The logs only store the PK, the event datetime and the kind of event
(insert, update, delete). When the replicator reads the log, it must
order the records in a timeline to respect the cronologic order. Then,
my replicator creates one file for each replicatable table and plus an
timeline file that points to a file and recno that contains tables data.
After write this files, it clear the logs (all in one transaction, of
course). Then, it zips all files and put on ftp server. The other side,
on your time, gets the zip files from ftp server, unzip in file datetime
order and apply the changes on database, on cronological order from
timeline file. When some error occur, the zip file (i call pack) stays
on same directory, and a error message is sended to an configured
e-mail. The error must be solved and the pack can be applied again. But
if the two databases contains the same metadata, erros are sparse...
When the pack is successfully applied. it's moved to another directory.
My replicator also sends an e-mail each day at same predefined time I
call stayalive message to say that the service is active.

Regards,

--

*__________________________________________________________________*
*Marcos Vinicius Dufloth*
*Diretor de Tecnologia
**Cyber Informática Empresarial**

marvin@...* <mailto:marvin@...>
*http://www.tecnocyber.com.br* <http://www.tecnocyber.com.br/>
*__________________________________________________________________*



guido.klapperich@... wrote:

>>You can also read logs from database, put this changes in a zip and send
>>it via ftp, so:
>>
>>
>
>This sounds interesting. Can you please go more into detail. You don't need
>only the logs, but also some data. For example in the log is an entry like
>"new customer", then you have to export this customer as well. And what
>about error handling?
>
>Regards
>
>Guido
>
>
>
>
>
>
>Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>

[Non-text portions of this message have been removed]