From: "R, Siva" <sivasubr@amazon.com>
To: Alexander Korotkov <a.korotkov@postgrespro.ru>,
        "sawada.mshk@gmail.com"
	<sawada.mshk@gmail.com>
CC: pgsql-hackers <pgsql-hackers@postgresql.org>
Subject: Re: Bug in ginRedoRecompress that causes opaque data on page to be
 overrun
Thread-Topic: Bug in ginRedoRecompress that causes opaque data on page to be
 overrun
Thread-Index: 
 AQHURITWziSocye1S0ygUmXh1SVXsaTgquEAgAANUlGAALPkAIAAJPwAgAAo8QCAAHJWAIAACpEY
Date: Wed, 5 Sep 2018 21:49:05 +0000
Message-ID: <1536184145340.65860@amazon.com>
References: <1536091151804.6588@amazon.com>
 <CAH2-WzneRxhYubPMh-x0L7h_1iH5kddVH5HWx2RFWKh=ELdgzw@mail.gmail.com>
 <1536101117759.63175@amazon.com>
 <CAPpHfdv1FkM3Lk7xGh1Wmo+u7xKV9GqNzLkc2mxiiX5OOqkN1Q@mail.gmail.com>
 <CAPpHfduH0P9onLnkGGMM_Pze5J47qjHFxFxouWkhYxBebq0mrQ@mail.gmail.com>
 <CAPpHfdtS2hHJBT4bOnk2FJDzE1yCvRDveAjrW_Oi3aAK1MtrGw@mail.gmail.com>,<CAPpHfduRz9qeqxVQh+2FAj+uQ040qzijYOM6Lc2o7873UzVL3w@mail.gmail.com>
In-Reply-To: 
 <CAPpHfduRz9qeqxVQh+2FAj+uQ040qzijYOM6Lc2o7873UzVL3w@mail.gmail.com>
Accept-Language: en-US
Content-Language: en-US
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
Precedence: Bulk
Precedence: bulk

On Tue, Sep 5, 2018 at 08:55 PM, Alexander Korotkov <a.korotkov@postgrespro=
.ru> wrote:=0A=
> Finally I managed to reproduce the bug.  The scenario is following.=0A=
> Underlying idea is that when insertion of multiple tuples goes to the=0A=
> beginning of the page and this insertion succeed only thanks to=0A=
> collapse of some short segments together, then this insertion wouldn't=0A=
> fit to the page if given alone.=0A=
=0A=
Sorry for the late reply.=0A=
Thank you so much for working on this and reproducing the issue!=0A=
Like you mentioned, the WAL record where we detected this problem=0A=
has future segments deleted due to compaction and written out=0A=
as an insert segment.=0A=
=0A=
> alter index test_idx set (fastupdate =3D on);=0A=
Just curious why does this help with the repro? This is related to only=0A=
using the Gin pending list vs the posting tree.=0A=
=0A=
I will try to reproduce the issue with the above workload and=0A=
also test the fix with the same and report back.=0A=
=0A=
On Wed, Sep 5, 2018 at 5:24 PM, Masahiko Sawada <sawada.mshk@gmail.com> wro=
te:=0A=
> Hmm, could you share the sequence of what kind of WAL has applied to=0A=
> the broken page? I suspect the segment list contains=0A=
> GIN_SEGMENT_REPLACE before GIN_SEGMENT_INSERT.=0A=
=0A=
These are the segment operations on the WAL in sequence:=0A=
- 1 Replace action on segment N=0A=
- 5 Insert actions after segment N - the 5th insert action is essentially=
=0A=
   replacing the last 3 remaining segments with a new one.=0A=
- 3 delete actions on the remaining segments.=0A=
=0A=
Best=0A=
Siva=0A=