Re: [HACKERS] GSoC 2017 : Patch for predicate locking in Gist index

Поиск
Список
Период
Сортировка
От Alexander Korotkov
Тема Re: [HACKERS] GSoC 2017 : Patch for predicate locking in Gist index
Дата
Msg-id CAPpHfdsJ9JfgDgTKzN7s-Qt84jYrAGHidTVMcFFh6t+1AS13_Q@mail.gmail.com
обсуждение исходный текст
Ответ на Re: [HACKERS] GSoC 2017 : Patch for predicate locking in Gist index  (Andrey Borodin <x4mmm@yandex-team.ru>)
Ответы Re: [HACKERS] GSoC 2017 : Patch for predicate locking in Gist index
Список pgsql-hackers
Дерево обсуждения
Hi!

On Tue, Mar 27, 2018 at 11:16 AM, Andrey Borodin <x4mmm@yandex-team.ru> wrote:
> 27 марта 2018 г., в 12:53, Teodor Sigaev <teodor@sigaev.ru> написал(а):
>
> I have a question: why do not CheckForSerializableConflictIn() move  into begining of gistplacetopage()? Seems, it is the single function which actually changes page and all predicate locking stuff will be placed in single function...

gistplacetopage() is called from
1. Buffered build - probably harmless

Yes, harmless, but useless.
 
2. Finish split - i'm not sure about this. It seems to me that it is necessary... then your version is correct.

Yes, it's necessary, because GiST scan can end up on non-leaf page.  So, scan and modify of same non-leaf page should conflict.

Checking for serializable conflicts from buffering build seems useless overhead.  gistplacetopage()
is called from only two places: gistinserttuples() and gistbufferinginserttuples().  In order to evade
useless overhead for buffering build, I've moved CheckForSerializableConflictIn() into gistinserttuples().

Also, I find that we call PredicateLockPageSplit() for every page produced by split including
original.  That also seems to cause extra overhead.  This is why I've moved
PredicateLockPageSplit() into loop where we do assign new buffers.

------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company 
Вложения

В списке pgsql-hackers по дате отправления: