Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 10.2 \(3259\))
Subject: Re: logical decoding of two-phase transactions
From: Stas Kelvich <s.kelvich@postgrespro.ru>
In-Reply-To: 
 <CAMsr+YH0aRgnk8XwyYypV1E10DNXwgGpWTLq_P2MGjACpiwMpA@mail.gmail.com>
Date: Thu, 26 Jan 2017 10:43:58 +0300
Cc: Simon Riggs <simon@2ndquadrant.com>,
 PostgreSQL mailing lists <pgsql-hackers@postgresql.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <89073436-925E-4299-A9AF-88D04E55C231@postgrespro.ru>
References: <02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru>
 <CANP8+jJGRpczm8L=LQCzuiEG3qHbY1cQB+T4A-cct7xgFmjj9g@mail.gmail.com>
 <CANP8+jJ+=hwLcDSNAqBmZwHh0Pb46SERQsBC1i4VCpkgkL9Jjg@mail.gmail.com>
 <D11D3072-6A72-439C-8B62-1D2628134A60@postgrespro.ru>
 <CANP8+j+xxYB5GoFebq1NGm17TEAaZ=JvM-Ok-Hp+n6De3pseeA@mail.gmail.com>
 <DED204B8-0877-4BAB-B2FE-0F450BD56186@postgrespro.ru>
 <CAMsr+YH0aRgnk8XwyYypV1E10DNXwgGpWTLq_P2MGjACpiwMpA@mail.gmail.com>
To: Craig Ringer <craig@2ndquadrant.com>
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

>>=20
>> Yes, that=E2=80=99s also possible but seems to be less flexible restrict=
ing us to some
>> specific GID format.
>>=20
>> Anyway, I can measure WAL space overhead introduced by the GID=E2=80=99s=
 inside commit records
>> to know exactly what will be the cost of such approach.
>=20
> Stas,
>=20
> Have you had a chance to look at this further?

Generally i=E2=80=99m okay with Simon=E2=80=99s approach and will send send=
 updated patch. Anyway want to
perform some test to estimate how much disk space is actually wasted by ext=
ra WAL records.

> I think the approach of storing just the xid and fetching the GID
> during logical decoding of the PREPARE TRANSACTION is probably the
> best way forward, per my prior mail.

I don=E2=80=99t think that=E2=80=99s possible in this way. If we will not p=
ut GID in commit record, than by the time
when logical decoding will happened transaction will be already committed/a=
borted and there will
be no easy way to get that GID. I thought about several possibilities:

* Tracking xid/gid map in memory also doesn=E2=80=99t help much =E2=80=94 i=
f server reboots between prepare=20
and commit we=E2=80=99ll lose that mapping.=20
* We can provide some hooks on prepared tx recovery during startup, but tha=
t approach also fails
if reboot happened between commit and decoding of that commit.
* Logical messages are WAL-logged, but they don=E2=80=99t have any redo fun=
ction so don=E2=80=99t helps much.

So to support user-accessible 2PC over replication based on 2PC decoding we=
 should invent
something more nasty like writing them into a table.

> That should eliminate Simon's
> objection re the cost of tracking GIDs and still let us have access to
> them when we want them, which is the best of both worlds really.

Having 2PC decoding in core is a good thing anyway even without GID trackin=
g =3D)

--
Stas Kelvich
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company


--=20
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers