On 02/14/07 16:00, Joshua D. Drake wrote:
> Tom Lane wrote:
>> Zach Bagnall <zach.bagnall@bulletinwireless.com> writes:
>>>> Seriously though, the PITR recovery mechanism is quite mature and usable.
>>> I would say it is a good start - and progress is clearly being made. But
>>> it wont be "usable" for most serious applications until we have
>>> per-database transaction logging and backup recovery.
>
>> Don't hold your breath; that's not even on the agenda, much less
>> something that's likely to appear soon. If you need separable recovery
>> then run a different postmaster instance for each database.
>
> I would have to concur... if you need such a feature I would wonder why
> you are running multiple databases within the same cluster at all.
It's not exactly an exotic feature. Sybase has had it for as long as I
can remember, as has Informix and DB2 (I'm told).
The obvious reason for not running multiple instances is that postgresql
will know how to use the available memory most efficiently. By splitting
them up, each postmaster must be tuned and adjusted by a stupid human.
Unless you meant using a new hardware server for each database..