[sheepdog-users] Testing snapshot-object-reclaim

[Valerio Pachera]

I've been trying to:

create a new vdi, install debian, export the vdi in qcow and raw
format (by dog vdi read).
I tried then the usual snapshot test but there are no differences.
I also tried to poweroff the guest each time before taking a snapshot
to be sure there was no flush related issues, but no luck anyway.

If I take only 3 snapshot (before writing the file in the guest,
after, and after removing it), then the un-removed objects are much
less (12M).

I'm going to change the sheep file system as suggested by Hitoshi,
just to see if it makes nay difference.

root at test006:~# dog vdi snapshot test
root at test006:~# dog vdi snapshot test
root at test006:~# dog vdi snapshot test

root at test006:~# dog vdi list
  Name        Id    Size    Used  Shared    Creation time   VDI id  Copies  Tag
s test         1  5.0 GB  868 MB  0.0 MB 2014-03-31 13:51   7c2b25
2:1
s test         2  5.0 GB  4.5 GB  236 MB 2014-03-31 14:10   7c2b26
2:1
s test         3  5.0 GB   36 MB  4.7 GB 2014-03-31 14:12   7c2b27
2:1
  test         0  5.0 GB  0.0 MB  4.7 GB 2014-03-31 14:13   7c2b28
2:1

root at test006:~# for n in 1 2 3; do dog vdi delete -s $n test; done

root at test006:~# dog vdi list
  Name        Id    Size    Used  Shared    Creation time   VDI id  Copies  Tag
  test         0  5.0 GB  0.0 MB  4.7 GB 2014-03-31 14:13   7c2b28
2:1

root at test006:~# dog node info
Id      Size    Used    Avail   Use%
 0      233 GB  2.4 GB  230 GB    1%
 1      452 GB  2.4 GB  450 GB    0%
 2      500 GB  2.4 GB  497 GB    0%
Total   1.2 TB  7.1 GB  1.2 TB    0%
Total virtual image size        5.0 GB
root at test006:~# dog vdi delete test

root at test006:~# dog node info
Id      Size    Used    Avail   Use%
 0      233 GB  4.0 MB  233 GB    0%
 1      452 GB  4.0 MB  452 GB    0%
 2      500 GB  4.0 MB  500 GB    0%
Total   1.2 TB  12 MB   1.2 TB    0%
Total virtual image size        0.0 MB