On Nov 22, 2006, at 11:00 AM, Shane Ambler wrote:
> Simon Riggs wrote:
>> On Wed, 2006-11-22 at 10:44 -0300, Alvaro Herrera wrote:
>>> Simon Riggs wrote:
>>>> Anyway, I'm very interested in getting Synchronous Replication
>>>> into PostgreSQL 8.3. Can I gauge your interest in making Mammoth
>>>> Replicator
>>>> Open Source to assist with that project? I'd be very happy to
>>>> work with
>>>> you in an open manner on that.
>>> Mammoth Replicator is not synchronous anyway ...
>> That's a shame. I thought we might be able get a head start in
>> that way.
>
> Have you looked at PGCluster? - Synchronous multi master replication
The downside to pgCluster (and the synchronous pgpool replication as
well) is that it's statement-based. That means something like
INSERT INTO table VALUES( ..., now(), ... );
Doesn't work right at all. Continuent has statement-based replication
that handles common cases (like now() and random()), but it doesn't
handle everything (and it's limited to JDBC connections).
I'll agree with Josh Drake that the demand for synchronous
replication isn't as great as the hype (or the demand for async), but
it definitely does exist.
Regarding putting stuff in core, ISTR a group was setup on pgFoundry
to discuss what features could be added to core to help various
replication systems (such as being able to put triggers on DDL), but
I don't know if anything every came of it.
--
Jim Nasby jim.nasby@enterprisedb.com
EnterpriseDB http://enterprisedb.com 512.569.9461 (cell)