Live Snapshot problem: Xen 4.4+ with qcow2 disk

Live Snapshot problem: Xen 4.4+ with qcow2 disk

[kevin.zhang]

[-- Attachment #1.1: Type: text/plain, Size: 826 bytes --]

Hi all,

    Today, I tried to test whether Live Snapshot schema can be adopted to qemu-xen with qcow2 disk image as
what we do with qemu-xen-traditional using vhd disk image.
    My live snapshot procedure is as follow:
( my vm1 is currently running with qcow2 disk file 1.qcow2 ) 
 1. xl pause vm1
 2. mv 1.qcow2 snap.qcow2
 3. qemu-img create  -b snap.qcow2 -f qcow2 1.qcow2
 4. xl unpause vm1

   However, as I unpaused vm1, the write operation still went to snap.qcow2 instead of  the newly created snapshot file 1.qcow2.
I remember that xen 4.0 also had similar problem if we did vhd live snapshot, and the problem was fixed in xen 4.1.
   My question is : whether the failure of making live qcow2 snapshot is a known bug, or  whether I've missed some key operations to make it work.





Best Regards

[-- Attachment #1.2: Type: text/html, Size: 5019 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Live Snapshot problem: Xen 4.4+ with qcow2 disk

[Ian Campbell]

On Fri, 2014-07-18 at 18:26 +0800, kevin.zhang@octlink.com wrote:
> Hi all,

Please don't post colourful HTML emails to xen-devel, stick to plain
text.

Ian.

Re: Live Snapshot problem: Xen 4.4+ with qcow2 disk

[kevin.zhang]

[-- Attachment #1.1: Type: text/plain, Size: 878 bytes --]


Hi Campbell,

Sorry for the wrong format, 
Then, paste into non-colorful plain text.

Hi all, Today, I tried to test whether Live Snapshot schema can be adopted to qemu-xen with qcow2 disk image as what we do with qemu-xen-traditional using vhd disk image. 
My live snapshot procedure is as follow:
 ( my vm1 is currently running with qcow2 disk file 1.qcow2 ) 
1. xl pause vm1 
2. mv 1.qcow2 snap.qcow2 
3. qemu-img create -b snap.qcow2 -f qcow2 1.qcow2 
4. xl unpause vm1 
However, as I unpaused vm1, the write operation still went to snap.qcow2 instead of the newly created snapshot file 1.qcow2.
I remember that xen 4.0 also had similar problem if we did vhd live snapshot, and the problem was fixed in xen 4.1. 
My question is : 
whether the failure of making live qcow2 snapshot is a known bug, or whether I've missed some key operations to make it work. 

[-- Attachment #1.2: Type: text/html, Size: 2113 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Live Snapshot problem: Xen 4.4+ with qcow2 disk

[Ian Campbell]

On Fri, 2014-07-18 at 23:05 +0800, kevin.zhang@octlink.com wrote:
>  
> Hi Campbell,
>  
> Sorry for the wrong format, 
> Then, paste into non-colorful plain text.

It was still an HTML mail. Please find a way to disable this.

> Hi all, Today, I tried to test whether Live Snapshot schema can be
> adopted to qemu-xen with qcow2 disk image as what we do with
> qemu-xen-traditional using vhd disk image. 
> My live snapshot procedure is as follow:
>  ( my vm1 is currently running with qcow2 disk file 1.qcow2 ) 
> 1. xl pause vm1 
> 2. mv 1.qcow2 snap.qcow2 
> 3. qemu-img create -b snap.qcow2 -f qcow2 1.qcow2 
> 4. xl unpause vm1 
> However, as I unpaused vm1, the write operation still went to
> snap.qcow2 instead of the newly created snapshot file 1.qcow2.
> I remember that xen 4.0 also had similar problem if we did vhd live
> snapshot, and the problem was fixed in xen 4.1. 
> My question is : 
> whether the failure of making live qcow2 snapshot is a known bug, or
> whether I've missed some key operations to make it work. 

You've not told the toolstack what you've done, so it (or more
specifically qemu) still has the filehandle open to the original file,
renaming the file has no effect on that.

Bamvor has been posting designs for integrating snapshotting into the xl
toolstack, perhaps you could consider contributing to his efforts.

Also I should note that just pausing the vm may not be sufficient to
guarantee a consistent disk snapshot since there might be I/O in flight.

Cheers,
Ian.