On 07/18/2011 05:39 PM, Stefan Hajnoczi wrote:
On Mon, Jul 18, 2011 at 2:08 PM, sirmax<sirmax@xxxxxxxxxxxxx> wrote:
> Node B with VM started:
>
> VM_IMAGE="/usr/local/virtual/img/win-xp-test-migrate.img"
> VM_NAME="win-xp-test-m"
> DEV0="kvm_m1"
>
> TMPDIR="/var/tmp/" /usr/bin/kvm \
> -hda $VM_IMAGE \
> -net nic -net tap,ifname=${DEV},script=no,downscript=no \
> -boot c \
> -vnc :15 \
> -usbdevice tablet \
> -usb \
> -snapshot \
> -monitor tcp:127.0.0.1:4415,server,nowait \
> -monitor unix:/tmp/${VM_NAME}.fifo,server,nowait \
> -incoming tcp:172.16.253.1:9999
The -snapshot command works by creating a temporary qcow2 image file
based on the disk image. In your case the effect is something like:
qemu-img create -f qcow2 -o backing_file=$VM_IMAGE /tmp/...
This temporary file is now your disk image. All write operations from
the guest will touch the temporary file and *not* the base image
($VM_IMAGE).
When you combine -snapshot with live migration QEMU creates the
temporary image before performing the block migration. Therefore the
block migration copies over all data into the temporary file (doh!).
Seems like a special case to me that would need to be carefully fixed
by performing snapshotting after block migration.
Why? It seems to me -snapshot means "don't write into this image".
What I'd expect is that all data be written into the temporary image,
perhaps with the optimization that unchanged data is taken from the
backing file.
--
error compiling committee.c: too many arguments to function
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
