Notes from MonadLUG, 9-Oct-2008, Patrick Galbraith and MySQL Replication

[Lloyd Kvam]

On Mon, 2008-10-13 at 08:43 -0400, Ted Roche wrote:
> There were lots of related discussions and side conversations, too. An
> intriguing thread involved "blackhole" data storage engines, where the
> data actually never is written to disk, but the engine exists purely
> for posting log entries, which can then be replicated. Wow.

I think I've been doing something along those lines.  I would have liked
to attend.  A MySQL "slave" points to a single master, so a graph of
your replication servers will either be a tree or a cycle (possibly with
branches).  You need a cycle if there are multiple masters that should
all contain each others inserts and updates.  

I have a client running multiple masters.  Since a cycle is vulnerable
to individual failures and delays, I have what I called a "collector"
that takes log files from the individual masters and consolidates them
into a combined log which acts as a single master.  I think that is the
processing you describe.

I'm certainly interested in sharing code if that would help others
dealing with a similar setup.

-- 
Lloyd Kvam
Venix Corp
DLSLUG/GNHLUG library
http://dlslug.org/library.html
http://www.librarything.com/catalog/dlslug
http://www.librarything.com/rsshtml/recent/dlslug
http://www.librarything.com/rss/recent/dlslug