Re: Patch for ginCombineData

From: Alexander Korotkov <a(dot)korotkov(at)postgrespro(dot)ru>
To: Robert Abraham <robert(dot)abraham86(at)googlemail(dot)com>
Cc: pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Patch for ginCombineData
Date: 2015-08-05 11:11:01
Message-ID: CAPpHfdsOA6Q-N=X6OjcRAd1TqkB8aKTn5hZ=8wVL3V=aYo=SYQ@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi!

On Wed, Aug 5, 2015 at 1:17 PM, Robert Abraham <
robert(dot)abraham86(at)googlemail(dot)com> wrote:

> we are using gin indexes on big tables. these tables happen to have
> several billion rows.
> the index creation fails in ginCombineData in src/backend/access/ginbulk.c
> because repalloc is limited to 1 gb.
> this limitation makes no sense in this context (compare comments in
> src/include/utils/memutils.h).
> To overcome this limitation on tables with lots of rows repalloc_huge is
> used.
> The test suite still succeeds on my machine.
> Find the patch attached,
>

Thank you for notice and for the patch!
You should have maintenance_work_mem > 1gb and some very frequent entry so
that it's posting list exceeds 1 gb itself.
These circumstances shouldn't be very rare in modern systems. I think it
could be backpatched.

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

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Andres Freund 2015-08-05 11:24:04 Re: Reduce ProcArrayLock contention
Previous Message Nikolay Shaplov 2015-08-05 10:52:27 Re: pageinspect patch, for showing tuple data