[Csync2] csync2 db asymetry

[Vincent Régnard]

Art -kwaak- van Breemen wrote:
> Hello,
> 
> On Fri, Dec 19, 2008 at 10:28:26AM +0100, Vincent Régnard wrote:
>> If the difference has to do with compression or history, why is it
>> always bigger on the same side (master) and not randomly bigger one side
>> and also the other ?
> 
> It's pretty normal for a master database to be bigger than the
> slave. The reason is that the master usually does more
> transactions, and hence is more fragmented. The slack will
> increase due to that, and since sqlite stores everything in a
> single file, the fragmentation difference is very large.
> After a dump/restore  or a full vacuum, the databases should be
> about equal size.
> 
> The master touches:
> - hints
> - dirty
> - file
> 
> The slave touches:
> - file
> 
> With postgres replication we even see a file size difference of 10 to 50%
> of the master. (We actually have database size differences of 10G and then
> we are not talking 50%, but less ;-) ).

Hi,

Thanks a lot for this very clear answer. I can confirm that very large
database with unfrequent (weekly) activity do not suffer this kind of
problem. I have on the same servers another csync2 synchronisation group
with 400MB database that only shows a 1% assymetry ( 387855K/383303K
=1.011 ).

-- 
Vincent Régnard
TBS-internet.com

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 5806 bytes
Desc: S/MIME Cryptographic Signature
Url : http://lists.linbit.com/pipermail/csync2/attachments/20081223/98d1ebd4/smime.bin