Re: VACUUM process running for a long time

Поиск
Список
Период
Сортировка
От Scott Marlowe
Тема Re: VACUUM process running for a long time
Дата
Msg-id v2rdcc563d11004160233r4d981993haf2898a4340561@mail.gmail.com
обсуждение исходный текст
Ответ на VACUUM process running for a long time  (Jan Krcmar <honza801@gmail.com>)
Список pgsql-general
On Wed, Apr 14, 2010 at 8:01 AM, Jan Krcmar <honza801@gmail.com> wrote:
> hi
>
> i've got the database (about 300G) and it's still growing.
>
> i am inserting new data (about 2G/day) into the database (there is
> only one table there) and i'm also deleting about 2G/day (data older
> than month).
>
> the documentation says, one should run VACUUM if there are many
> changes in the database, but the vacuumdb never finishes sooner than
> the new data should be imported.
>
> is there any technique that can solve this problem?

Are you running autovacuum?  Has it been tuned to be more aggresive
than the default.  I've got some large heavily updated dbs for which
I've had to turn down the autovacuum_vacuum_cost_delay to 2 or 5 ms to
get it to keep up.  But I've got a pretty good IO subsystem that can
handle the more aggresive autovacuum.

If you're doing one big insert and one big delete a day, then you
should be able to just kick off a regular vacuum at the end of the
delete, with low cost_delay and higher cost_limit that might keep up.
However, if you're on the edge on your IO subsystem then it isn't
gonna help much because it's gonna slow down the system too much.

В списке pgsql-general по дате отправления:

Предыдущее
От: Adrian von Bidder
Дата:
Сообщение: Re: VACUUM process running for a long time
Следующее
От: Peter Bex
Дата:
Сообщение: Tuple storage overhead