Horton

Re: Potential G2-item cycles under serializable isolation

Daniel Verite <daniel@manitou-mail.org>

View thread View on postgr.es
From: "Daniel Verite" <daniel@manitou-mail.org>
To: "Kyle Kingsbury" <aphyr@jepsen.io>
Cc: "PostgreSQL mailing lists" <pgsql-bugs@lists.postgresql.org>
Date: 2020-06-03T11:51:19Z
Lists: pgsql-bugs 
	Kyle Kingsbury wrote:

> SQL 92 says that the difference between repeatable read and serializable is 
> *exactly* the prohibition of P3 ("phantom").

You're probably refering to Table-9 in SQL-92, showing
that P3 can happen under Repeatable Read and cannot happen
under Serializable.

But it doesn't say that the *only* difference between RR and Serializable
is avoiding P3. When defining P1, P2, P3, it doesn't explicitly say
that these are the only anomalies that can arise from concurrency.
The PG doc refers to the other cases as "serialization anomalies".

Compared to the manual, https://wiki.postgresql.org/wiki/Serializable
has more in-depth explanations on the specifics of PG serializable
implementation.

Best regards,
-- 
Daniel Vérité
PostgreSQL-powered mailer: http://www.manitou-mail.org
Twitter: @DanielVerite

Commits

  1. Doc: Add references for SI and SSI.

  2. Improve comments for [Heap]CheckForSerializableConflictOut().

  3. Avoid update conflict out serialization anomalies.

  4. Implement genuine serializable isolation level.