On Mon, Sep 29, 2014 at 10:33:37AM +0100, Stefan Hajnoczi wrote:
On Thu, Sep 25, 2014 at 3:26 PM, Kashyap Chamarthy
<kchamart(a)redhat.com> wrote:
> This notes is based on an IRC conversation with Eric Blake, to have
> efficient non-shared storage live migration. Thought I'd post my notes
> here before I forget. Please review and spot if there are any
> inaccuracies.
What are you trying to achieve?
Hmm, I clearly failed to articulate what I was trying to do then. Allow
me to rephrase.
So, this was just an attempt out of curiosity derived from an IRC
conversation with Eric Blake on #virt (OFTC) some months ago -- once
libvirt starts using QEMU 2.0's (block) commit of active layer of a disk
image chain into its immediate base or one of the intermediate images,
live migration with non-shared storage should be more efficient (with a
combination of blockcopy+blockcommit).
Now that all the relevant bits are in place for awhile, So, I was just
trying to test that works out.
- Relevant git commit from libvirt:
318cfabdb442f41c9a7016889526c67aad26a503 (blockcommit: document
semantics of committing active layer)
- Relevant git commit from QEMU:
20a63d2cec838c2dde4d246c4d7abe747d9b7a11 (commit: Support commit
active layer)
Hope I explain it a bit more clear.
More context: I intend to a do user/admin style talk at the upcoming
LinuxCon/CloudOpen in Düsseldorf, so was just testing and experimenting
to see what kind of practical steps can virt/cloud software (like
OpenStack) users/adminstrators can take advantage of the newer features
in libvirt/QEMU when dealing with snapshots and large disk image
chains/merge. To keep things less complicated, I'm mainly tinkering with
file-based snapshots.
https://kashyapc.fedorapeople.org/virt/lcce-2014/Abstract-CloudOpen-Eu-20...
--
/kashyap