Horton

Re: Potential G2-item cycles under serializable isolation

Peter Geoghegan <pg@bowt.ie>

View thread View on postgr.es
From: Peter Geoghegan <pg@bowt.ie>
To: Kyle Kingsbury <aphyr@jepsen.io>
Cc: PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>
Date: 2020-06-01T04:47:13Z
Lists: pgsql-bugs 
On Sun, May 31, 2020 at 9:29 PM Kyle Kingsbury <aphyr@jepsen.io> wrote:
> Right, right. I was thinking "Oh, repeatable read is incomparable with snapshot, so it must be that read committed is snapshot, and repeatable is serializable."

Right.

We used to call snapshot isolation (i.e., the behavior we now provide
under RR mode) SERIALIZABLE, which was wrong (still is in Oracle).
This was how Postgres worked before the SSI feature was added back in
2011. SSI became the new SERIALIZABLE at that time. Ordinary snapshot
isolation was "demoted" to being called RR mode.

-- 
Peter Geoghegan

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.