discrepancies in used space after cpio

[Paul Wootton]

Max Laier wrote:
> On Tuesday 19 May 2009 05:48:51 Tim Kientzle wrote:
>   
>> Paul Wootton wrote:
>>     
>>> I am currently in the process of moving all my data around, going
>>> from a single zfs drive (ex-mirror) to a zfs raidz.
>>> I have used cpio to copy the data to the new pool, but a du shows a
>>> big difference in the results.
>>>
>>> Does anyone have any ideas, or does a "du -h ." not do what I think
>>> it should?
>>>       
>> ... <snip> ...
>>
>>     
>>> demophon# pwd
>>> /DemoPool/var/tmp/kdecache-paul/kpc
>>> demophon# ls -lah
>>> total 1282522
>>> drwx------   2 paul  paul    25B May 15 19:35 .
>>> drwx------  11 paul  paul    16B May 15 19:36 ..
>>> -rw-r--r--   1 paul  paul   5.0M May  6 16:47 kapman_cache.data
>>> -rw-r--r--   1 paul  paul   1.3M May  6 16:47 kapman_cache.index
>>>       
>> ... <snip> ...
>>
>>     
>>> demophon# du -h .
>>> 1.2G    .
>>>
>>> demophon# pwd
>>> /var/tmp/kdecache-paul/kpc
>>> demophon# ls -lah
>>> total 7833
>>> drwx------   2 paul  paul    25B May 18 09:37 .
>>> drwx------  11 paul  paul    16B May 18 09:12 ..
>>> -rw-r--r--   1 paul  paul   5.0M May  6 16:47 kapman_cache.data
>>> -rw-r--r--   1 paul  paul   1.3M May  6 16:47 kapman_cache.index
>>>       
>> ... <snip> ...
>>
>>     
>>> demophon# du -h .
>>> 7.6M    .
>>>       
>> Dmitry Morozovsky wrote:
>>  > Ehmm, possibly stupid question: sparse files?
>>
>> Dmitry's probably right here:  Files .data/.index are
>> probably some kind of database package, which are
>> often highly sparse files.
>>
>> Try "du -k *" in each of these directories to see
>> how much disk space is actually allocated to each file;
>> that would verify that file sparseness is at issue here.
>>     
>
> You can also use "du -hA ." to get the apparent size.  But, as I assume 
> that "/DemoPool" is the destination and "/var" is the source of your cpio 
> copy, it is quite obvious that sparse files are in play at the source.
>
>   
Yes /DemoPool is a raidz pool that is going to replace my single disk 
pool. Dmitry was right about sparse files
demophon# pwd
/var/tmp/kdecache-paul/kpc
demophon# du -hA .
1.2G    .
demophon# du -h .
8.9M    .

Is there a there a better way instead of using cpio for moving an entire 
filing system from a single disk zfs pool to a raidz zfs pool?
Or does making a sparse file in to a none sparse file just consume more 
disk space and no other side affects

Cheers
Paul