Content-Type: text/plain;
	charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.80.23.2.2\))
Subject: Re: Confusions regards serializable transaction
From: Hannah Huang <hannah.huang.y@gmail.com>
In-Reply-To: 
 <CAH2-WzkdNKZ+fZ9r+98ZbYR7iV4x7NJr6sh6og=R3xge5fUVvQ@mail.gmail.com>
Date: Fri, 4 Sep 2020 16:05:50 +1000
Cc: pgsql-novice@lists.postgresql.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <31DA4198-1B8E-4999-9A19-49AD87F36BDB@gmail.com>
References: <AD1ECF70-71F3-404B-84A4-A074D7F08ED3@gmail.com>
 <CAH2-WzkdNKZ+fZ9r+98ZbYR7iV4x7NJr6sh6og=R3xge5fUVvQ@mail.gmail.com>
To: Peter Geoghegan <pg@bowt.ie>
Precedence: bulk


> On 4 Sep 2020, at 2:21 pm, Peter Geoghegan <pg@bowt.ie> wrote:
>=20
> On Thu, Sep 3, 2020 at 5:20 PM Hannah Huang <hannah.huang.y@gmail.com> =
wrote:
>> I=E2=80=99m confused about what will trigger the serializable error. =
My test environment is PG v12.
>>=20
>> This is how you can produce the test table:
>=20
> Serializable isolation level promises to emulate serial transaction
> execution for all committed transactions. I believe that your example
> does not show behavior that breaks that specific promise.
>=20
> --=20
> Peter Geoghegan

Hi Peter,
Thank you very much for replying my question.

My confusion is actually:

Transaction B updated the n value of  id 2, while transaction A needs to =
update n value for id 1 referencing the n value of id 2.

If the transaction is executed in a serialized way, then the n value of =
id 2 will be changed in transaction B first, and the change will be =
reflected in the value of id 1in transaction A. However, the two =
transactions are not executed concurrently - transaction A executed =
successfully without seeing changes made in transaction B. Is that a =
break of serializable isolation level?

Thanks,
Hannah=