[sheepdog-users] Testing snapshot-object-reclaim

Hitoshi Mitake mitake.hitoshi at gmail.com
Fri Mar 14 15:08:53 CET 2014


At Fri, 14 Mar 2014 21:58:32 +0800,
Liu Yuan wrote:
> 
> On Fri, Mar 14, 2014 at 09:27:00PM +0900, Hitoshi Mitake wrote:
> > At Fri, 14 Mar 2014 10:10:23 +0100,
> > Valerio Pachera wrote:
> > > 
> > > [1  <text/plain; UTF-8 (7bit)>]
> > > 2014-03-14 3:42 GMT+01:00 Hitoshi Mitake <mitake.hitoshi at gmail.com>:
> > > 
> > > > Thanks a lot for the log. BTW, was the log produced by the node 0? I
> > > > want to confirn this point.
> > > 
> > > No, the log is from node 2.
> > > I imported the vdi from node 0 but the guest was running on node 2.
> > > 
> > > > Can I see an output of "tree <sheepdog dir>" direct after this status?
> > > > I want to know what kind of object is consuming disk space.
> > > 
> > > 
> > > It's attached.
> > 
> > Thanks for your log. BTW, I think your problem is caused by a bug
> > which is already in the discard operation. For solving the problem, I
> > required to change both of sheepdog and qemu.
> > 
> > The change of sheepdog side is pushed to our tree:
> > https://github.com/sheepdog/sheepdog/tree/snapshot-object-reclaim
> > 
> > The change of QEMU side is pushed to here:
> > https://github.com/sheepdog/qemu/tree/snapshot-object-reclaim
> > 
> > Could you update both of sheepdog and qemu, then test it again?
> > 
> 
> Discard? Probably no, discard can only be enabled by virtio-scsi, which is not
> widely adopted by ordinary users. 

I checked Valerio's log and it seems that the discard operation was
not issued. As you say, the problem wouldn't be related to the
discard.

But I could write a test for the discard operation which causes
failure under the new object reclaim scheme:
https://github.com/sheepdog/sheepdog/commit/6bf6ad3c04df89a21c1a66cd3bf263c91015c70a

Thanks,
Hitoshi



More information about the sheepdog-users mailing list