Horton

Re: Potential G2-item cycles under serializable isolation

Kyle Kingsbury <aphyr@jepsen.io>

View thread View on postgr.es
From: Kyle Kingsbury <aphyr@jepsen.io>
To: Peter Geoghegan <pg@bowt.ie>
Cc: PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>
Date: 2020-06-01T04:29:57Z
Lists: pgsql-bugs 
On 6/1/20 12:20 AM, Peter Geoghegan wrote:
> It's confusing because the standard only requires that the isolation  > levels avoid certain read phenomena, but implementations are free to > go 
above and beyond. For example, you can ask Postgres for READ > UNCOMMITTED, but 
you'll get READ COMMITTED. (So RC, RR, and SI each > provide distinct behavior.)
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." 
This way around, Postgres "repeatable read" actually gives you behavior that 
violates repeatable read! But I understand the pragmatic rationale of "we need 3 
levels, and this is the closest mapping we could get to the ANSI SQL names". :)

--Kyle

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.