[sheepdog-users] Fwd: Testing snapshot-object-reclaim

[Liu Yuan]

On Thu, May 15, 2014 at 04:55:08PM +0200, Valerio Pachera wrote:
> Tthe current snapshot algorithm (master branch) also leak some objects:
> 
> Sheepdog daemon version 0.8.0_152_g65dde97
> 
> 
> dog vdi list
>   Name        Id    Size    Used  Shared    Creation time   VDI id  Copies
> Tag
> s test         1  5.0 GB  1.0 GB  0.0 MB 2014-05-15 15:45   7c2b25
> 2:1
>   test         0  5.0 GB  0.0 MB  1.0 GB 2014-05-15 15:48   7c2b26
> 2:1
> c testClone     0  5.0 GB   52 MB 1020 MB 2014-05-15 15:48   a90c42
> 2:1
> 
> dog node info
> Id      Size    Used    Avail   Use%
>  0      450 GB  574 MB  449 GB    0%
>  1      235 GB  574 MB  234 GB    0%
>  2      100 GB  574 MB  99 GB     0%
> Total   784 GB  1.7 GB  783 GB    0%
> Total virtual image size        10 GB
> 
> dog vdi delete -s 1 test
> dog vdi delete test
> dog vdi delete testClone
> 
> 
> dog vdi list
>   Name        Id    Size    Used  Shared    Creation time   VDI id  Copies
> Tag
> 
> dog node info
> Id      Size    Used    Avail   Use%
>  0      450 GB  12 MB   450 GB    0%
>  1      235 GB  12 MB   235 GB    0%
>  2      100 GB  12 MB   100 GB    0%
> Total   784 GB  36 MB   784 GB    0%
> 
> Total virtual image size        0.0 MB

This is a inherent problem with the current reclaim algoritm, we don't reclaim
inode object (but we can reuse them if new vdi created)

Thanks
Yuan