Re: Problem with reading data from standby server?

Поиск
Список
Период
Сортировка
От Condor
Тема Re: Problem with reading data from standby server?
Дата
Msg-id c69174615613525decf74d87a4332406@stz-bg.com
обсуждение исходный текст
Ответ на Re: Problem with reading data from standby server ?  (Andres Freund <andres@anarazel.de>)
Список pgsql-general
On 21.04.2012 12:59, Andres Freund wrote:
> On Friday, April 20, 2012 10:39:25 AM Condor wrote:
>> array_accumulate aggregate is:
>>
>>
>> CREATE AGGREGATE array_accumulate (
>>      sfunc = array_append,
>>      basetype = anyelement,
>>      stype = anyarray,
>>      initcond = '{}'
>> );
> Btw, if you replace that by array_agg which is builtin in any version
> support
> HS/SR the whole query might finish faster and thus is less likely to
> cause
> conflicts.

No, the speed is the same, I just check with explain but yes, I will
switch to use array_agg

>
> If youre already on 9.1 you might also want to look into
> hot_standby_feedback
> that can also reduce the likelihood of conflicts by informing the
> master what
> is going on on the standby.
>

This option hot_standby_feedback look interesting, but when I read what
do more questions coming like:

This can cause database bloat on the primary for some workloads. Well
if I run one query every hour and take
30 min to finish the job, with how much my database on primary will be
increase and when I run vaccumdb on master do this
blob will gone. Well I watch one treat here, one guy with 540 gb
database with huge blob and Im little scared do I
will have the same result. Also if something happened on the master
server and it's going down do I will have all
wal files in archive directory on standby server and did postgres will
apply them auto after finish the query or
I need to do this manually.


> Greetings,
>
> Andres

Greetings,
Condor

В списке pgsql-general по дате отправления:

Предыдущее
От: Guillaume Lelarge
Дата:
Сообщение: Re: Starnge things with big datas...
Следующее
От: Guillaume Lelarge
Дата:
Сообщение: Re: Starnge things with big datas...