[sheepdog-users] cache performance test

[Liu Yuan]

On Thu, Jul 18, 2013 at 11:43:00AM +0200, Valerio Pachera wrote:
> 2013/7/18 Liu Yuan <namei.unix at gmail.com>:
> >> Is it normal that /mnt/sheep/meta/cache/  is still full of data?
> > No. I think the remained cache is something related to copy-on-write mechanism
> > regards of snapshot.
> 
> vdi list
> s backup_data     3  1.0 TB  4.0 MB  501 GB 2013-07-12 12:40   c8d12a
>    2         lug15
> s backup_data     4  1.0 TB  240 MB  501 GB 2013-07-15 10:04   c8d12b
>    2      lug15bis
> s backup_data     5  1.0 TB  2.1 GB  501 GB 2013-07-15 10:08   c8d12c
>    2  lug17_144450
> s backup_data     6  1.0 TB  964 MB  503 GB 2013-07-17 14:55   c8d12d
>    2  lug17_150000
> s backup_data     7  1.0 TB  5.9 GB  500 GB 2013-07-17 14:59   c8d12e
>    2
>   backup_data     0  1.0 TB   15 GB  505 GB 2013-07-17 15:15   c8d12f     2
> 
> root at test004:/mnt/sheep/meta/cache# du -sh *
> 633M    19093f
> 8,1G    c8d128
> 8,0M    c8d129
> 4,0M    c8d12a
> 229M    c8d12b
> 1,9G    c8d12c
> 61M     c8d12d
> 5,9G    c8d12e
> 17M     c8d12f
> 
> After removing the cache, by rm (with no guests running), the vdi
> backup_data (1T) got corrupted.
> I guess it was better by 'vdi cache delete -s tag'.
> 
> vdi check of 'backup' (10G) and 'whezzy' (10G) fixed something but
> seems fine (there were no snapshot for these vdi).
> 
> Notice what happens when I delete backup_data (more than 200G used, 1T virtual).
> 
> root at test004:/mnt/sheep/meta/cache# collie vdi list
>   Name        Id    Size    Used  Shared    Creation time   VDI id  Copies  Tag
>   backup       0   10 GB  3.1 GB  0.0 MB 2013-07-10 16:58   19093f
> 2
>   wheezy       0   10 GB  1.8 GB  0.0 MB 2013-07-16 15:08   9533ed
> 2
> s backup_data     3  1.0 TB  4.0 MB  501 GB 2013-07-12 12:40   c8d12a
>    2         lug15
> s backup_data     4  1.0 TB  236 MB  501 GB 2013-07-15 10:04   c8d12b
>    2      lug15bis
> s backup_data     5  1.0 TB  2.1 GB  501 GB 2013-07-15 10:08   c8d12c
>    2  lug17_144450
> s backup_data     6  1.0 TB  964 MB  503 GB 2013-07-17 14:55   c8d12d
>    2  lug17_150000
> s backup_data     7  1.0 TB  5.9 GB  500 GB 2013-07-17 14:59   c8d12e
>    2
>   backup_data     0  1.0 TB   15 GB  505 GB 2013-07-17 15:15   c8d12f
>    2
> root at test004:/mnt/sheep/meta/cache# collie vdi delete -s 3 backup_data
> root at test004:/mnt/sheep/meta/cache# collie vdi delete -s 4 backup_data
> root at test004:/mnt/sheep/meta/cache# collie vdi delete -s 5 backup_data
> root at test004:/mnt/sheep/meta/cache# collie vdi delete -s 6 backup_data
> root at test004:/mnt/sheep/meta/cache# collie vdi delete -s 7 backup_data
> 
> root at test004:/mnt/sheep/meta/cache# collie vdi delete backup_data
> failed to read from socket: -1, Resource temporarily unavailable
> failed to read a response
> 

This means that deletion process takes too long and collie timeout. I'm writing
a patch to have collie wait tight for ever (it should).

Thanks
Yuan