Re: [HACKERS] Re: PD_ALL_VISIBLE flag was incorrectly set happend during repeatable vacuum

On Tue, Mar 1, 2011 at 10:43 AM, David Christensen <david(at)endpoint(dot)com> wrote:
> Was this cluster upgraded to 8.4.4 from 8.4.0?  It sounds to me like a known bug in 8.4.0 which was fixed by this commit:
>
> commit 7fc7a7c4d082bfbd579f49e92b046dd51f1faf5f
> Author: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
> Date:   Mon Aug 24 02:18:32 2009 +0000
>
>    Fix a violation of WAL coding rules in the recent patch to include an
>    "all tuples visible" flag in heap page headers.  The flag update *must*
>    be applied before calling XLogInsert, but heap_update and the tuple
>    moving routines in VACUUM FULL were ignoring this rule.  A crash and
>    replay could therefore leave the flag incorrectly set, causing rows
>    to appear visible in seqscans when they should not be.  This might explain
>    recent reports of data corruption from Jeff Ross and others.
>
>    In passing, do a bit of editorialization on comments in visibilitymap.c.
>
> oy:postgresql machack$ git describe --tag 7fc7a7c4d082bfbd579f49e92b046dd51f1faf5f
> REL8_4_0-190-g7fc7a7c
>
> If the flag got twiddled while running as 8.4.0, the incorrect PD_ALL_VISIBLE flag would (obviously) not be fixed by the upgrade to 8.4.4.  (Is this a separate issue?)

Yes, it's a different issue. I observed it on 8.4.2.

Regards,

--
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center