Re: incremental dumps

On Thu, Aug 1, 2013 at 10:59 AM, <hamann(dot)w(at)t-online(dot)de> wrote:

> However, the diff files seem to be considerably larger than one would expect.
> One obvious part of the problem is the fact that diff shows old and new text,
> so e.g. changing the amount of stock for a product with a 1kB description
> would generate at least 2kb of text in the diff file.
>
> What would be a better approach?

I suppose wal archiving or PITR would be better, but assuming you want
text files I guess you need to change your database structure to
either:
1) include a watermark on data and dump only data since the last dump
(to do manually);
2) partition your tables and backup specific tables/partitions
depending on the timing.

Hope this helps.
Luca