Reallocation without MySQL cluster (was: Allocating MySQL cluster nodes for MogileFS database)

[Jared Klett]

hi Arthur,

	you're quite right. I worry a lot about downtime, which makes me
gravitate towards MySQL cluster. replication with failover should do
nicely for now - as you say, until we outgrow it. :)

	thanks again for the advice on memory for the file servers. I
just ordered 8 x 2 GB DIMMs. :)

	so now my planned setup will look like this:

file1 - mogstored, mogilefsd
file2 - mogstored, mogilefsd
file3 - mogstored, mogilefsd
file4 - mogstored, mogilefsd
db1   - mysqld (master)
db2   - mysqld (slave)

	which I may want to change yet again, since I'm now exploring
the different options on how to serve files out of MogileFS via HTTP -
either Apache 2 or Lighttpd.

	but I have more research to do on those before I start asking
questions... unless someone has a recommended method. :)

cheers,

- Jared

-----Original Message-----
From: mogilefs-bounces at lists.danga.com
[mailto:mogilefs-bounces at lists.danga.com] On Behalf Of Arthur Bebak
Sent: Wednesday, February 07, 2007 3:09 PM
To: mogilefs
Subject: Re: Allocating MySQL cluster nodes for MogileFS database

Jared Klett wrote:
> hi Jonathan,
> 
> 	I had the same thought, initially, but as it turns out in the 
> latest/greatest (gulp) version of MySQL cluster it's possible to use 
> disk storage on your data nodes:
> 
> http://dev.mysql.com/doc/refman/5.1/en/mysql-cluster-disk-data.html

Honestly, from an operations point of view I'm not convinced that the
complexity of running a cluster is really needed. You're probably better
off with a simple master/slave setup with some automatic failover. When
your DB traffic outgrows your single DB box you use your application to
balance between multiple database boxes.

For example, store the data for users A-L in one M/S pair, and users M-Z
in the other pair, or split independent tables between different M/S
pairs - you get the idea.
That will take you a long way up the performance curve with little
complexity.
At least until you get to be as big as LiveJournal :)

--
Arthur Bebak
abebak at fabrikinc.com