Re: TOAST behavior in 8.3 and 8.4 - Mailing list pgsql-admin

From Lewis Kapell
Subject Re: TOAST behavior in 8.3 and 8.4
Date
Msg-id 4BC4C592.7020400@setonhome.org
Whole thread Raw
In response to Re: TOAST behavior in 8.3 and 8.4  ("Kevin Grittner" <Kevin.Grittner@wicourts.gov>)
Responses Re: TOAST behavior in 8.3 and 8.4
List pgsql-admin
Thanks for the clarification.

In tuptoaster.h, just above the declaration of TOAST_TUPLES_PER_PAGE,
there is a comment which begins: "while these can be modified without
initdb..."

Does this mean that if I reduce the value of TOAST_TUPLE_THRESHOLD and
rebuild, I can use an existing database cluster without having to run
initdb again?


On 4/13/2010 3:02 PM, Kevin Grittner wrote:
> Lewis Kapell<lkapell@setonhome.org>  wrote:
>
>> I don't understand what that '256 bytes' refers to.  That is a far
>> cry from 2kb.  I would be grateful if anyone can fill in the
>> evident gap in my knowledge here.
>
> It doesn't try to compress anything unless the tuple (row instance)
> as a whole is above TOAST_TUPLE_THRESHOLD.  In trying to reduce the
> tuple size, it won't consider compressing column values below a
> certain size.  That is where the 256 versus 32 bytes comes in.
>
> -Kevin


pgsql-admin by date:

Previous
From: Alvaro Herrera
Date:
Subject: Re: TOAST behavior in 8.3 and 8.4
Next
From: "Kevin Grittner"
Date:
Subject: Re: TOAST behavior in 8.3 and 8.4