Re: Cost problem

Поиск
Список
Период
Сортировка
От Ragnar Hafstað
Тема Re: Cost problem
Дата
Msg-id 1123084966.19874.51.camel@localhost.localdomain
обсуждение исходный текст
Ответ на Cost problem  (Victor <victor@ambra.ro>)
Список pgsql-general
On Tue, 2005-08-02 at 17:32 +0300, Victor wrote:
> Hello.
>
> I use psql (PostgreSQL) 7.4.5.
>
> I have a cost problem.
>
> A simple select lasts too long.
> I have a table with aprox 900 rows.
> All rows are deleted and reinserted once a minute.
>

> EXPLAIN ANALYZE SELECT * FROM logati;
>                                                       QUERY PLAN
>
-----------------------------------------------------------------------------------------------------------------------
>  Seq Scan on logati  (cost=0.00..100741.22 rows=1198722 width=340)
> (actual time=35927.945..35944.272 rows=842 loops=1)
>  Total runtime: 35945.840 ms

The rows are not actually removed when DELETE happens.
the next VACUUM vill put them into the Free Space Map
if they are older than the oldest running transaction,
so that their space can be reused.

this means that you must VACUUM this table frequently
to keep it from bloating. If the data distribution keeps
changing , you should also ANALYZE frequently (but not
when the table in empty), but that is another matter.

to fix an already severely bloated table, you might
need to do a VACUUM FULL on it once, but after that
frequent-enough regular (non-FULL) VACUUMS should do.

depending on the nature of your processes, you might
want to use TRUNCATE to empty your table.

gnari



В списке pgsql-general по дате отправления:

Предыдущее
От: Dr NoName
Дата:
Сообщение: Re: Failure to use indexes (fwd)
Следующее
От: Edmund Dengler
Дата:
Сообщение: Re: Failure to use indexes (fwd)