[prev in list] [next in list] [prev in thread] [next in thread] 

List:       postgresql-general
Subject:    Re: [HACKERS] WAL format changes break the suppression of do-nothing checkpoints.
From:       Heikki Linnakangas <hlinnaka () iki ! fi>
Date:       2015-03-31 13:19:58
Message-ID: 551A9EFE.5080702 () iki ! fi
[Download RAW message or body]

On 03/30/2015 09:01 PM, Jeff Janes wrote:
> commit 2c03216d831160bedd72d45f7 has invalidated the part of the docs
> saying "If no WAL has been written since the previous checkpoint, new
> checkpoints will be skipped even if checkpoint_timeout has passed",
> presumably by accident.
>
> It seems that this part is no longer true when it should be true:
>
>          if (curInsert == ControlFile->checkPoint +
>              MAXALIGN(SizeOfXLogRecord + sizeof(CheckPoint))
>
> MAXALIGN(SizeOfXLogRecord + sizeof(CheckPoint) is now 96, but the amount by
> which curInsert gets advanced is still 104, like it was before the commit.

Hmm. Wasn't this a bit broken before too, when the checkpoint record 
crosses a page boundary?

Instead of trying to calculate where the checkpoint record ends, I think 
we could check that the prev-pointer points to the last checkpoint record.

- Heikki



-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
[prev in list] [next in list] [prev in thread] [next in thread]