Luca Ferrari wrote:
On Fri, Aug 2, 2013 at 6:55 PM, <hamann.w@t-online.de> wrote:
> thanks for the hint - this is probably one of the things to do.
> I have something else in mind, but at present I just suspect that this might happen:
> when I modify data and select _without an ordering_, I am pretty sure to get the data
> in a different sequence than before. So I wonder whethet forcing the dump to honor
> a particular ordering (if that is at all possible) would also reduce the size of dumps ... or the
> time diff takes to produce them
>
May I ask what is the final purpose? Because if it is to take a backup
I guess this is not the right way to go, while if it is keeping (and
rebuilding) an history of data, than using a more specific approach
(like logging) could give you less headaches.
Luca
--------------
Hi Luca,
we recently decided to have off-site backups rather than burning piles of DVDs that are kept
on-site. The backup server sits in a data center and is fed nightly via rsync.
The link is not too fast.
One thought in favor of text files: if disaster really strikes (the regular machine goes on fire)
it is quite likely that a replacement would be installed with latest versions of all software.
Now, if I had binary files, I would probably have to install the old version of the software
just to be able to do a regular dump and then reload into newer one
With the planned setup, I would be able to look up previous states of the database as well.
(Sample scenario: when was the price of product xyz actually changed?)
This is likely not too convenient ... but loading successive dumps into a secondary installation
of the database is definitely worse.
Regards
Wolfgang
