[Pljava-dev] Developing a db-clustering system using PL/Java

Hello,

I'm considering building a clustering solution for Postgres using PL/Java. Before embarking on this ambitious task, I'd like to ask your advice.

Any feedback on the following Qs would be greatly appreciated!
Also, let me know if this project would be of interest to you.

1. What state is PL/Java in and is it actively maintained?

2. Am I re-inventing the wheel?
My immediate requirements are to do _partial_ master-master replication.
I've checked out several existing solutions such as Slony-I, but they seem to offer all-or-nothing replication.

3. The architecture I'm considering is to use triggers to maintain a table of local edits, with edits then sent across the cluster, using embedded
Jetty servers to handle the low-level networking via http.
Do you see any objections to this?
Is there anything about how PL/Java works that might cause problems?
Is there a better architecture I should be considering?

Thank you for your time & any advice you have.

Best regards,
-Daniel

--------------------------------------------
Dr Daniel Winterstein
tel: 0772 5172 612 @winterstein
http://winterwell.com http://sodash.com
Registered in Scotland, company no. SC342991

Daniel Winterstein <daniel at winterwell.com> writes:

> Hello,
>
> I'm considering building a clustering solution for Postgres using
> PL/Java. Before embarking on this ambitious task, I'd like to ask
> your advice.

> 1. What state is PL/Java in and is it actively maintained?

It is actively maintained. Though slow on development.

> 2. Am I re-inventing the wheel?

> My immediate requirements are to do _partial_ master-master
> replication. I've checked out several existing solutions such as
> Slony-I, but they seem to offer all-or-nothing replication.

Below, you seem to be describing Londiste. Have you looked at that
project?

> 3. The architecture I'm considering is to use triggers to maintain a
> table of local edits, with edits then sent across the cluster, using
> embedded Jetty servers to handle the low-level networking via http.
> Do you see any objections to this?

> Is there anything about how PL/Java works that might cause problems?
> Is there a better architecture I should be considering?

There shouldn't be anything in PL/Java that should hamper you.
Compared to Londiste, it seems like a duplication of effort though.

What part of Londiste does not suit your purpose?

--
Johann Oskarsson http://www.2ndquadrant.com/ |[]
PostgreSQL Development, 24x7 Support, Training and Services --+--
|
Blog: http://my.opera.com/myrkraverk/blog/

Hi Johann,

Thanks for your email.
From http://wiki.postgresql.org/wiki/Londiste_Tutorial:

> Londiste is a single master to several slaves solutions.

So not partial master-master replication. However the underlying PgQ system does look interesting. It might work to have 1 queue per server, and some
SQL-level logic to manage the replication. I will take a closer look at it.

Thanks &
Best regards,
- Daniel

On 03/11/11 17:36, Johann 'Myrkraverk' Oskarsson wrote:
> 2. Am I re-inventing the wheel?
>> My immediate requirements are to do _partial_ master-master
>> replication. I've checked out several existing solutions such as
>> Slony-I, but they seem to offer all-or-nothing replication.
> Below, you seem to be describing Londiste. Have you looked at that
> project?

--
--------------------------------------------
Dr Daniel Winterstein
tel: 0772 5172 612 @winterstein
http://winterwell.com http://sodash.com
Registered in Scotland, company no. SC342991