Re: [siesta-dev] multimaster replication

[prev] [thread] [next] [lurker] [Date index for 2006/05/22]

From: morty
Subject: Re: [siesta-dev] multimaster replication
Date: 17:52 on 22 May 2006
On Mon, May 22, 2006 at 05:02:20PM +0100, Richard Clamp wrote:

> Unlikely, if only because we use auto-incrementing ints as primary
> keys everywhere, and from what I remember clustered databases aren't
> terribly happy when you do that.

The clustered DB I have in mind deals with this by assigning
auto-incrementing IDs from non-colliding spaces.  E.g. in a two-server
setup, server0 assigns evens, server1 assigns odds.  So unless you
need monotonically increasing IDs, you don't have a problem.

Other potential problem places are local file stores, and server
processes that don't lock rows while working.


- Morty
There's stuff above here

Generated at 14:00 on 12 Aug 2006 by mariachi 0.52