Jim Nasby wrote:
> On Mar 7, 2006, at 2:58 PM, Louis Gonzales wrote:
>
>> Based on:
>> Postgres based Replication Projects
>>
>> PG Replication
>> Postgres-R: Dr. Kemme's Site, Paper, Publications, Replication Work,
>> The Horus Project and Emsemble
>> DRAGON: Database Replication based on Group Communication
>> The Slony-1 Project
>> PGCluster
>> DBBalancer
>> PostgreSQL Replicator: Tutorial, Paper, Techdocs
>> eRServer: Techdocs
>> DBMirror: Will be in the 7.3 contrib directory.
>> Usogres: Techdocs
>> From: http://gborg.postgresql.org/project/pgreplication/
>> genpage.php?replication_research
>> I was under the impression that "Postgres-R" - I misspelled it as
>> "postgresql-R" - is a database clustering software. I downloaded
>> the sources for this and it's supposed to work with "spread." I
>> apologize if I'm mistaken?
>
>
> Well, AfAIK it's quasi-clustering. I'd describe it as a mix between
> sync and async multi-master replication. Its concepts are the basis
> behind Slony-II AFAIK.
>
>> I will certainly take a look into Slony-1, as you suggest. Did you
>> have any good links to this? Have you gone through an install on
>> Solaris 9 w/Postgresl v8.0?
>>
>> WAN != LAN, where my database server is not in the same city as the
>> target system to cluster with; for my worse case scenario, could be
>> from east coast to west coast.
>
>
> The issues you'll run into are that in a high-load scenario, the
> latency on that link could become a real issue, as could bandwidth
> (actually, bandwidth is probably the bigger issue). The other
> downside is that slony is decidedly async; if you have to fail over
> you're probably going to lose some recently committed data. As long
> as that's not an issue then it's your best bet. If it is an issue,
> you'll probably be stuck rolling your own solution using something
> like 2PC.
Actually, one more question. Upon further reading, the documentation
stated that any modification done on the master, will have to be
committed to all cluster participants, otherwise it will not commit? Is
it still possible to lose any committed data, with that being true?
As a supplement, I was thinking about using some sort of PITR, in
conjuction with Slony-I to ensure all bases are covered.
>
> You should hop on the slony mailing list if you have any questions
> about it.
> --
> Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com
> Pervasive Software http://pervasive.com work: 512-231-6117
> vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461
>