On 3/19/07, Jan Poslusny <pajout@gingerall.cz> wrote:
> Hi all,
> I participate on realization of warm standby pg 8.2.3, according to
> documentation. Mostly all works fine, but one problem remains. The
> recovering rdbms eats transfered wal logs fluently, but I cannot see the
> way how to recognize if the recovered wal log file is really processed
> into db and, consequently, is obsolete and does not need to be archived
> for this recovery. Specially, when recovering rdbms is stopped, some wal
> log can be processed (and automatically deleted), but the rdbms asks the
> same file after restart in recovery mode. Is it some way how to
> recognize such logs?
> I know that I should keep base backup AND all wal logs after
> pg_start_backup, but my question is NOT about safe archiving, but about
> log shipping.
you should check out the pg_standby utility. They solved the problem
by letting you set the maximum amount of files in your archive
transfer folder, which should be the maximum amount of WAL traffic you
would see between checkpoints plus a good safety margin. It works...
merlin