-on-read first.

This seems odd, since as far as I know the tunneling code is quite separate
to the migration code; I thought the only thing that the migration
code sees different is the file descriptors it gets past.
(Having said that, again I don't know storage stuff, so if this
is a storage special there may be something there...)

Do you have the libvirt commands you're using and ideally the logs
to show what qmp commands are being sent to qemu to see what's
actually doing the COR ?

> The problem: VMs (VM disks) that are larger than available memory can't
> be migrated due to space constraints.
> 
> I'd like to disable this COR opration but I think it's important to
> understand why the code is written this way. I've yet to run across any
> comments that explain the need for this "extra" copy, and am looking
> for background and advice.
> 
> Any insights are welcome.

Dave

> 
> -- 
> Gary R Hook
> Senior Kernel Engineer
> NIMBOXX, Inc
> 
> 
> 
--
Dr. David Alan Gilbert / address@hidden / Manchester, UK