zfs discuss - Mar 2008 - zfs snapshots sometimes don''t report correct used space

Hi,

I''m using Tim Foster''s zfs automatic snapshot to keep copies
of a zfs filesystem.

I keep 30 days, seven weeks and 12 months, these snapshots get created at 00:00
every day and when a new week starts (or a new month) there are two or three
snapshots that start at the same time.

Using zfs list I see that when more than one snapshot starts at the same time,
the USED column of the report shows no space occupied, while, when just a single
snapshot is done, I can see how much space is being used.

Snaphots are ok, I mean, if I go inside one of the 0 byte used one I can see
that the files are ok and their content is what they really contained at that
point in time, so it seem just a ''cosmetic'' problem.

Here his zfs list on that filesystem

-----------8<---------------------------------
# zfs list
NAME                                                   USED  AVAIL  REFER  MOUNT
POINT
nas                                                    262G   879G  11,2G  /nas
nas/nwserv                                            13,0G   879G  6,86G  /nas/
nwserv
nas/nwserv at zfs-auto-snap:daily-2008-02-27-00:00:00     519M      -  6,99G  -
nas/nwserv at zfs-auto-snap:daily-2008-02-28-00:00:00     491M      -  7,02G  -
nas/nwserv at zfs-auto-snap:daily-2008-02-29-00:00:00        0      -  6,91G  -
nas/nwserv at zfs-auto-snap:weekly-2008-02-29-00:00:00       0      -  6,91G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-01-00:00:00        0      -  6,93G  -
nas/nwserv at zfs-auto-snap:monthly-2008-03-01-00:00:00      0      -  6,93G  -
nas/nwserv at zfs-auto-snap:weekly-2008-03-01-00:00:00       0      -  6,93G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-02-00:00:00     241M      -  6,94G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-03-00:00:00     242M      -  6,94G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-04-00:00:00     538M      -  6,96G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-05-00:00:00     499M      -  6,74G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-06-00:00:00     519M      -  6,80G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-07-00:00:00     514M      -  6,81G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-08-00:00:00        0      -  6,80G  -
nas/nwserv at zfs-auto-snap:weekly-2008-03-08-00:00:00       0      -  6,80G  -
nas/nwserv at zfs-auto-snap:daily-2008-03-09-00:00:00     256M      -  6,85G  -
----------->8---------------------------------

As you can see, when there is a weekly or monthly snapshot used space is 0.

Is this a known issue?

Best regards.

Maurilio.

PS. nas/nwserv is compressed and has atime=off

# zfs get all nas/nwserv | grep local
nas/nwserv  compression                    on                             local
nas/nwserv  atime                          off                            local
nas/nwserv  snapdir                        hidden                         local
nas/nwserv  com.sun:auto-snapshot:weekly   true                           local
nas/nwserv  com.sun:auto-snapshot:monthly  true                           local
nas/nwserv  com.sun:auto-snapshot:daily    true                           local
#
 
 
This message posted from opensolaris.org

>Here his zfs list on that filesystem
>
>-----------8<---------------------------------
># zfs list
>NAME                                                   USED  AVAIL  REFER 
MOUNT
>POINT
>nas                                                    262G   879G  11,2G 
/nas
>nas/nwserv                                            13,0G   879G  6,86G 
/nas/
>nwserv
>nas/nwserv at zfs-auto-snap:daily-2008-02-27-00:00:00     519M      -  6,99G
-
>nas/nwserv at zfs-auto-snap:daily-2008-02-28-00:00:00     491M      -  7,02G
-
>nas/nwserv at zfs-auto-snap:daily-2008-02-29-00:00:00        0      -  6,91G
-
>nas/nwserv at zfs-auto-snap:weekly-2008-02-29-00:00:00       0      -  6,91G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-01-00:00:00        0      -  6,93G
-
>nas/nwserv at zfs-auto-snap:monthly-2008-03-01-00:00:00      0      -  6,93G
-
>nas/nwserv at zfs-auto-snap:weekly-2008-03-01-00:00:00       0      -  6,93G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-02-00:00:00     241M      -  6,94G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-03-00:00:00     242M      -  6,94G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-04-00:00:00     538M      -  6,96G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-05-00:00:00     499M      -  6,74G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-06-00:00:00     519M      -  6,80G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-07-00:00:00     514M      -  6,81G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-08-00:00:00        0      -  6,80G
-
>nas/nwserv at zfs-auto-snap:weekly-2008-03-08-00:00:00       0      -  6,80G
-
>nas/nwserv at zfs-auto-snap:daily-2008-03-09-00:00:00     256M      -  6,85G
-
>----------->8---------------------------------
>
>As you can see, when there is a weekly or monthly snapshot used space is 0.
>
>Is this a known issue?

Why do you think it''s an issue in the first place?

The space allocated for a snapshot is the amount of data freed after it
was created which is not also allocated elsewhere.  The way this
appears to work is that the space is counted towards the most recently
created snapshot which the data is shared with.

No data is freed when that snapshot is destroyed.

Casper

>So, every next day, I see how much data changed since the day before, but I
>can''t see this info when more than one snapshot gets executed.
>
>Hope this clarifies my issue a little bit.

I understand, but i don''t think the accounting is meant to be that way;
it''s more or less "if you remove this data, at most X data will be
freed".

If you remove one of the non-0 sized  snapshots, it will show more space 
allocated toward the now 0 sized snapshots.  It seems that zfs will only
count data toward one snapshot. 

Casper