Re: [table partitioning] How many partitions are possibel?

Поиск
Список
Период
Сортировка
От Michelle Konzack
Тема Re: [table partitioning] How many partitions are possibel?
Дата
Msg-id 338af873d776ec0128e9fdc9ded31506.squirrel@webmail.tamay-dogan.net
обсуждение исходный текст
Ответ на Re: [table partitioning] How many partitions are possibel?  (Michael Paquier <michael.paquier@gmail.com>)
Ответы Re: [table partitioning] How many partitions are possibel?  (Michael Paquier <michael.paquier@gmail.com>)
Список pgsql-hackers
Hi,

Am 2017-12-26 hackte Michael Paquier in die Tasten:
------------------------------------------------------------------------
> On Tue, Dec 26, 2017 at 11:04:55AM +0100, Michelle Konzack wrote:
>> I like to make the partitions smaller, but the documentation say, you
>> should not dare to make several 1000 partitions..
>
> Even more than a hundred may be already risky in my opinion here. The
> issue with a large number of partitioning using the pre-9.6 grammar
> comes from the planning time which sky-rockets because the optimizer
> uses a O(N^2) algorithm to consider all the partitions. The v10 grammar
> does not take care of this planning problem as far as I recall, but v11
> gets that better for partition-wise joins if my memory does not fail me.
> --
> Michael

I think on migrating to v10, but I have to backport the PostgreSQL.

Currently I am thinking about hot to backup 28 TByte of data...
Have to dump and compress table by table and thios take endless time,
especially, if I am in Estonia and the database is in Germany.

Is there already a rease date for v11?

Thanks in avance

-- 
Michelle Konzack
00372-54541400



В списке pgsql-hackers по дате отправления:

Предыдущее
От: Fabien COELHO
Дата:
Сообщение: Re: [HACKERS] pow support for pgbench
Следующее
От: Michael Paquier
Дата:
Сообщение: Re: Getting rid of "tuple concurrently updated" elog()s withconcurrent DDLs (at least ALTER TABLE)